Home Uncategorized Economic Evaluation and Health Policy

Uncategorized

Economic Evaluation and Health Policy

Nursing shortages are a source of feared for many health care administrators. Describe what the nursing workforce is like where you live. (Your state board of nursing website is a good place to gain information.) Next, discuss what aspects are leading to an overall nursing shortage, and what steps are being done to counteract this shortage.

Return-to-work-after-sickness-absence–a-register-based-compari_2019_Public-

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 0
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Return to work after sickness absence: a register-
based comparison of two indigenous population
groups
K. Reini*, J. Saarela

Demography Unit, Faculty of Education and Welfare Studies, �Abo Akademi University, Vaasa, Finland
a r t i c l e i n f o

Article history:

Received 20 August 2018

Received in revised form

18 January 2019

Accepted 31 January 2019

Available online 8 March 2019

Keywords:

Sickness absence

Native groups

Unemployment

Finland
* Corresponding author. �Abo Akademi Vaasa
E-mail address: kaarina.reini@abo.fi (K. R

https://doi.org/10.1016/j.puhe.2019.01.016
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objectives: The objective of the article was to analyse how Finnish and Swedish speakers in

Finland differ in health and labour market outcomes after sickness absence. Apart from

many similarities, these two population groups differ in life expectancy and union stability

and are supposed to be culturally distinct. Our analyses, therefore, help to shed light on the

interrelation between culture and health.

Study design: We monitored health and labour market-related status 3 years after the first

sickness absence.

Methods: The register-based longitudinal data covered the years 1988e2010. Multinomial

logistic regressions were used to quantify the odds of being unemployed, retired due to

disability, otherwise outside the labour force or dead, as compared with being employed.

The analyses were controlled for age, educational level, region of residence, population

density, birth region, family status, job industry, income, homeownership, time period and

time on sick leave.

Results: Unemployment after sickness absence was notably more common for Finnish

speakers than for Swedish speakers. In the fully adjusted models, the odds ratios were 1.48

(95% confidence interval [CI] 1.23e1.67) in men and 1.29 (95% CI 1.07e1.48) in women.

Disability pension, being outside the labour force and having died were also more frequent

outcomes for Finnish speakers than for Swedish speakers, although most of this variation

could be attributed to socio-economic and demographic characteristics.

Conclusions: The article illustrates that register-based analyses can be effective tools for

assessing and identifying persons with latent problems that impede their functioning in

the labour market. These findings also suggest that culturally related factors presumably

play an important role in this concern.

© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
, P.O. Box 311, 65101 VAASA, Finland.
eini).

mailto:kaarina.reini@abo.fi

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.016&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.016

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 094
Introduction

In the Nordic context, public sector incentives to help people

return to work (RTW) after sickness absence are high. In

Finland, approximately 8% of working-aged persons received

sickness allowance in 2016, and the associated costs were

774 million euros. The most common diagnostic causes of

long-term sickness absence in recent years in Finland have

been musculoskeletal diseases, mental disorders and in-

juries.1 The first two are also the main causes of disability

pension.2

Successful RTW varies by a multitude of factors such as

persons’ age, sex, education and socio-economic status and
labour market activity, the severeness of the injury or illness,

adequate RTW coordination and multidisciplinary in-

terventions.3 In general, individuals on sick leave with mental

health reasons have a higher risk of not returning to work,

although RTW expectations seem to abolish this excess risk.4

We take a novel approach to the issue of successful RTW by

comparing two groups that are supposedly culturally distinct,

Finnish speakers and Swedish speakers, in Finland. In the

country’s population register, people can be uniquely identi-
fied according to ethnolinguistic affiliation, which enables

separation of Finnish speakers and Swedish speakers. Finnish

speakers constitute approximately 90% of the country’s pop-
ulation, and Swedish speakers constitute about 5.5%. Both

groups are native and have an equal position with respect to

legislation, the same access to the social and healthcare sys-

tem, similar educational institutions and roughly equal socio-

economic position. They offer an interesting setting for com-

parisons because apart from the many similarities, they differ

in two central aspects that are, at least partly, related to

culturally related practices.5,6

First, they differ in life expectancy. At working ages, the

standardised death risk of Swedish speakers is only about 0.65

times that of Finnish speakers.7 The difference is particularly

marked for causes of death that associate with behaviours

and lifestyles, such as alcohol consumption, suicide and other

external causes. Second, the two groups differ in union sta-

bility. The divorce rate of Finnish speakers, as well as their risk

of separation from cohabiting unions, is close to twice that of

Swedish speakers.8,9 Because the difference in divorce risk

persists over very long marital durations also, it has been

claimed to be dependent on culturally related practices.10

Cultures often merge and change, but human diversity as-

sures that different lifestyles and beliefs will persist.11

Although culture does not equate solely with ethnic identity

or ethnolinguistic affiliation, we argue that Finnish speakers

and Swedish speakers in Finland, in a rough sense, represent

two culturally distinct groups. A comparison of them with

regard to successful RTW may, therefore, shed light on the

relation between culture and health.

Previous research has revealed that Finnish-speaking men

are 30% more likely to receive sickness allowance than

Swedish-speaking men, whereas the difference in women is

about 15%.12 Our aim was to examine whether these two

population groups differ with respect to successful RTW and

health-related outcomes also after all-cause sickness absence,

which is an issue that has not been investigated before. This
study is needed as it may help to gain new insights into the

development of the social security system and targeted health

promotion to more effectively address health and labour

market inequalities.
Methods

Data

The data used, with permission number TK-53-768-12, were

obtained from Statistics Finland. They consist of a random

sample of 5% of all Finnish speakers and a similarly con-

structed 20% random sample of Swedish speakers, observed

throughout the period 1988e2010. The analysis concerned

persons aged 20e56 years because sickness absence is very rare

before 20 years and disability pension begins to dominate at

older age.12

Study design

All study individuals were first-time sickness allowance

recipients. In Finland, the maximum period on sickness

allowance is approximately 1 year. However, time on the

sickness allowance can be divided across 3 consecutive cal-

endar years. For this reason and because individuals may

have several transitions after sickness absence, we observed

the situation 3 years after the first sickness absence. This

approach consequently ensured that we captured the end

state of the sickness absence process.

The analyses were geographically restricted to the south-

ern and western coastal part of Finland because it has both

Swedish-speaking and Finnish-speaking settlements and few

Swedish speakers in Finland live outside this area. The pop-

ulation in this area is generally healthier than that living

elsewhere in the country.12 In total, there were 24,107 study

individuals; of whom, 15,053 were Finnish speakers and 9054

were Swedish speakers.

Outcome and explanatory variables

The outcome variables measured whether a person was

employed, unemployed, disability pensioner, otherwise

outside the labour force or deceased 3 years after the first

sickness absence. This status was determined by the situation

at the end of the calendar year, based on information about the

main economic activity, disability pension receipt and records

from the death registry.

The explanatory variables included age, educational level,

region of residence, population density, region of birth, family

status, job industry, income quintile, homeownership, time

period and the approximated time on sick leave during the

first and subsequent calendar years with sickness allowance.

Except for the time on sick leave, the explanatory variables

were measured 1 year before the first sickness absence.

The data do not allow for the separation of different sick-

ness absence spells. However, because the total amount of

sickness allowance received in each calendar year and the

taxable income in each calendar year are known for every

individual, we could approximate the total time on sickness

https://doi.org/10.1016/j.puhe.2019.01.016

Table 1 e Variable distributions (%) by language and sex.

Variables Men Women

Finnish Swedish Finnish Swedish

Age (years)

20e24 8.2 8.9 6.4 5.9

25e29 12.9 11.3 11.8 9.3

30e34 12.2 10.0 13.0 11.5

35e39 12.8 11.3 14.3 14.2

40e44 15.2 14.3 14.5 16.2

45e49 16.1 17.1 16.8 16.5

50e54 15.9 18.9 16.5 18.5

55e56 6.6 8.2 6.7 8.0

Education

Basic 37.6 39.7 34.0 32.9

Secondary 47.0 42.4 41.5 41.2

Tertiary 15.4 17.9 24.4 25.9

Region of residence

Helsinki areaa 51.9 17.0 54.1 20.3

Rest of Uusimaa 27.7 27.6 25.2 26.9

Turku region 12.8 15.4 13.3 17.0

Pohjanmaa 7.7 39.9 7.4 35.8

Population density

Rural 6.4 40.0 6.4 35.1

Semi-urban 19.3 27.9 18.7 27.9

Urban 74.3 32.2 74.9 37.0

Region of birth

Southern Finland 53.6 57.7 47.4 60.1

Western Finland 22.1 41.2 23.8 38.6

Eastern Finland 16.7 0.9 19.1 1.1

Northern Finland 7.6 0.2 9.6 0.3

Family situation

With partner 55.8 61.9 60.5 70.8

Single 26.5 17.2 32.4 20.5

Other 17.7 20.9 7.0 8.7

Job industry

Primary industries 1.5 15.2 1.5 9.5

Manufacturing and

construction

32.9 27.1 11.6 9.1

Trade, hotel and

restaurants

11.6 11.6 18.8 13.8

Transport and

communication

10.5 11.8 5.7 6.2

Financial and

business services

9.2 5.9 11.8 7.5

Public and other services 8.4 7.8 26.7 30.5

Unemployed 14.9 10.0 10.8 9.8

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 0 95
absence. This approach was facilitated by the fact that the

sickness allowance in Finland has a nearly constant ratio

to income. The reimbursement rate is about 70% of the pre-

viously taxed income. Our measure for time on sickness

absence was, therefore, also a fairly good proxy for the

severity of the illness. Length of sickness absence is known to

be a better indicator of future reduced workability than spell

frequency,13 and most individuals are expected to have only

one medically certified sickness absence spell per calendar

year.14 Length of sickness absence has also been found to

relate positively to the reimbursement rate.15

Table 1 gives variable distributions by sex and population

group. Finnish speakers and Swedish speakers differ on region

of residence, population density, region of birth and job in-

dustry. This variation is largely explained by the fact that per-

sons in the latter group have lived in the study area for many

generations, whereas a considerable part of those in the former

group has moved into the area from other parts of Finland.16 In

terms of socio-economic and demographic characteristics, the

two groups are similar, which is consistent with previous

research and the situation in the general population.17,18

Statistical analyses

Because the outcome variable had several categories and ad-

justments were needed for effects of the explanatory vari-

ables, multinomial logistic regression models were used for

estimation. The contribution of the socio-economic and de-

mographic variables was examined in a sequential manner,

meaning that the explanatory variables were added stepwise

to the regressions. The analyses were performed separately

for men and women because the patterns of sickness absence

differ by sex.19 Combined effects of the population groups and

the explanatory variables were also assessed to evaluate if the

population group differentials differed across specific socio-

economic and demographic attributes. All regression results

are presented as odds ratios (ORs), with 95% confidence in-

tervals (CIs), between Finnish speakers and Swedish speakers.

In all regressions, the reference category of the outcome var-

iable is being employed 3 years after sickness absence. The

analyses were weighted to account for the different sampling

proportions of Finnish speakers and Swedish speakers.

Outside labour force 11.0 10.5 13.0 13.6

Income quintile

1st 18.6 17.7 17.6 19.7

2nd 14.1 14.2 20.5 23.4

3rd 12.6 16.3 23.5 23.6

4th 21.3 20.4 21.8 19.2

5th 33.5 31.4 16.6 14.1

Home

Rented or other 45.4 25.4 44.1 28.5

Home owner 54.6 74.6 55.9 71.5

Time period

1988e1992 19.0 16.9 18.4 18.8

1993e1995 22.1 22.2 21.2 22.5

1996e2001 24.9 28.7 25.9 25.6

2002e2005 15.0 14.7 15.6 15.1

2006e2008 11.5 10.3 11.2 10.3

2009e2010 7.5 7.2 7.7 7.6

Sickness absence

<2 months 66.7 71.4 71.2 72.7 (continued on next page) Results Swedish speakers, both men and women, were more often employed after sickness absence than Finnish speakers, whereas Finnish speakers were more likely to be unem- ployed or on disability pension than Swedish speakers (Table 2). For men, there were population-group differences also in the proportion of persons outside the labour force for any other reason than disability pension, and in the pro- portion having died, whereas in women there were no such differentials. According to the unadjusted regression results (Table 3, first row), Finnish-speaking men had twice the risk of being unemployed compared with Swedish-speaking men (OR 2.00, 95% CI 1.77e2.25). The ratio was lower when having controlled for all explanatory variables, but still as high as 1.48 (95% CI https://doi.org/10.1016/j.puhe.2019.01.016 https://doi.org/10.1016/j.puhe.2019.01.016 Table 1 e (continued) Variables Men Women Finnish Swedish Finnish Swedish 2e6 months 19.0 16.6 18.0 16.9 >6 months 14.3 12.0 10.8 10.4
Number of individuals 7370 4490 7683 4564

a Helsinki area: Helsinki, Espoo, Vantaa and Kauniainen.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 096
1.23e1.67). Gaining employment after sickness absence

appeared troublesome for Finnish-speaking men. Their risk of

being outside the labour force was nearly 1.5 times higher

than that for Swedish-speaking men (OR 1.41, 95% CI

1.25e1.60). Disability pension and death were also notably

more common events for Finnish-speaking men. An inclusion
Table 2 e Outcome variable distributions by language and sex

Outcome Men

Finnish

n % n

Employed 4434 60.2 31

Unemployed 1215 16.5 4

Disability pension 880 11.9 4

Outside labour force (other reason) 618 8.4 3

Deceased 223 3.0 1

Table 3 e Odds ratios of unemployment, disability pension, ot
between Finnish speakers and Swedish speakers, based on m

Sex Unemployed Disability pensio

Men

No controls 2.00 (1.77e2.25) 1.41 (1.25e1.60)

þ age 2.00 (1.78e2.26) 1.58 (1.40e1.80)
þ education 2.03 (1.81e2.29) 1.61 (1.42e1.83)
þ region of residence 1.79 (1.50e1.95) 1.51 (1.27e1.68)
þ population density 1.73 (1.39e1.83) 1.34 (1.12e1.51)
þ region of birth 1.66 (1.34e1.78) 1.24 (1.03e1.42)
þ family status 1.62 (1.30e1.74) 1.22 (1.02e1.41)
þ job industry 1.43 (1.19e1.61) 1.13 (0.97e1.34)
þ income 1.47 (1.21e1.64) 1.15 (0.98e1.35)
þ home owner 1.45 (1.21e1.63) 1.15 (0.98e1.35)
þ time period 1.47 (1.23e1.67) 1.12 (0.97e1.34)
þ length of sickness absence 1.48 (1.23e1.67) 1.19 (1.01e1.47)
Women

No controls 1.48 (1.30e1.68) 1.22 (1.07e1.38)

þ age 1.49 (1.31e1.69) 1.32 (1.15e1.49)
þ education 1.48 (1.30e1.68) 1.32 (1.15e1.49)
þ region of residence 1.62 (1.37e1.82) 1.27 (1.07e1.42)
þ population density 1.47 (1.23e1.64) 1.19 (0.99e1.33)
þ region of birth 1.37 (1.12e1.53) 1.16 (0.96e1.32)
þ family status 1.35 (1.11e1.52) 1.13 (0.95e1.31)
þ job industry 1.26 (1.04e1.43) 1.09 (0.92e1.27)
þ income 1.25 (1.03e1.42) 1.09 (0.92e1.27)
þ home owner 1.24 (1.03e1.42) 1.09 (0.92e1.27)
þ time period 1.28 (1.06e1.47) 1.09 (0.93e1.28)
þ length of sickness absence 1.29 (1.07e1.48) 1.13 (0.95e1.37)
The reference category for the outcome variable is employed. Odds ratio

intervals in parentheses).
of region of residence and job industry somewhat evened the

differences between the population groups.

The results for women mimicked those for men, although

the population group differentials were less emphasised. The

greatest difference between the population groups in women

was observed in the odds of unemployment (OR 1.48, 95% CI

1.30e1.68). When all explanatory variables were added, the

odds ratio was reduced to 1.29 (95% CI 1.07e1.48). The odds

of disability pension were also higher for Finnish-speaking

women than for Swedish-speaking women (OR 1.22, 95% CI

1.07e1.38), although the difference was statistically not sig-

nificant once age, education, region of residence and popula-

tion density had been controlled for. In contrast to men, there

was no difference between Finnish-speaking women and

Swedish-speaking women with regard to being outside the

labour force and death.
.

Women

Swedish Finnish Swedish

% n % n %

98 71.2 4870 63.4 3105 68.0

39 9.8 906 11.8 394 8.6

50 10.0 801 10.4 421 9.2

02 6.7 990 12.9 576 12.6

01 2.2 116 1.5 68 1.5

her reason for being outside the labour force and mortality
ultinomial logistic regression models.

n Other reason for outside labour force Deceased

1.46 (1.27e1.69) 1.58 (1.25e2.00)

1.50 (1.30e1.74) 1.70 (1.34e2.17)

1.51 (1.30e1.75) 1.73 (1.36e2.21)

1.24 (1.04e1.44) 1.44 (1.03e1.75)

1.16 (0.93e1.31) 1.31 (0.90e1.57)

1.11 (0.88e1.26) 1.21 (0.83e1.49)

1.09 (0.86e1.24) 1.18 (0.81e1.46)

1.03 (0.85e1.22) 1.06 (0.76e1.39)

1.04 (0.85e1.23) 1.08 (0.77e1.40)

1.03 (0.85e1.22) 1.07 (0.77e1.39)

1.03 (0.85e1.22) 1.04 (0.76e1.37)

1.05 (0.85e1.24) 1.04 (0.75e1.37)

1.11 (0.99e1.24) 1.10 (0.82e1.49)

1.04 (0.92e1.16) 1.19 (0.87e1.61)

1.03 (0.92e1.16) 1.19 (0.87e1.60)

1.03 (0.90e1.16) 1.03 (0.72e1.41)

1.03 (0.88e1.14) 1.07 (0.71e1.44)

1.02 (0.85e1.12) 0.95 (0.63e1.35)

1.02 (0.85e1.12) 0.93 (0.61e1.31)

1.01 (0.84e1.12) 0.92 (0.61e1.32)

1.00 (0.83e1.11) 0.93 (0.61e1.32)

1.01 (0.85e1.13) 0.94 (0.62e1.35)

1.03 (0.86e1.15) 0.96 (0.64e1.38)

s in bold are statistically significant at the 5% level (95% confidence

https://doi.org/10.1016/j.puhe.2019.01.016

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 0 97
To inspect the robustness and validity of our approach, we

expanded and restricted the data (all results are not shown,

but available on request). Estimates remained largely the

same when we removed the geographic restriction to the

southern and western coastal areas and performed analyses

on the whole country. The findings were also robust when the

variables for region of residence and population density were

substituted with a variable representing smaller geographical

areas, based on each individual’s subregion of residence. We
also ran the regressions only for persons who were employed

1 year before sickness absence. The results were similar,

although for men, the population group difference in the odds

of being unemployed was slightly smaller, and when all

explanatory variables were included, no difference in the odds

of disability pension was observed. For women, no difference

in unemployment was observed in the fully adjusted model.

Because the region of residence and job industry evened

the population group difference in outcomes after sickness

absence, additional robustness checks were performed to

examine how they interacted with the population group vari-

able. Joint effects, together with the main effects of the other

variables, were therefore estimated. Length of sickness

absence was also examined on this account as it is known to

play an important role in RTW.20 Finnish-speaking men were

found to have higher odds for unemployment than Swedish-

speaking men in all study regions, whereas in women, the

population group difference was statistically not significant in

the Helsinki area and rest of Uusimaa (Table 4). For the other

outcomes, no clear and significant pattern was found. With

regard to job industry, we observed differences predominantly

for unemployment in the categories manufacturing, trade,

public services and if having been unemployed or outside the

labour force before sickness absence (Table 5). For men, there

was also a population group difference for being outside the

labour force, if previously employed in the primary industries

or if having been unemployed before sickness absence. For

women, population group differentials by job industry were

more modest and primarily linked to having been outside the

labour force before sickness absence. Length of sickness

absence displayed significant associations for unemployment

only (Table 6). There was a marked population group difference
Table 4 e Odds ratios of unemployment, disability pension, ot
between Finnish speakers and Swedish speakers within each
effects between population group and region of residence.

Region Unemployed Disability pension

Men

Helsinki area 1.30 (1.01e1.68) 1.11 (0.82e1.57)

Rest of Uusimaa 1.45 (1.12e1.84) 1.28 (0.94e1.73)

Turku region 1.48 (1.03e2.12) 1.09 (0.77e1.80)

Pohjanmaa 1.74 (1.22e2.38) 1.22 (0.86e1.94)

Women

Helsinki area 1.08 (0.82e1.43) 1.10 (0.83e1.51)

Rest of Uusimaa 1.15 (0.86e1.47) 0.94 (0.72e1.33)

Turku region 1.82 (1.23e2.64) 1.45 (0.95e2.22)

Pohjanmaa 1.43 (1.03e1.98) 1.26 (0.82e1.84)

The reference category for the outcome variable is employed. Odds rat

confidence intervals in parentheses). Each row refers to a comparison of

adjusting for the effects of all other explanatory variables.
in unemployment for men if sickness absence had been at

most 6 months and for women if it had been 2e6 months.

Thus, the population groups appeared to be more equal in

terms of long-term (over 6 months) sickness spells.
Discussion

In line with previous research, we have seen that sickness

absence is not only reflective of health but also strongly

related to individual labour market problems. The specific

contribution of this article nevertheless lies in the notion that

cultural systems of values are strongly associated with health

behaviours and health outcomes.11 The need to understand

the relation between culture and health, especially the cul-

tural factors that affect health-improving behaviours, is

therefore crucial. We have, therefore, studied two equal and

native population groups, which can be considered culturally

distinct, and compared them with regard to RTW and health-

related outcomes after sickness absence. The Finnish setting

we have used is informative in an international perspective

also because when studying the cultureehealth nexus, it is

important to know what any outcomes would be without

discrimination, negative attitudes, unsuccessful integration,

and large social and economic inequalities. In many circum-

stances, this is not possible because differences in cultural

practices coincide with economic, political, legal or ethical

inequalities across population subgroups. The case of Swed-

ish speakers and Finnish speakers in Finland does not suffer

from these impediments.

We have found that, particularly not only for Finnish-

speaking men but also for Finnish-speaking women, there

was an elevated risk of becoming unemployed after sickness

absence compared with their Swedish-speaking counterparts.

Disability pension receipt after sickness absence is also

notably more common among Finnish speakers. For men, in

particular, Finnish speakers tend to move outside the labour

force and die more frequently after having been on sick leave.

Part of the explanation to the substantial difference in un-

employment incidence may be more favourable labour market

position of Swedish speakers in general, which persists even
her reason for being outside the labour force and mortality
region, based on multinomial logistic regression with joint

Other reason for outside labour force Deceased

1.08 (0.82e1.49) 0.85 (0.56e1.38)

1.30 (0.92e1.75) 0.94 (0.58e1.54)

0.68 (0.42e0.96) 1.20 (0.58e2.42)

1.00 (0.60e1.45) 1.42 (0.72e2.85)

0.93 (0.74e1.19) 1.02 (0.60e1.96)

1.00 (0.77e1.25) 0.78 (0.41e1.59)

1.00 (0.70e1.34) 0.46 (0.18e1.16)

1.20 (0.82e1.60) 1.68 (0.64e3.50)

ios in bold are statistically significant at the five percent level (95%

Finnish speakers with Swedish speakers within that specific region,

https://doi.org/10.1016/j.puhe.2019.01.016

Table 5 e Odds ratios of unemployment, disability pension, other reason for being outside the labour force and mortality
between Finnish speakers and Swedish speakers within each job industry, based on multinomial logistic regression with
joint effects between population group and job industry.

Job industry Unemployed Disability pension Other reason for outside labour force Deceased

Men

Primary industries 1.00 (0.31e2.86) 1.58 (0.60e4.53) 4.15 (1.06e15.80) 0.58 (0.07e4.64)

Manufacturing and construction 1.65 (1.24e2.05) 1.19 (0.89e1.67) 1.09 (0.75e1.53) 1.42 (0.78e2.34)

Trade, hotel and restaurants 1.60 (1.04e2.25) 1.14 (0.70e1.89) 1.09 (0.66e1.77) 0.82 (0.34e1.87)

Transport and communication 0.97 (0.60e1.45) 1.07 (0.71e1.77) 0.98 (0.53e1.81) 0.76 (0.34e1.57)

Finance and business services 1.27 (0.75e2.07) 1.03 (0.61e1.90) 1.46 (0.74e2.83) 1.00 (0.36e2.63)

Public and other services 2.02 (1.01e3.81) 1.57 (0.96e2.69) 0.81 (0.45e1.52) 0.89 (0.44e1.71)

Unemployed 1.53 (1.08e1.99) 1.20 (0.81e1.83) 1.68 (1.10e2.51) 1.76 (0.90e3.17)

Outside labour force 1.53 (1.04e2.12) 1.20 (0.77e2.00) 0.73 (0.53e0.98) 0.60 (0.29e1.27)

Women

Primary industries 1.05 (0.39e2.62) 1.35 (0.43e4.03) 2.29 (0.99e4.73) 2.47 (0.37e12.17)

Manufacturing and construction 1.22 (0.80e1.78) 1.15 (0.73e1.88) 0.80 (0.51e1.20) 0.62 (0.19e1.78)

Trade, hotel and restaurants 1.14 (0.78e1.55) 1.06 (0.68e1.69) 0.93 (0.66e1.25) 0.72 (0.33e1.52)

Transport and communication 1.07 (0.57e1.94) 2.47 (1.20e5.10) 1.42 (0.71e2.69) 0.82 (0.25e2.91)

Finance and business services 1.48 (0.90e2.38) 0.79 (0.50e1.30) 1.19 (0.73e1.89) 1.22 (0.44e3.56)

Public and other services 1.38 (0.94e1.89) 0.97 (0.73e1.30) 0.81 (0.61e1.04) 0.62 (0.36e1.12)

Unemployed 1.21 (0.87e1.61) 1.38 (0.88e2.18) 1.48 (0.99e2.06) 1.66 (0.55e4.86)

Outside labour force 1.90 (1.29e2.66) 1.34 (0.88e2.03) 1.06 (0.80e1.32) 3.78 (1.03e13.09)

The reference category for the outcome variable is employed. Odds ratios in bold are statistically significant at the five percent level (95%

confidence intervals in parentheses). Each row refers to a comparison of Finnish speakers with Swedish speakers within that specific job in-

dustry, adjusting for the effects of all other explanatory variables.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 098
when age, education and the municipality of residence are

accounted for.21,22 To some extent, Swedish speakers work in

other industries compared with Finnish speakers,23 but the

industrial distribution can explain their lower unemployment

rate only to a limited extent.22 The message from our study is

similar. We noticed that the job industry accounted for some

of the population group differences, and it was particularly

marked within certain industries for men.

Swedish speakers have previously been found to have

significantly lower unemployment rates than Finnish

speakers in Finnish-dominated areas also.22 Thus, the popu-

lation group difference in unemployment propensity persists

even after accounting for population group concentration.

Here, we observed that the region of residence accounted

for some of the differences in outcomes, especially in unem-

ployment after sickness absence. One possible explanation

may be language proficiency. Although no official data exist, it

is fairly obvious that more Swedish speakers can speak both
Table 6 e Odds ratios of unemployment, disability pension, ot
between Finnish speakers and Swedish speakers within each
multinomial logistic regression with joint effects between pop

Length of sickness absence Unemployed Disability pensi

Men

<2 months 1.42 (1.15e1.63) 1.22 (0.92e1.68) 2.0e6.0 months 1.83 (1.30e2.44) 1.25 (0.96e1.71) >6 months 1.34 (0.84e2.05) 1.12 (0.83e1.57)
Women

<2 months 1.20 (0.98e1.40) 1.01 (0.79e1.32) 2.0e6.0 months 1.93 (1.32e2.69) 1.25 (0.94e1.68) >6 months 1.23 (0.74e1.93) 1.32 (0.95e1.86)

The reference category for the outcome variable is employed. Odds rat

confidence intervals in parentheses). Each row refers to a comparison of Fi

length of sickness absence, adjusting for the effects of all other explanat
languages fluently than Finnish speakers.24 This is supposed

to favour their job search and strengthen the relative position

in the labour market. Furthermore, if Swedish speakers form a

majority in certain local areas, the language proficiency

requirements are most likely to be higher in those areas and

the relative labour market position of Finnish speakers to be

consequently weaker. We found some support for this argu-

ment because in areas where Swedish can be considered

particularly important, the population group differential in

unemployment after sickness absence was the highest.

The observed differentials may potentially reflect variation

in lifestyles and social behaviours also. The elevated mortality

risk of Finnish-speaking men is particularly marked in work-

ing ages,7,17,18,25 and for both sexes, this variation is most

pronounced for alcohol-related and external causes.7,25

Swedish speakers have lower disability retirement rates

than Finnish speakers, and likewise, with mortality, the dif-

ferences are greater in men than in women.17 It may also be
her reason for being outside the labour force and mortality
category of length of sickness absence, based on
ulation group and length of sickness absence.

on Other reason for outside labour force Deceased

1.00 (0.80e1.21) 1.14 (0.80e1.57)

1.36 (0.92e1.96) 0.73 (0.39e1.32)

0.87 (0.51e1.40) 0.95 (0.44e1.98)

0.98 (0.81e1.11) 0.96 (0.61e1.44)

1.12 (0.79e1.46) 1.10 (0.46e2.53)

1.50 (0.91e2.35) 0.88 (0.35e2.18)

ios in bold are statistically significant at the five percent level (95%

nnish speakers with Swedish speakers within that specific category of

ory variables.

https://doi.org/10.1016/j.puhe.2019.01.016

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 0 99
that Swedish speakers possess more social capital in terms of

social networks and trust towards other community members

and that behavioural differences of this kind positively affect

health.26 Social support, work motivation and a positive atti-

tude towards work are known to promote RTW after sickness

absence,27 but there has been no research about whether

Finnish speakers and Swedish speakers differ in this respect.

Studies on mortality and sickness absence have suggested

that the difference between Finnish and Swedish speakers

cannot be found in the most socially successful group, that is,

employed persons with a family.12,18 Here, we found some

support for this claim because among people who were

employed before sickness absence, the population group dif-

ferences in unemployment and disability retirement were

lower. One explanation might be that Finnish speakers and

Swedish speakers in this category are more equal with regard

to health literacy, health behaviours and social networks that

support employment.25,28

Strengths and limitations of the study

A strength of our analyses was the use of register-based data,

meaning that information on sickness absence was based on a

medically certified condition. Unfortunately, we lacked in-

formation about the specific cause, although we know that the

main causes of sickness absence in Finland are mental

disorders and musculoskeletal diseases. Periods of sickness

absence shorter than 10 working days were also omitted from

the study simply because the sickness allowance is received

after a waiting period of 10 working days. This limitation

nevertheless has the advantage that minor health problems,

such as common colds, were not recorded.

Another caveat to our findings is the inability to identify

people with ethnolinguistically mixed background. Multigen-

erational population register data are needed to study these

issues. There is some previous evidence to suggest that, in

terms of all-cause mortality in working-aged men, persons

with a mixed background may be positioned in between per-

sons with endogamous Finnish and endogamous Swedish

background.7 The present article hints towards a similar

conclusion for unemployment as the language group differ-

ence in unemployment after sickness absence was lower in

regions where intermarriage is more common (the Helsinki

area and rest of Uusimaa).

It has previously been found that RTW increased when

employers were obliged to give notice of prolonged sickness

absence to the occupational health services.29 Similarly, the

introduction of part-time sick leave enhanced RTW.30 Partial

sickness absence was introduced in Finland in 2007, but our

data cannot differentiate between partial and full sickness

absence periods. However, our results for 1988e2006 do not

differ markedly from those for 2007e2010.

In Finland, people typically receive disability pension after

being on sick leave for 1 year. Research has shown that, if the

application for disability pension is rejected, it is common to

live on different benefits and that 70% of all rejected claimants

receive some unemployment benefits.31 Such behaviours may

affect inference drawn from our data, particularly because

Finnish speakers have a higher disability retirement rate than

Swedish speakers.15,17
Conclusions

This article has illustrated that register-based analyses can

be informative tools for assessing and identifying persons

with latent problems that impede their functioning in

the labour market. We have also shown that culturally

related factors may play an important role in this concern.

The reintegration barrier of Finnish speakers seems notably

higher than that of Swedish speakers as the former appears

to encounter greater difficulties in gaining employment after

sickness absence. In general, outcomes that reflect poor

health and less advantageous labour market positions after

sickness absence are more visible for Finnish speakers than

for Swedish speakers and the difference is particularly

marked in men.

Our results are relevant for policymakers as working-age

disability benefits are one of the greatest challenges in form-

ing effective social and labour market policies. An especially

vulnerable group in the labour market studied here appears to

be Finnish-speaking men with a history of both sickness and

unemployment. Their needs for health care and rehabilitation

are likely unmet as the workability assessments are closely

connected to the access to employer-organised healthcare

services. Future studies may, therefore, tentatively examine if

policy changes, the employers’ obligation to give notice of
prolonged sickness absence and the introduction of part-time

sick leave will narrow the differences in RTW across

subgroups of the population.

In correspondence with previous research,31 our results

raise questions and concerns about sickness absence policies,

particularly about the practical processes involved. In Finland,

long sickness absence periods require several health in-

spections, but in practice, they only work adequately for per-

sons with access to healthcare services organised by the

employer.2 The interrelation of unemployment and sickness

absence demands carefully designed social security systems

that support social cohesion, evade marginalisation and avoid

having people living on unintended benefits.
Author statements

Data availability

The data used in this study are available from Statistics

Finland, but restrictions apply to their availability because

they were used under licence (permission number: TK-53-

768-12). They are consequently not publicly accessible, but

available from Statistics Finland on request, subject to service

fees.

Acknowledgements

This project was supported by funding from Svenska Litter-

aturs€allskapet i Finland and H€ogskolestiftelsen i €Osterbotten.

Ethical approval

Not required because the research involved anonymised

records and data sets.

https://doi.org/10.1016/j.puhe.2019.01.016

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 9 3 e1 0 0100
Funding

This project was supported by funding from Svenska

Litteraturs€allskapet i Finland and H€ogskolestiftelsen i
€Osterbotten.

Competing interests

None declared.
r e f e r e n c e s

1. Pekkala J, Blomgren J, Pietil€ainen O, Lahelma E, Rahkonen O.
Occupational class differences in diagnostic-specific sickness
absence: a register-based study in the Finnish population,
2005e2014. BMC Public Health 2017;17:670.

2. Laaksonen M, Blomgren J, Gould R. Sickness allowance,
rehabilitation and unemployment history of disability
retirees: a register-based study. Finnish Centre for Pensions.
Reports 05/2014. https://www.etk.fi/wp-content/uploads/
2015/10/rap_05_2014 . Accessed 13 Jun 2018.

3. Cancelliere C, Donovan J, Stochkendahl MJ, Biscardi M,
Ammendolia C, Myburgh C, et al. Factors affecting return to
work after injury or illness: best evidence synthesis of
systematic reviews. Chiropr Man Therap 2016;24:32.

4. Pedersen P, Lund T, Lindholdt L, Nohr EA, Jensen C,
Sogaard HJ, et al. Labour market trajectories following
sickness absence due to self-reported all cause morbidity e a
longitudinal study. BMC Public Health 2016. https://doi.org/
10.1186/s12889-016-3017-x.

5. Saarela J, Cederstr€om A, Rostila M. Birth order and mortality
in two ethno-linguistic groups: register-based evidence from
Finland. Soc Sci Med 2016;158:8e13.

6. Saarela J, Rostila M. Mortality after the death of a parent in
adulthood: a register-based comparison of two ethno-
linguistic groups. Eur J Public Health 2018. https://doi.org/
10.1093/eurpub/cky189.

7. Saarela J, Finn€as F. The ethno-linguistic community and
premature death: a register based study of working-aged
men in Finland. J Racial Ethn Health Disparities
2016;3:373e80.

8. Finn€as F. Social integration, heterogeneity, and divorce: the
case of the Swedish-speaking population in Finland. Acta
Sociologica 1997;40:263e77.

9. Saarela J, Finn€as F. Transitions within and from ethno-
linguistically mixed and endogamous first unions in Finland.
Acta Sociologica 2014;57:77e92.

10. Saarela J, Finn€as F. Ethno-linguistic exogamy and divorce:
does marital duration matter? Socio Focus 2018;51:279e303.

11. Napier D, Ancarno C, Butler B, Calabrese J, Chater A,
Chatterjee H, et al. Culture and health. Lancet
2014;384:1607e39.

12. Reini K, Saarela J. Differences in sickness allowance receipt
between Swedish speakers and Finnish speakers in Finland:
a register-based study. Finnish Yearb Popul Res
2017;52:43e58.

13. Stapelfeldt CM, Nielsen CV, Andersen NT, Krane L, Borg V,
Fleten N, et al. Sick leave patterns as predictors of disability
pension or long-term sick leave: a 6.75-year follow-up study
in municipal eldercare workers. BMJ Open 2014. https://
doi.org/10.1136/bmjopen-2013-003941.
14. Askildsen JE, Bratberg E, Nilsen ØA. Unemployment, labour
force composition and sickness absence: a panel data study.
Health Econ 2005;14:1087e101.

15. B€ockerman P, Kanninen O, Suoniemi I. A kink that makes you
sick: the effect of sick pay on absence. J Appl Econ
2018;33:568e79.

16. Saarela J, Finn€as F. Regional mortality variation in Finland: a
study of two population groups. Genus 2006;62:169e211.

17. Saarela J, Finn€as F. Geographical extraction and the Finnish-
Swedish health differential in Finland. Yearbk Popul Res
Finland 2005;41:61e73.

18. Saarela J, Finn€as F. Mortality inequality in two native
population groups. Popul Stud 2005;59:313e20.

19. Thorsen SV, Friborg C, Lundstrøm B, Kausto J, €Ornelius K,
Sundell T, et al. Sickness absence in the Nordic countries.
Nordic Soc Stat Comm 2015;59.

20. Høgelund J, Filges T, Jensen S. Long-term sickness absence e
what happens and how it goes?. Report No 03:20 Copenhagen:
National Institute of Social Research; 2003.

21. Saarela J, Finn€as F. Unemployment and native language: the
Finnish case. J Socio Econ 2003;32:59e80.

22. Saarela J, Finn€as F. Can the low unemployment rate of
Swedish speakers in Finland be attributed to structural
factors? J Socio Econ 2006;35:498e513.

23. Finn€as F. Swedish Finns in front of 21st century e a statistical
overview [Finlandssvenskarna inf€or 2000-talet e En statistik
€Oversikt]. Finlandssvensk Rapport No 40. Helsingfors: Svenska
Finlands Folkting; 2001.

24. Finn€as F. Bilingual families in statistical light [Tvåspråkiga
familjer i statistikens ljus]. Forskningsrapport No 37. Vasa:
Institutet f€or Finlandssvensk Samh€allsforskning; 2000.

25. Sipil€a P, Martikainen P. Language-group mortality
differentials in Finland in 1988-2004: assessment of the
contribution of cause of death, sex and age. Eur J Public Health
2009;19:492e8.

26. Nyqvist F, Finn€as F, Jakobsson G, Koskinen S. The effect of
social capital on health: the case of two language groups in
Finland. Health Place 2008;14:347e60.

27. Brouwer S, Krol B, Reneman MF, Bültmann U, Franche RL, van
der Klink JJL, et al. Behavioral determinants as predictors of
return to work after long term sickness absence: an
application of the theory of planned behavior. J Occup Rehabil
2009;19:166e74.

28. Palj€arvi T, Suominen S, Koskenvuo M, Winter T, Kauhanen J.
The differences in drinking patterns between Finnish-
speaking majority and Swedish-speaking minority in Finland.
Eur J Public Health 2009;19:278e84.

29. Halonen JI, Solovieva S, Virta LJ, Kivim€aki M, Vahtera J,
Viikari-Juntura E. Sustained return to work and work
participation after a new legislation obligating employers to
notify prolonged sickness absence. Scand J Public Health
2018;46(19_suppl):65e73.

30. Viikari-Juntura E, Virta LJ, Kausto J, Autti-R€am€o I, Martimo KP,
Laaksonen M, et al. Legislative change enabling use of early
part-time sick leave enhanced return to work and work
participation in Finland. Scand J Work Environ Health
2017;43:447e56.

31. Perhoniemi R, Blomgren J, Laaksonen M. What next after a
rejected disability pension application? Unemployment,
sickness and rehabilitation benefits and new disability
pension decisions in a four-year follow-up [Mit€a hylk€a€av€an
ty€okyvytt€omyysel€akep€a€at€oksen j€alkeen? Ty€ott€omyys-,
sairausp€aiv€araha- ja kuntoutusrahaetuudet sek€a uudet
el€akep€a€at€okset nelj€an vuoden seurannassa].
Yhteiskuntapolitiikka 2018;83:117e31.

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref1

https://www.etk.fi/wp-content/uploads/2015/10/rap_05_2014

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref3

https://doi.org/10.1186/s12889-016-3017-x

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref5

https://doi.org/10.1093/eurpub/cky189

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref7

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref8

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref9

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref10

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref11

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref12

https://doi.org/10.1136/bmjopen-2013-003941

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref14

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref15

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref16

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref17

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref18

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref19

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref20

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref21

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref22

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref23

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref24

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref25

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref26

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref27

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref28

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref29

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref30

http://refhub.elsevier.com/S0033-3506(19)30022-8/sref31

https://doi.org/10.1016/j.puhe.2019.01.016

Return to work after sickness absence: a register-based comparison of two indigenous population groups
Introduction
Methods
Data
Study design
Outcome and explanatory variables
Statistical analyses

Results
Discussion
Strengths and limitations of the study
Conclusions

Author statements
Data availability
Acknowledgements

Ethical approval
Funding
Competing interests
References

Obesity-Kuznets-curve–international-evidence_2019_Public-Health

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Obesity Kuznets curve: international evidence
N. Windarti a, S.W. Hlaing b, M. Kakinaka c,*

a Fish Quarrantine and Inspection Agency, Ministry of Marine Affairs and Fisheries, Jl. Suratmo 28, Semarang,

Central Java 50148, Indonesia
b Ministry of Planning and Finance, Office No. 26, Naypyitaw, Myanmar
c Graduate School for International Development and Cooperation, Hiroshima University. 1-5-1 Kagamiyama,

Higashi-Hiroshima, Hiroshima 739-8529, Japan
a r t i c l e i n f o

Article history:

Received 29 August 2018

Received in revised form

7 December 2018

Accepted 3 January 2019

Available online 16 February 2019

Keywords:

Obesity Kuznets curve

BMI

Economic development
* Corresponding author.
E-mail addresses: nanik.w@gmail.com (N

https://doi.org/10.1016/j.puhe.2019.01.004
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objectives: The obesity epidemic has prevailed worldwide and is currently recognized as a

global pandemic disease. Given the argument that various socio-economic features

contribute to substantial differences in obesity rates across countries, economic develop-

ment can also be considered a crucial factor of such variation. This study examines the

relationship between economic development and each of three weight-related health

statuses (rates of overweight, obesity, and morbid obesity).

Study design: This study uses panel data analysis.

Methods: Using country-level panel data of 130 countries during the period from 1975 to

2010, we apply dynamic panel data analysis to mitigate possible endogeneity problems.

Results: The main results show a clear pattern of the obesity Kuznets curve, i.e. a non-linear

relationship between a country’s income per capita and its weight-related health status, for

both males and females. For low-income countries, as incomes increase, the weight-

related health status deteriorates; thus, an increase in incomes raises the health risk. In

contrast, for high-income countries, as incomes increase, the weight-related health status

improves; thus, an increase in incomes reduces the health risk.

Conclusions: The policy implications from our analysis include the argument that a strong

initiative for health policy targeting obesity prevention is required for middle-income

countries, many of which are currently experiencing high economic growth.

The epidemic of obesity has prevailed worldwide, including

advanced and developing countries, and it is currently regar-

ded as a global pandemic disease. Obesity increases the risk of

adverse health conditions because it is associated with
. Windarti), suwahimf@iu

ic Health. Published by E
chronic medical conditions, reduced health-related quality of

life, increased health care and medication spending,1 and

decreased life expectancy;2,3 thus, the burden is not only on

the individual’s healthcare cost but also on the indirect cost
incurred by society through the reduction in productivity and

income tax.4e6 Overweight and obese individuals have
j.ac.jp (S.W. Hlaing), kakinaka@hiroshima-u.ac.jp (M. Kakinaka).

mailto:nanik.w@gmail.com

mailto:suwahimf@iuj.ac.jp

mailto:kakinaka@hiroshima-u.ac.jp

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.004&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.004

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5 27
stereotypes as unproductive, undisciplined, and unmoti-

vated,7 and they can affect the self-confidence, and even the

wages, of young adults.8 A decrease in body mass index (BMI)

in overweight and obese men and women improves health

outcomes and cost savings.9 The causal link between obesity

and environmental emissions highlights the importance of

addressing the obesity epidemic on public health and envi-

ronmental grounds.10

According to non-communicable disease (NCD) Risk Factor

Collaboration (NCD-RisC), the prevalence of obesity increased

from 3.2% to 10.8% among men and from 6.4% to 14.9% among

women during the period from 1975 to 2014. Substantial dif-

ferences in obesity rates exist across countries. The updated

report by the Organization for Economic Co-operation and

Development (OECD) in 2017 shows that the top two countries

with the highest obesity rates are the United States (38.2%)

and Mexico (32.4%), whereas the countries with the lowest

rates are Japan (3.7%) and South Korea (5.3%). Many factors

contribute to the variation in obesity across countries,

including lifestyle, cultural, and socio-economic factors,11

genetic influences, globalization,12e15 and economic

freedom.16,17 Among them, technological change and eco-

nomic factors can also be considered crucial determinants of

the obesity epidemic.18e20

Similar to the concept of the Kuznets curve of a non-linear

relationship between income inequality and development,21

health inequality and development also have a non-linear

relationship called the health Kuznets curve.22e24 The argu-

ment of the non-linear relationship has recently extended to

personal health, proxied by obesity.25 Under the obesity Kuz-

nets curve, as incomes rise, weight gain occurs since in-

dividuals can afford excess food; thus, caloric imbalance leads

to an increase in obesity rates.26 Economic development with

technological advancement creates inexpensive and delicious

foods, pushing lives with more sedentary lifestyles and less

physical activity and thus causing the obesity epidemic.20,27,28

However, given the argument that health is a normal good,

continued increases in incomes enable people to shift con-

sumption to healthier foods and to invest more in their overall

personal health, which eventually reduces obesity rates. The

case of the United States shows a negative income gradient in

BMI at the obesity threshold and that increases in income are

correlated with healthier BMI values at the tails of the BMI

distribution.29 The state- and country-level panel data also

present the non-linear relationship between income levels

and obesity rates.25,30 A spline regression analysis over 175

countries reveals that income is positively related to BMI up to

US$ 3000, with a less significant relationship beyond that

level.31

Our study also extends the income-obesity analysis across

countries with an emphasis on obesity Kuznets curve con-

texts. We use country-level 5-year interval panel data of 130

countries during the period from 1975 to 2010. Differently

from previous studies, we apply dynamic panel data analysis

to estimate empirical models with obesity rates as a depen-

dent variable and income levels as an independent variable,

allowing for a partial adjustment or persistence of obesity

rates and unobserved panel-level fixed effects. One method-

ological issue is that the models may suffer from endogeneity

problems, including dynamic effects related to persistence of
obesity rates, as a result of which the ordinary least squares

(OLS) method derives biased estimators. To mitigate such

problems, this study applies system generalized method of

moments (GMM) estimators, which include additional

moment conditions under the assumptions that there is no

autocorrelation in the idiosyncratic errors and that the panel-

level effects are uncorrelated with the first difference of the

first observation of the dependent variable.32,33

The main results show a clear pattern of the obesity Kuz-

nets curve, i.e. a non-linear relationship between health sta-

tus and income level for both males and females. In addition,

the critical value of the income level differentiating the sign of

the relationship is larger for males than for females, implying

that as incomes increase, females tend to pay more attention

to health as a normal good. The policy implications from our

analysis suggest that a strong initiative for health policy tar-

geting obesity prevention is required for middle-income

countries, many of which are currently experiencing high

economic growth. The rest of this article is organized as fol-

lows: Section Methods explains the methodology and data,

and Section Results presents the estimated results, with

important implications related to the obesity Kuznets curve.

The Discussion section concludes.
Methods

The primary objective of this study is to evaluate the obesity

Kuznets curve by estimating the inverted U-shaped relation-

ship between obesity rates and economic development, which

can be captured by real gross domestic product (GDP) per

capita over 130 countries. The obesity Kuznets curve requires

that the prevalence of the obesity epidemic rises in the early

stage of development but eventually declines after the income

level reaches some critical level. To address this issue, we

estimate the following dynamic panel data model:

Yi;t ¼ b0Yi;t�1 þ b1LGDPPCi;t þ b2LGDPPC2i;t þ gZi;t þ mi þ tt þ εi;t;

where Yi;t is the weight-related health status in country i at

time t, LGDPPCi;t is the log of the income level of country i, Zi;t
is a vector of other control variables that are expected to relate

to the weight-related health status, mi is the country fixed ef-

fects, tt is the time-specific effects, and εi;t is the error term.

The model includes the squared term of LGDPPCi;t to capture

the concavity or the non-linear relationship, and it also in-

cludes the lag of the weight-related health status because

health status has a time persistent property.

This study considers three country-level measures of

weight-related health status: age standardized rates of

overweight, obesity, and morbid obesity for adults (OVR, OBE,

and MOBE, respectively), which are taken from the Global

Health Observatory data repository from the World Health

Organization. People are classified as overweight, obese, and

morbidly obese if their BMI is more than 25, 30, and 40,

respectively. Income level is measured by the log of real GDP

per capita (LRGDPPC), which is obtained from the Penn World

Table. The inclusion of the country fixed-effects controls for

time-invariant characteristics, such as climatic conditions

and unmeasured cultural factors, and the inclusion of time-

https://doi.org/10.1016/j.puhe.2019.01.004

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 528
specific effects absorbs any time-varying differences com-

mon to all countries.

In choosing our set of control variables, we follow standard

practices as much as possible. The model includes trade

openness (TRADE), which is measured by the sum of exports

and imports divided by GDP. Trade with interaction among

countries promotes health through the transfer of knowledge,

technology, and medical supplies.34 In contrast, TRADE pro-

motes a globalized lifestyle with the increased exposure to

and consumption of imported goods, which can be recognized

as a main driver of obesity.15 In addition, we include the ratio

of the urban population (URBAN) to capture urbanization

since the previous studies suggest that people in urban areas

generally have more access to food and participate in fewer

physical activities35 and that urbanization has been driving

the rise in overweight and obesity.30

As robustness checks, the models include two additional

control variables, female labor participation (FLP) and the Gini

index (GINI). Female working hours have an impact on healthy
Table 1 e List of sample countries.

East Asia & Pacific Europe & Central
Asia

Latin America &
Caribbean

Australia Albania Antigua and Barbuda U

Brunei Darussalam Austria Argentina D

Cambodia Bulgaria Bahamas A

China Cyprus Barbados E

Fiji Denmark Belize I

Indonesia Finland Bolivia I

Japan France Brazil I

Korea, Rep. Germany Chile Jo

Lao PDR Greece Colombia K

Malaysia Hungary Costa Rica L

Mongolia Iceland Dominica M

Myanmar Ireland Dominican Republic M

New Zealand Italy Ecuador S

Philippines Netherlands El Salvador T

Thailand Norway Grenada

Vietnam Poland Guatemala

Portugal Haiti

Romania Honduras

Spain Jamaica

Sweden Mexico

Switzerland Nicaragua

Turkey Panama

United Kingdom Paraguay

Peru

St. Kitts and Nevis

St. Lucia

St. Vincent and the Grenad

Suriname

Trinidad and Tobago

Uruguay

Venezuela, RB
eating and the physical activity of females and their family

members36e38 because females are often responsible for

maintaining the health and weight of their family members.

The GINI, which is a common measure of within-country in-

come inequality, is also included because income inequality

plays a role in explaining obesity and overweight.13,39 In-

creases in income have a sizable positive impact on public

health, but the strength of the relationship is influenced by

changing levels of poverty and inequality.40 Although this

argument suggests that our model specification may not be

enough, the inclusion of GINI in the model would be helpful to

check the empirical validity of our baseline results. The data

on trade flows and urbanization are taken from the Penn

World Table and the United Nations Population Division,

respectively. The data on FLP and the GINI are obtained from

the World Development Indicators. Table 1 shows the list of

sample countries in this study. Table 2 presents the de-

scriptions of the variables used in this study. Tables 3 and 4

show the summary statistics and correlation matrix. Our
iddle East & North
Africa

South
Asia

Sub-Saharan
Africa

North
America

nited Arab Emirates Bangladesh Angola Canada

jibouti India Benin United States

lgeria Sri Lanka Burkina Faso

gypt, Arab Rep. Maldives Botswana

ran, Islamic Rep. Nepal Central African Republic

raq Pakistan Côte d’Ivoire
srael Cameroon

rdan Congo, Rep.

uwait Cabo Verde

ebanon Gabon

orocco Ghana

alta Guinea

audi Arabia Gambia

unisia Guinea-Bissau

Kenya

Liberia

Lesotho

Madagascar

Mali

Mozambique

Mauritania

Mauritius

Malawi

Namibia

Niger

Nigeria

ines Rwanda

Senegal

Sierra Leone

S~ao Tom�e and Principe

Swaziland

Chad

Togo

Tanzania

Uganda

South Africa

Zambia

Zimbabwe

https://doi.org/10.1016/j.puhe.2019.01.004

T
a
b
le

2
e

K
e
y
v
a
ri
a
b
le
s
a
n
d
d
a
ta

s
o
u
rc
e
s
.

V
a
ri
a
b
le

D
e
s
c
ri
p
ti
o
n

S
o
u
rc
e

D
e
p
e
n
d
e
n
t
v
a
ri
a
b
le
s

O
b
e
s
it
y
ra
te

(O
B
E
)

T
h
e
p
e
rc
e
n
ta
g
e
o
f
th

e
p
o
p
u
la
ti
o
n
o
f
a
g
iv
e
n
c
o
u
n
tr
y
th

a
t
is

o
b
e
s
e
in

lo
g

G
lo
b
a
l
H
e
a
lt
h
O
b
s
e
rv
a
to
ry

D
a
ta

(W
o
rl
d
H
e
a
lt
h
O
rg
a
n
iz
a
ti
o
n
)

O
v
e
rw

e
ig
h
t
ra
te

(O
V
R
)

T
h
e
p
e
rc
e
n
ta
g
e
o
f
th

e
p
o
p
u
la
ti
o
n
o
f
a
g
iv
e
n
c
o
u
n
tr
y
th

a
t
is

o
v
e
rw

e
ig
h
t
in

lo
g

G
lo
b
a
l
H
e
a
lt
h
O
b
s
e
rv
a
to
ry

D
a
ta

(W
o
rl
d
H
e
a
lt
h
O
rg
a
n
iz
a
ti
o
n
)

M
o
rb

id
o
b
e
s
it
y
ra
te

(M
O
B
E
)

T
h
e
p
e
rc
e
n
ta
g
e
o
f
th

e
p
o
p
u
la
ti
o
n
o
f
a
g
iv
e
n
c
o
u
n
tr
y
th

a
t
is

m
o
rb

id
o
b
e
s
e
in

lo
g

G
lo
b
a
l
H
e
a
lt
h
O
b
s
e
rv
a
to
ry

D
a
ta

(W
o
rl
d
H
e
a
lt
h
O
rg
a
n
iz
a
ti
o
n
)

In
d
e
p
e
n
d
e
n
t
v
a
ri
a
b
le
s

L
G
D
P
P
C

E
x
p
e
n
d
it
u
re
-s
id
e
re
a
l
G
D
P
a
t
c
h
a
in
e
d
P
P
P
s
(i
n
m
il
.
2
0
1
1
U
S
$
)/
P
o
p
u
la
ti
o
n
(i
n
m
il
li
o
n
s
)
in

lo
g

P
e
n
n
W

o
rl
d
T
a
b
le

(P
W

T
)
9
.0

T
R
A
D
E

T
ra
d
e
is

th
e
s
u
m

o
f
e
x
p
o
rt
s
a
n
d
im

p
o
rt
s
o
f
m
e
rc
h
a
n
d
is
e
m
e
a
s
u
re
d
a
s
a
s
h
a
re

o
f
c
u
rr
e
n
t
P
P
P
s

P
W

T
9
.0

U
R
B
A
N

U
rb

a
n
p
o
p
u
la
ti
o
n
re
fe
rs

to
p
e
o
p
le

li
v
in
g
in

u
rb

a
n
a
re
a
s
a
s
d
e
fi
n
e
d
b
y
n
a
ti
o
n
a
l
s
ta
ti
s
ti
c
a
l
o
ffi
c
e
s

U
n
it
e
d
N
a
ti
o
n
s
P
o
p
u
la
ti
o
n
D
iv
is
io
n

F
L
P

F
e
m
a
le

la
b
o
r
fo
rc
e
a
s
a
p
e
rc
e
n
ta
g
e
o
f
th

e
to
ta
l
la
b
o
r
fo
rc
e

W
o
rl
d
D
e
v
e
lo
p
m
e
n
t
In
d
ic
a
to
rs

(W
D
I)

G
IN

I
G
in
i
in
d
e
x
m
e
a
s
u
re
s

W
o
rl
d
B
a
n
k
,
D
e
v
e
lo
p
m
e
n
t
R
e
s
e
a
rc
h
G
ro

u
p

F
L
P
,
fe
m
a
le

la
b
o
r
p
a
rt
ic
ip
a
ti
o
n
;
G
IN

I,
G
in
i
in
d
e
x
;
U
R
B
A
N
,
u
rb

a
n
p
o
p
u
la
ti
o
n
;
L
G
D
P
P
C
,
lo
g
o
f
re
a
l
G
D
P
p
e
r
c
a
p
it
a
;
T
R
A
D
E
,
tr
a
d
e
o
p
e
n
n
e
s
s
.

Table 3 e Summary statistics of key variables.

Variable No. of obs. Mean Std. dev. Max Min

OVR (Both) 910 34.624 16.411 70.800 5.600

OVR (Male) 910 31.822 18.829 70.400 4.100

OVR (Female) 910 37.185 15.387 73.400 6.500

OBE (Both) 910 10.516 7.417 34.600 0.200

OBE (Male) 910 7.474 6.688 31.200 0.100

OBE (Female) 910 13.445 8.914 43.200 0.300

MOBE (Both) 896 10.808 7.710 36.408 0.186

MOBE (Male) 910 0.230 0.382 4.362 0.001

MOBE (Female) 910 1.130 1.300 8.328 0.001

LGDPPC 910 8.667 1.216 12.272 4.997

TRADE 910 0.481 0.342 1.839 0.005

URBAN 910 51.266 23.238 98.263 4.721

FLP 630 0.5069 0.166 0.876 0.081

GINI 356 42.762 9.284 65.800 23.000

FLP, female labor participation; GINI, Gini index; URBAN, urban

population; LGDPPC, log of real GDP per capita; TRADE, trade

openness; OBE, obesity rate; OVR, overweight rate, MOBE, morbid

obesity rate.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5 29
sample covers 130 countries during the 5-year interval period

from 1975 to 2010, and the number of observations is 910.

Because variables of FLP and GINI are limited, the number of

observations is decreased in the model estimations for the

robustness checks.

One crucial limitation is that our model may suffer from

endogeneity problems. Dynamic panel data models include

some lags of the dependent variable as covariates, as a result

of which the model allows for a partial adjustment mecha-

nism, and they also contain unobserved panel-level fixed ef-

fects. By construction, the lagged dependent variables are

correlated with the unobserved panel-level effects. In addi-

tion, explanatory variables are not strictly exogenous; thus,

they are correlated with past and possibly current realizations

of the error. For example, obesity could influence the income

level. Moreover, the model may contain heteroskedasticity

and autocorrelation within individual units’ errors, but not
across them. In such cases, standard estimators become

inconsistent. Arellano and Bond develop a consistent GMM

estimator for the model.41 However, the Arellano-Bond esti-

mator performs poorly if the autoregressive parameters or the

ratio of the variance of the panel-level effect to the variance of

the idiosyncratic error are relatively large. Poor instruments in

the difference GMM estimator cause inefficient and biased

coefficient estimates.41,42

To address this issue, the system GMM, which includes

additional moment conditions, is developed.33 This esti-

mator is designed for panel data with many panels and short

periods, under the assumptions that there is no autocorre-

lation in the idiosyncratic errors and that the panel-level

effects are uncorrelated with the first difference of the first

observation of the dependent variable. The system GMM

estimator combines the use of lagged levels of the series as

instruments for the predetermined and endogenous vari-

ables in equations in first differences and the use of lagged

differences of the dependent variable as instruments for

equations in levels. The system GMM estimator derives more

https://doi.org/10.1016/j.puhe.2019.01.004

Table 4 e Correlation matrix of key variables (full sample).

Variable OVR (B) OVR (M) OVR (F) OBE (B) OBE (M) OBE (F) MOBE (B) MOBE (M) MOBE (F) LGDPPC TRADE URBAN FLP

OVR (B) 1.000

OVR (M) 0.972 1.000

OVR (F) 0.955 0.858 1.000

OBE (B) 0.971 0.925 0.950 1.000

OBE (M) 0.939 0.972 0.820 0.940 1.000

OBE (F) 0.914 0.809 0.972 0.962 0.811 1.000

MOBE (B) 0.972 0.928 0.949 1.000 0.942 0.960 1.000

MOBE (M) 0.640 0.666 0.554 0.727 0.784 0.618 0.728 1.000

MOBE (F) 0.668 0.573 0.731 0.802 0.660 0.843 0.798 0.776 1.000

LGDPPC 0.767 0.824 0.631 0.730 0.808 0.603 0.733 0.605 0.490 1.000

TRADE 0.363 0.394 0.290 0.341 0.395 0.263 0.341 0.299 0.200 0.507 1.000

URBAN 0.839 0.856 0.752 0.786 0.821 0.693 0.789 0.534 0.478 0.774 0.373 1.000

FLP �0.375 �0.307 �0.439 �0.391 �0.260 �0.470 �0.390 �0.081 �0.285 �0.238 �0.003 �0.293 1.000
GINI �0.083 �0.227 0.101 �0.078 �0.258 0.068 �0.082 �0.215 0.054 �0.270 �0.181 �0.098 0.051
FLP, female labor participation; GINI, Gini index; URBAN, urban population; LGDPPC, log of real GDP per capita; TRADE, trade openness; OBE,

obesity rate; OVR, overweight rate, MOBE, morbid obesity rate.

(B), (M), and (F) mean both sex, males, and females, respectively.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 530
efficient results. Thus, this study employs the two-step sys-

tem GMM estimator to estimate the empirical model. Lacking

valid instruments for the income level, we cannot claim to

have fully resolved all endogeneity issues. However, the

system GMM estimator mitigates some of them since this

methodology is suitable for the adjustment process of the

dependent variable with independent variables that are not

strictly exogenous.
Results

Tables 5e7 present the estimated results of the OLS, the fixed

effects models, and the dynamic panel data models with the

two-step system GMM estimators in the overweight, obesity,

and morbid obesity equations, for all people (both sex), males,
Table 5 e Estimated results e Both sex.

Variable Overweight ratios

OLS FE GMM O

Lag of dependent variable 1.022*** 0.814*** 1.006*** 1.0

(0.003) (0.020) (0.009) (0.0

LGDPPC 2.913*** 1.470** 4.403*** 1.3

(0.368) (0.571) (1.017) (0.1

LGDPPC∧2 �0.171*** �0.082** �0.261*** �0.0
(0.021) (0.036) (0.060) (0.0

TRADE 0.256*** 0.002 0.361 0.1

(0.087) (0.120) (0.311) (0.0

URBAN 0.003 0.079*** 0.018*** 0.0

(0.002) (0.010) (0.006) (0.0

AR(2) 0.251

Hansen test 0.281

No. of obs. 910 910 910 910

No. of countries 130 130 130 130

R-squared 0.998 0.991 0.9

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI,

ordinary least squares; FE, fixed effects; GMM, generalized method of mo

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels,

gressions include time dummies.
and females, respectively. Table 8 summarizes the results of

the two-step system GMM estimators and the peak-out level

of real GDP per capita when the results support the obesity

Kuznets curve. When applying the system GMM estimators,

we need to confirm that there is no second-order serial cor-

relation. As the tests for the specification for the absence of

serial correlation, the Arellano-Bond test statistic for second

order autocorrelation (AR(2)) show that in every model, the

null hypothesis of no second-order serial correlation cannot

be rejected, as required by the specification. In addition, the

Hansen tests for exogeneity show that the J-statistic has a P-

value greater than 0.10 in all models; thus, we cannot reject

the null hypothesis that the instruments as a group are

exogenous in the system GMM estimation, as required by the

specification. The main interest of our study is whether the

income-obesity relationship shows the obesity Kuznets curve.
Obesity ratios Morbid obesity ratios

LS FE GMM OLS FE GMM

88*** 0.996*** 1.074*** 1.087*** 0.995*** 1.083***

04) (0.015) (0.010) (0.004) (0.015) (0.009)

43*** 0.329 2.184*** 1.384*** 0.098 1.652***

76) (0.330) (0.542) (0.190) (0.298) (0.412)

78*** �0.018 �0.132*** �0.080*** �0.003 �0.097***
10) (0.021) (0.032) (0.011) (0.019) (0.024)

54*** 0.125 0.159 0.140*** 0.068 0.250

51) (0.078) (0.154) (0.053) (0.072) (0.257)

01 0.029*** 0.011*** 0.001 0.032*** 0.005*

01) (0.005) (0.003) (0.001) (0.005) (0.003)

0.257 0.292

0.119 0.246

910 910 896 896 896

130 130 130 130 130

97 0.993 0.997 0.993

Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ments; AR(2), ArellanoeBond test for second order autocorrelation.

respectively. Robust standard error is reported in parenthesis. All re-

https://doi.org/10.1016/j.puhe.2019.01.004

Table 6 e Estimated results e Males.

Variable Overweight ratios Obesity ratios Morbid obesity ratios

OLS FE GMM OLS FE GMM OLS FE GMM

Lag of dependent variable 1.021*** 0.891*** 0.985*** 1.110*** 1.053*** 1.086*** 1.410*** 1.380*** 1.394***

(0.003) (0.015) (0.011) (0.004) (0.011) (0.011) (0.022) (0.028) (0.017)

LGDPPC 3.701*** 1.603** 5.652*** 1.224*** 0.083 2.127*** 0.027*** �0.044 0.056*
(0.386) (0.626) (0.981) (0.143) (0.268) (0.444) (0.009) (0.033) (0.034)

LGDPPC∧2 �0.211*** �0.082** �0.317*** �0.068*** 0.000 �0.119*** �0.001*** 0.003 �0.003
(0.022) (0.039) (0.056) (0.008) (0.017) (0.026) (0.001) (0.002) (0.002)

TRADE 0.087 0.121 �0.035 0.050 0.095 �0.022 0.012*** 0.002 0.017
(0.088) (0.136) (0.314) (0.040) (0.063) (0.133) (0.004) (0.008) (0.011)

URBAN 0.016*** 0.082*** 0.044*** 0.005*** 0.027*** 0.014*** 0.000*** 0.001 0.000

(0.002) (0.010) (0.007) (0.001) (0.005) (0.002) (0.000) (0.001) (0.000)

AR(2) 0.422 0.784 0.818

Hansen test 0.216 0.202 0.122

# of obs. 910 910 910 910 910 910 910 910 910

# of countries 130 130 130 130 130 130 130 130 130

R-squared 0.999 0.996 0.998 0.995 0.995 0.993

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI, Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ordinary least squares; FE – Fixed effects; GMM, generalized method of moments; AR(2), ArellanoeBond test for second order autocorrelation.

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Robust standard error is reported in parenthesis. All re-

gressions include time dummies.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5 31
For the results of both sex in Table 5, the estimations show

that the coefficients on LGDPPC are significantly positive and

that those on squared LGDPPC are significantly negative for

the OLS and system GMM estimations, regardless of the types

of weight-related health status. Our analysis presents clear

evidence supporting an inverted U-shaped Kuznets curve

relationship between weight-related health status and real

income per capita. For low-income countries, as incomes in-

crease, the rates of overweight, obesity, and morbid obesity

increase; thus, an increase in incomes raises the health risk. In

contrast, for high-income countries, as incomes increase, the

rates of overweight, obesity, and morbid obesity decrease;
Table 7 e Estimated results e Females.

Variable Overweight ratios

OLS FE GMM OL

Lag of dependent variable 1.027*** 0.822*** 1.024*** 1.0

(0.003) (0.029) (0.007) (0.0

LRGDPPC 2.247*** 2.057*** 3.360*** 1.8

(0.380) (0.700) (1.058) (0.2

LRGDPPC∧2 �0.139*** �0.126*** �0.214*** �0.1
(0.022) (0.045) (0.064) (0.0

TRADE 0.425*** �0.263* 1.017*** 0.2
(0.100) (0.159) (0.347) (0.0

URBAN �0.012*** 0.070*** �0.006 �0.0
(0.002) (0.013) (0.006) (0.0

AR(2) 0.372

Hansen test 0.245

# of obs. 910 910 910 910

# of countries 130 130 130 130

R-squared 0.997 0.987 0.9

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI,

ordinary least squares; FE, fixed effects; GMM, generalized method of mo

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, r

gressions include time dummies.
thus, an increase in incomes reduces the health risk. A recent

study on the case of the United States also recommends that

improving people’s income from low- to middle-income levels
will help control the rising obesity trend in the US adult pop-

ulation.43 Our system GMM estimators also show that the

rates of overweight, obesity, and morbid obesity for both sex

peak at the levels of real income per capita of approximately

US$ 4600, 4000, and 4800, respectively (Table 8). It should be

noted that the peak-out levels depend highly on empirical

methods and model specifications, so that our estimated

peak-out levels under the system GMM may not be precise but

should be considered just a reference. Our results based on
Obesity ratios Morbid obesity ratios

S FE GMM OLS FE GMM

78*** 0.951*** 1.075*** 1.221*** 1.157*** 1.219***

04) (0.019) (0.008) (0.008) (0.016) (0.012)

15*** 0.547 2.462*** 0.289*** 0.095 0.375***

57) (0.415) (0.653) (0.045) (0.068) (0.112)

08*** �0.033 �0.152*** �0.017*** �0.005 �0.022***
15) (0.027) (0.039) (0.003) (0.004) (0.006)

81*** �0.054 0.451* 0.035*** �0.015 0.057
69) (0.094) (0.237) (0.012) (0.019) (0.049)

04*** 0.037*** 0.002 �0.001** 0.005*** �0.000
01) (0.007) (0.004) (0.000) (0.001) (0.001)

0.649 0.453

0.225 0.223

910 910 910 910 910

130 130 130 130 130

96 0.988 0.994 0.986

Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ments; AR(2), ArellanoeBond test for second order autocorrelation.

espectively. Robust standard error is reported in parenthesis. All re-

https://doi.org/10.1016/j.puhe.2019.01.004

Table 8 e System-GMM estimations.

Variable Overweight Obesity Morbid obesity

Both sex Male Female Both sex Male Female Both sex Male Female

Lag of dependent variable 1.006*** 0.985*** 1.024*** 1.074*** 1.086*** 1.075*** 1.083*** 1.394*** 1.219***

(0.009) (0.011) (0.007) (0.010) (0.011) (0.008) (0.009) (0.017) (0.012)

LGDPPC 4.403*** 5.652*** 3.360*** 2.184*** 2.127*** 2.462*** 1.652*** 0.056* 0.375***

(1.017) (0.981) (1.058) (0.542) (0.444) (0.653) (0.412) (0.034) (0.112)

LGDPPC∧2 �0.261*** �0.317*** �0.214*** �0.132*** �0.119*** �0.152*** �0.097*** �0.003 �0.022***
(0.060) (0.056) (0.064) (0.032) (0.026) (0.039) (0.024) (0.002) (0.006)

TRADE 0.361 �0.035 1.017*** 0.159 �0.022 0.451* 0.250 0.017 0.057
(0.311) (0.314) (0.347) (0.154) (0.133) (0.237) (0.257) (0.011) (0.049)

URBAN 0.018*** 0.044*** �0.006 0.011*** 0.014*** 0.002 0.005* 0.000 �0.000
(0.006) (0.007) (0.006) (0.003) (0.002) (0.004) (0.003) (0.000) (0.001)

AR(2) 0.251 0.422 0.372 0.257 0.784 0.649 0.292 0.818 0.453

Hansen test 0.281 0.216 0.245 0.119 0.202 0.225 0.246 0.122 0.223

# of obs. 910 910 910 910 910 910 896 910 910

# of countries 130 130 130 130 130 130 130 130 130

Peak-Out level of GDPPC 4627 7469 2530 4023 7434 3272 4795 – 4700

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI, Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ordinary least squares; GMM, generalized method of moments AR(2), ArellanoeBond test for second order autocorrelation.

Notes: *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Robust standard error is reported in parenthesis. All

regressions include time dummies. No peak-out level of GDPPC is shown in the morbid obesity equation for males because the coefficient on

squared LGDPPC is insignificant.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 532
country-level data are partly consistent with the previous

findings that income per capita is positively related to BMI up

to US$ 3000, with no significant relationship beyond this

level,31 that both overweight and obesity are non-linearly

related to income per capita,44 and that there is an empirical

basis for the obesity Kuznets curve for white females within

the United States.25

Our models cannot discuss possible links between the

characteristics of individuals and the obesity epidemic

because our sample accounts only for aggregate features at

the country level. However, our analysis confirms some
Table 9 e Models with additional control variables, FLP (system

Variable Overweight

Both sex Male Female Both

Lag of dependent variable 1.004*** 0.991*** 1.019*** 1.0

(0.011) (0.012) (0.010) (0.0

LGDPPC 3.876*** 5.412*** 3.371** 1.6

(1.159) (1.303) (1.335) (0.6

LGDPPC∧2 �0.239*** �0.320*** �0.220*** �0.1
(0.069) (0.076) (0.081) (0.0

TRADE 1.100*** 0.977** 1.455*** 0.4

(0.386) (0.391) (0.446) (0.2

URBAN 0.009 0.035*** �0.012 0.0
(0.009) (0.008) (0.008) (0.0

FLP �1.817 �1.805 �0.857 �0.4
(1.381) (1.368) (1.565) (0.6

AR(2) 0.635 0.869 0.643 0.9

Hansen test 0.246 0.394 0.282 0.1

# of obs. 630 630 630 63

# of countries 126 126 126 12

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI,

ordinary least squares; GMM, generalized method of moments; AR(2), Ar

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels,

gressions include time dummies.
important national trends of the income-obesity relationship,

which could help regulators plan and implement sound

health policies to reduce people’s health risk as well as the
financial burden associated with healthcare costs for in-

dividuals and governments. The policy implications suggest

that a strong initiative for health policy targeting obesity

prevention is required, particularly for middle-income coun-

tries, many of which are currently experiencing high eco-

nomic growth.

To check possible gender differences in the income-obesity

relationship, we estimate the empirical models for each
-GMM estimation).

Obesity Morbid obesity

sex Male Female Both sex Male Female

74*** 1.082*** 1.072*** 1.073*** 1.382*** 1.216***

12) (0.014) (0.013) (0.012) (0.016) (0.015)

94** 1.881*** 2.556*** 2.246*** 0.043 0.396**

75) (0.566) (0.805) (0.571) (0.036) (0.154)

05*** �0.111*** �0.163*** �0.136*** �0.002 �0.024***
40) (0.033) (0.048) (0.034) (0.002) (0.009)

12* 0.001 1.066*** 0.535* 0.010 0.160**

37) (0.236) (0.363) (0.274) (0.023) (0.081)

09* 0.020*** �0.005 0.006 0.001** �0.001
05) (0.005) (0.006) (0.004) (0.000) (0.001)

01 �0.387 �0.921 �0.689 0.027 �0.238
73) (0.745) (1.138) (0.582) (0.047) (0.152)

68 0.793 0.320 0.253 0.593 0.621

68 0.220 0.187 0.130 0.172 0.182

0 630 630 630 630 630

6 126 126 126 126 126

Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ellanoeBond test for second order autocorrelation.

respectively. Robust standard error is reported in parenthesis. All re-

https://doi.org/10.1016/j.puhe.2019.01.004

Table 10 e Models with additional control variables, GINI (system-GMM estimation).

Variable Overweight Obesity Morbid obesity

Both sex Male Female Both sex Male Female Both sex Male Female

Lag of dependent variable 0.982*** 0.995*** 1.006*** 1.058*** 1.087*** 1.054*** 1.057*** 1.361*** 1.211***

(0.013) (0.013) (0.014) (0.013) (0.012) (0.013) (0.012) (0.026) (0.018)

LGDPPC 5.796*** 7.560*** 3.778*** 2.363*** 2.152*** 2.787*** 2.370*** �0.022 0.510***
(1.274) (1.189) (1.209) (0.722) (0.548) (1.016) (0.733) (0.073) (0.167)

LGDPPC∧2 �0.334*** �0.432*** �0.230*** �0.135*** �0.121*** �0.165*** �0.135*** 0.002 �0.030***
(0.072) (0.068) (0.069) (0.043) (0.033) (0.058) (0.043) (0.004) (0.010)

TRADE 0.251 0.173 0.327 0.003 0.095 0.106 0.066 �0.008 0.011
(0.470) (0.492) (0.443) (0.253) (0.226) (0.301) (0.297) (0.030) (0.068)

URBAN 0.027*** 0.036*** 0.005 0.012*** 0.014*** 0.008 0.013*** 0.001** 0.002

(0.009) (0.011) (0.008) (0.005) (0.003) (0.006) (0.004) (0.001) (0.001)

GINI 0.032*** 0.021* 0.045*** 0.020*** 0.009* 0.032*** 0.020*** 0.001* 0.002

(0.010) (0.011) (0.012) (0.007) (0.006) (0.009) (0.007) (0.001) (0.002)

AR(2) 0.587 0.247 0.791 0.394 0.449 0.555 0.721 0.397 0.513

Hansen test 0.371 0.323 0.390 0.411 0.309 0.339 0.377 0.366 0.586

# of obs. 356 356 356 356 356 356 356 356 356

# of countries 110 110 110 110 110 110 110 110 110

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI, Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ordinary least squares; GMM, generalized method of moments; AR(2), ArellanoeBond test for second order autocorrelation.

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Robust standard error is reported in parenthesis. All re-

gressions include time dummies.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5 33
gender. The estimated results of the system GMM estimators,

shown in Tables 6 and 7, generally confirm the existence of

the obesity Kuznets curve for males and females, except for

the case of the morbid obesity rates for males. In addition,

Table 8 shows that the system GMM estimations of the peak-

out levels of real income per capita. Although we admit that

our estimated peak-out levels should be considered just a

reference, as mentioned in the previous discussion, the esti-

mated peak-out levels are consistently higher for males than
Table 11 e Models with all additional control variables, FLP, an

Variable Overweight

Both sex Male Female Both

Lag of dependent variable 0.992*** 0.997*** 1.002*** 1.0

(0.013) (0.014) (0.017) (0.0

LGDPPC 5.307*** 7.200*** 3.567*** 2.2

(1.164) (1.825) (0.938) (0.6

LGDPPC∧2 �0.313*** �0.415*** �0.221*** �0.1
(0.065) (0.103) (0.053) (0.0

TRADE 0.238 0.216 0.359 0.2

(0.421) (0.431) (0.378) (0.2

URBAN 0.023** 0.040*** �0.003 0.0
(0.010) (0.010) (0.009) (0.0

FLP �0.086 1.159 �1.303 �0.1
(1.004) (1.165) (1.203) (0.5

GINI 0.036*** 0.020* 0.047*** 0.0

(0.012) (0.012) (0.016) (0.0

AR(2) 0.620 0.190 0.538 0.6

Hansen test 0.154 0.188 0.232 0.3

# of obs. 356 356 356 356

# of countries 110 110 110 110

LGDPPC, log of real GDP per capita; FLP, female labor participation; GINI,

ordinary least squares; GMM, generalized method of moments; AR(2), Ar

*, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, r

gressions include time dummies.
for females, irrespective of the choices of obesity measures

and empirical methods (the OLS, fixed effects, and system

GMM estimations). As an economy develops, the weight-

related health status of females tends to peak at the earlier

stage of development than it does for males. Females and

males face different types of social pressure in regard to being

overweight. Since females are likely to experience more social

pressure over obesity than males, their concern regarding

overweight is relatively large.25,45 Even with a similar body
d GINI (system-GMM estimation).

Obesity Morbid obesity

sex Male Female Both sex Male Female

69*** 1.097*** 1.059*** 1.059*** 1.369*** 1.220***

12) (0.015) (0.016) (0.015) (0.030) (0.017)

21*** 2.323*** 2.480*** 2.358*** �0.015 0.468***
18) (0.715) (0.605) (0.848) (0.093) (0.159)

30*** �0.134*** �0.151*** �0.135*** 0.002 �0.027***
35) (0.040) (0.035) (0.050) (0.006) (0.009)

45 0.202 0.243 0.114 0.003 0.074

71) (0.245) (0.286) (0.262) (0.037) (0.061)

08 0.014*** 0.002 0.013*** 0.001* 0.000

05) (0.004) (0.008) (0.004) (0.001) (0.001)

59 0.523 �0.821 �0.135 0.089 �0.065
10) (0.559) (1.119) (0.857) (0.068) (0.148)

21*** 0.009 0.033*** 0.017** 0.002** 0.003

07) (0.006) (0.010) (0.007) (0.001) (0.002)

18 0.532 0.906 0.870 0.430 0.530

18 0.183 0.220 0.235 0.179 0.293

356 356 356 356 356

110 110 110 110 110

Gini index; URBAN, urban population; TRADE, trade openness; OLS,

ellanoeBond test for second order autocorrelation.

espectively. Robust standard error is reported in parenthesis. All re-

https://doi.org/10.1016/j.puhe.2019.01.004

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 534
weight, females tend to make more efforts to keep a slimmer

body weight since they are dissatisfied with their weight.46

Thus, as incomes increase, weight-related health status

peaks at a lower income level for females than for males.

For the robustness checks, Tables 9 and 10 show the results

of the models including additional control variables, FLP and

GINI, respectively, and Table 11 presents the results of the

models including both FLP and GINI. The results are still

consistent with our baseline findings supporting the existence

of the obesity Kuznets curve. Concerning other control vari-

ables, the results of the system GMM estimations confirm a

positive relationship between urbanization (URBAN) and

weight-related health status, particularly for both sex and

males, which suggests that urbanization increases the obesity

epidemic by allowing urban males to improve their access to

food and to have a sedentary life. In addition, our analysis

verifies a positive relationship between obesity rates and the

GINI, which implies that income inequality increases the

obesity epidemic. However, our estimations generally present

less clear evidences of the links of the obesity epidemic with

TRADE and FLP.
Discussion

The obesity epidemic has been recognized as a global pandemic

disease, but its extent differs substantially across countries.

Among various socio-economic factors, economic develop-

ment can be considered a crucial factor of such international

variation. This study has examined the relationship between

economic development and weight-related health status,

which is measured by the rates of overweight, obesity, and

morbid obesity, over 130 countries during the 5-year interval

period from 1975 to 2010. The estimation has shown a clear

pattern of the obesity Kuznets curve. For low-income countries,

as incomes increase, weight-related health status deteriorates

with a high health risk. In contrast, for high-income countries,

as incomes increase, weight-related health status improves

with the reduction of the health risk. Our analysis supports the

argument that middle-income countries, many of which are

currently enjoying high economic growth, may be in danger of

damage to people’s health and the related financial burden,
requiring a strong initiative for health policy targeting obesity

prevention. In addition, the results have suggested different

types of social pressure in regard to being overweight between

males and females, i.e. concern regarding overweight is larger

for females than for males since females are likely to experi-

ence more social pressure over obesity than males.

Our study has several limitations. First, this study has

assumed that BMI reflects an individual’s health status. How-
ever, it may not be the best measure for capturing an in-

dividual’s health status because of its failure to measure body
fat and muscle. Second, itis acknowledged that the relationship

between income and health depends on changes in poverty and

inequality in a more complex manner. Our simple empirical

models may not evaluate such a relationship extensively, so

that more careful empirical analysis should be required to

verify the obesity Kuznets curve. Third, concerning the meth-

odologicalissues, thesystem GMMestimationswehaveapplied

in this study help capture dynamic effects over time that relate
to the potential recursive association. However, this method

cannot solve all endogeneity issues, such as omitted variable

problems. Although we admit these limitations, our study

focusing on international differences would still contribute to

the understanding of current situations and trends of the

obesity epidemic in relation to economic development.
Author statements

Ethical approval

None sought. All data used in this study are at the country-

level and are obtained from the open source.

Funding

None declared.

Competing interests

None declared.
r e f e r e n c e s

1. Sturm R. The effects of obesity, smoking, and drinking on
medical problems and costs. Health Aff 2002;21:245e53.

2. Fontaine KR, Redden DT, Wang C, Westfall AO, Allison DB.
Years of life lost due to obesity. JAMA 2003;289:187e93.

3. Finkelstein EA, Strombotne KL. The economics of obesity. Am
J Clin Nutr 2010;91:1520Se4S.

4. Lehnert T, Sonntag D, Konnopka A, Riedel-Heller S, K€onig HH.
Economic costs of overweight and obesity. Best Pract Res Clin
Endocrinol Metabol 2013;27:105e15.

5. Kjellberg J, Larsen AT, Ibsen R, Højgaard B. The socioeconomic
burden of obesity. Obes Facts 2017;10:493e502.

6. Tremmel M, Gerdtham UG, Nilsson PM, Saha S. Economic
burden of obesity: a systematic literature review. Int J Environ
Res Publ Health 2017;14:435.

7. Puhl RM, Heuer CA. Obesity stigma: important considerations
for public health. Am J Publ Health 2010;100:1019e28.

8. Mocan NH, Tekin E. Obesity, self-esteem and wages. National
Bureau of Economic Research; 2009. Working Paper No.
15101.

9. Verhaeghe N, De Greve O, Annemans L. The potential health
and economic effect of a body mass index decrease in the
overweight and obese population in Belgium. Publ Health
2016;134:26e33.

10. Squalli J. The environmental impact of obesity: longitudinal
evidence from the United States. Publ Health 2017;149:89e98.

11. Dinsa GD, Goryakin Y, Fumagalli E, Suhrcke M. Obesity and
socioeconomic status in developing countries: a systematic
review. Obes Rev 2012;13:1067e79.

12. Goryakin Y, Lobstein T, James WP, Suhrcke M. The impact of
economic, political and social globalization on overweight
and obesity in the 56 low and middle income countries. Soc Sci
Med 2015;133:67e76.

13. Costa-Font J, Mas N. ‘Globesity’? The effects of globalization
on obesity and caloric intake. Food Policy 2016;64:121e32.

14. Miljkovic D, Shaik S, Miranda S, Barabanov N, Liogier A.
Globalisation and obesity. World Econ 2015;38:1278e94.

15. Miljkovic D, de Miranda SH, Kassouf AL, Oliveira FC.
Determinants of obesity in Brazil: the effects of trade

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref1

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref2

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref3

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref4

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref5

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref6

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref7

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref8

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref9

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref10

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref11

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref12

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref13

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref14

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref15

https://doi.org/10.1016/j.puhe.2019.01.004

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 6 e3 5 35
liberalization and socio-economic variables. Appl Econ
2018;50:3076e88.

16. Ljungvall �A. The freer the fatter? a panel study of the relationship
between body-mass index and economic freedom. Department of
Economics, Lund University; 2013. Working Paper.

17. Lawson RA, Murphy RH, Williamson CR. The relationship
between income, economic freedom, and BMI. Publ Health
2016;134:18e25.

18. Philipson T. The world-wide growth in obesity: an economic
research agenda. Health Econ 2001;10:1e7.

19. Lakdawalla D, Philipson T. The growth of obesity and
technological change: a theoretical and empirical examination.
National Bureau of Economic Research; 2002. Working Paper
No.8946.

20. Finkelstein EA, Ruhm CJ, Kosa KM. Economic causes and
consequences of obesity. Annu Rev Public Health
2005;26:239e57.

21. Kuznets S. Economic growth and income inequality. Am Econ
Rev 1955;45:1e28.

22. Molini V, Nube M, van den Boom B. Adult BMI as a health and
nutritional inequality measure: applications at macro and
micro Levels. World Dev 2010;38:1012e23.

23. Sahn DE, Younger SD. Measuring intra-household health
inequality: explorations using the body mass index. Health
Econ 2009;18:S13e36.

24. Costa-Font J, Hernandez-Quevedo C, Sato A. A health
‘Kuznets’ curve’? Cross-sectional and longitudinal evidence
on concentration indices’. Soc Indicat Res 2018;136:439e52.

25. Grecu AM, Rotthoff KW. Economic growth and obesity:
findings of an obesity Kuznets curve. Appl Econ Lett
2015;22:539e43.

26. Koplan JP, Dietz WH. Caloric imbalance and public health
policy. JAMA 1999;282:1579e81.

27. Lakdawalla D, Philipson T. The growth of obesity and
technological change. Econ Hum Biol 2009;7:283e93.

28. Hruby A, Hu FB. The epidemiology of obesity: a big picture.
Pharmacoeconomics 2015;33:673e89.

29. Jolliffe D. Overweight and poor? On the relationship between
income and the body mass index. Econ Hum Biol
2011;9:342e55.

30. Goryakin Y, Suhrcke M. Economic development,
urbanization, technological change and overweight: what do
we learn from 244 Demographic and Health Surveys? Econ
Hum Biol 2014;14:109e27.

31. Egger G, Swinburn B, Islam FA. Economic growth and obesity:
an interesting relationship with world-wide implications.
Econ Hum Biol 2012;10:147e53.
32. Arellano M, Bover O. Another look at the instrumental
variable estimation of error-components models. J Econom
1995;68:29e51.

33. Blundell R, Bond S. Initial conditions and moment restrictions
in dynamic panel data models. J Econom 1998;87:115e43.

34. Owen AL, Wu S. Is trade good for your health? Rev Int Econ
2007;15:660e82.

35. Popkin BM. The nutrition transition in low-income countries:
an emerging crisis. Nutr Rev 1994;52:285e98.

36. Trost SG, Owen N, Bauman AE, Sallis JF, Brown W. Correlates
of adults’ participation in physical activity: review and
update. Med Sci Sports Exerc 2002;34:1996e2001.

37. Forss�en AS, Carlstedt G. “It’s heavenly to be alone!”: a room
of one’s own as a health-promoting resource for women.
Results from a qualitative study. Scand J Publ Health
2006;34:175e81.

38. Welch N, Hunter W, Butera K, Willis K, Cleland V, Crawford D,
Ball K. Women’s work. Maintaining a healthy body weight.
Appetite 2009;53:9e15.

39. Pickett KE, Kelly S, Brunner E, Lobstein T, Wilkinson RG.
Wider income gaps, wider waistbands? An ecological study of
obesity and income inequality. J Epidemiol Commun Health
2005;59:670e4.

40. Biggs B, King L, Basu S, Stuckler D. Is wealthier always
healthier? The impact of national income level, inequality,
and poverty on public health in Latin America. Soc Sci Med
2010;71:266e73.

41. Arellano M, Bond S. Some tests of specification for panel data:
Monte Carlo evidence and an application to employment
equations. Rev Econ Stud 1991;58:277e97.

42. Bound J, Jaeger DA, Baker RM. Problems with instrumental
variables estimation when the correlation between the
instruments and the endogenous explanatory variable is
weak. J Am Stat Assoc 1995;90:443e50.

43. Chen HJ, Xue H, Liu S, Huang TT, Wang YC, Wang Y. Obesity
trend in the United States and economic intervention options
to change it: a simulation study linking ecological
epidemiology and system dynamics modeling. Publ Health
2018;161:20e8.

44. Lobstein T, Leach RJ. Tackling obesities: future choices-
International comparisons of obesity trends, determinants and
responses-Evidence Review� 2. children. Foresight, Government
Office of the Chief Scientit; 2007.

45. Cawley J. The impact of obesity on wages. J Hum Resour
2004;39:451e74.

46. Pingitore R, Spring B, Garfieldt D. Gender differences in body
satisfaction. Obes Res 1997;5:402e9.

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref15

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref16

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref17

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref18

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref19

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref20

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref21

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref22

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref23

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref24

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref25

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref26

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref27

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref28

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref29

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref30

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref31

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref32

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref33

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref34

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref35

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref36

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref37

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref38

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref39

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref40

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref41

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref42

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref43

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref44

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref45

http://refhub.elsevier.com/S0033-3506(19)30004-6/sref46

https://doi.org/10.1016/j.puhe.2019.01.004

Obesity Kuznets curve: international evidence
Introduction
Methods
Results
Discussion
Author statements
Ethical approval
Funding
Competing interests
References

Prevention–of-what–in-whom–and-how-_2019_Public-Health

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) A 1 eA 2
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Editorial
Prevention: of what, in whom, and how?
The vast majority, if not all, of those of us working in the field

of public health will have some grasp on the concept of

different types of prevention. Primary prevention focuses on

preventing incident disease; secondary prevention involves

early detection of disease, before symptom onset when in-

terventions can halt or reverse progress; and tertiary preven-

tion refers to limiting the adverse outcomes of established

disease. Typically, prevention, when unqualified, refers to

primary prevention.

With an ever growing classification of diseases, where

should we be focussing our prevention efforts? Should we,

for example, be seeking lifestyle-based, risk factor modifi-

cation as primary (or secondary) prevention or should we be

concentrating elsewhere? The Global Burden of Diseases

2010 study demonstrated that across the globe, 54% of all

Disability Adjusted Life Years (DALYs) are due to non-

communicable diseases (NCDs), compared with 35% due to

communicable, maternal, neonatal and nutritional disorders

and 11% due to injuries.1 Given the relationship between

NCDs and lifestyle choices, it is unsurprising that ‘lifestyle

risk factors’ account for 40% of lost DALYs in England.2 This

is not the same worldwide. In low-income countries, there

are higher proportions of communicable disease and injury

attributable lost DALYs, and the biggest lifestyle risk factor is

malnutrition. While here in the UK we struggle with getting

people to eat less and move more, the picture is very different

elsewhere, where, for example, access to clean air and water

to prevent communicable diseases or workplaces safe from

injury needs to take priority. Understanding locale specific

risk factors is paramount and frequently our articles do just

that. For example, in this issue of Public Health, Kumar et al.3

investigate the risk of common cooking practices using

contaminated fuels on children’s respiratory health in rural
regions of India, and Donaldson et al.4 look at outdoor tem-

perature and the relationship with mortality in England

among others.

Regardless of the topic, one would hope that some educa-

tional and behaviour supportive interventions would tran-

scend specific risk factors. This is reflected in the growing

trend in large-scale preventative strategies involving sup-

porting the population to make behavioural changes. We use

the word population here purposely as opposed to patients for

two reasons: firstly, because good primary prevention will be
supporting people before they become patients and secondly,

because many of these initiatives are wider than the health

service. There are many opportunities to be proactive in

supporting people to make healthy decisions about their lives.

The Making Every Contact Count initiative relies on the

premise that ‘… staff across health, local authority and

voluntary sectors have thousands of contacts every day with

individuals and are ideally placed to promote health and

healthy lifestyles’,5 and its evaluation reports it as a low-cost

intervention adaptable to the setting and health issue.6

However, the value of this initiative for other countries and

other risk factors remains to be demonstrated.

This issue also sees the publication of a special section

curated by our guest editors Brian Ferguson (Chief Economist,

Public Health England) and Annalisa Belloni7 (Senior Health

Economist, Public Health England) reporting on the health

economics of prevention in public health. In their editorial,

Ferguson and Belloni reflect on the disconnect between recent

UK government policy promoting prevention strategies and

ever-diminishing budgets to deliver such interventions. While

the section mainly considers a UK perspective on the topic, we

think it will be of interest to all, as healthcare financing is a

challenging topic universally. As these articles show, the ev-

idence that preventative strategies are cost-effective options

does exist. We also have a growing body of evidence that

shows how we can help the population benefit from preven-

tative strategies when implemented. How we effect organ-

isational behaviour change to ensure that the commissioners

of health and health care invest in such strategies is another

question entirely.
r e f e r e n c e s

1. Murray CJL, Vos T, Lozano R, Naghavi M, Flaxman A,
Michaud C, et al. Disability-adjusted life years (DALYs) for 291
diseases and injuries in 21 regions, 1990e2010: a systematic
analysis for the Global Burden of Disease Study 2010. Lancet
2012;380(9859):2197e223.

2. Newton JN, Briggs ADM, Murray CJL, Dicker D, Foreman KJ,
Wang H, et al. Changes in health in England, with analysis by
English regions and areas of deprivation, 1990e2013: a
systematic analysis for the Global Burden of Disease Study
2013. Lancet 2015;386:2257e74.

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref1

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref2

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.03.010&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.03.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) A 1 eA 2A2
3. Kumar PS. Effects of cooking fuel sources on the respiratory
health of children: evidence from the Annual Health Survey,
Uttar Pradesh, India. Publ Health 2019;169:59e68.

4. Donaldson GC. Changes in cold-related mortalities between
1995-2016 in South-East England. Publ Health 2019;169:36e40.

5. Health Education England. Making every contact count. National
Health Service. Available online at: https://www.
makingeverycontactcount.co.uk/. [Accessed 4 March 2019].

6. Nelson A, de Normanville C, Payne K, Kelly MP. Making every
contact count: an evaluation. Publ Health 2013;127:653e60.

7. Ferguson B, Belloni A. The economics of prevention. Publ Health
2019;169:149e50.
In this issue

This issue contains our usual range of articles and additional

the health economics of prevention in public health. The six art

general reader with an interestdnot just those well versed in th

primary prevention of NCD runs through the section. As we a

alcohol in Scotland, we start our special section with a discussio

comprehensive review of the cost-effectiveness of public healt

and Care Excellence (NICE) guidelines published between Marc

vary depending on the evaluation approach adopted.

Among our other articles, we are introduced to the impact o

intent (a reduced gross domestic product, as a result of econom

economic theme, we have a country-level panel study of 130

development and weight-related health status. In case reader

formation, we move to the other end of the spectrum and h

emergency care access.
J. Morling*, F. Sim, P. Mackie

*Corresponding author.

E-mail address: publichealth@rsph.org.uk

https://doi.org/10.1016/j.puhe.2019.03.010
0033-3506/© 2019 The Royal Society for Public Health. Published

by Elsevier Ltd. All rights reserved.
ly a carefully curated special section of six articles exploring

icles examine a variety of topics and are all accessible to the

e health economics literature. The theme of predominantly

pproach a year’s experience of minimum unit pricing for
n of the use of taxation to promote healthy choices. We have

h interventions considered in National Institute for Health

h 2006 and March 2018, demonstrating how estimates can

f high-level political moves with no specific health-related

ic sanctions) on population health in Iran. Along a similar

countries examining the relationship between economic

s had not had enough finance-related health outcome in-

ave a report on the impact of individual copayments for

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref3

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref4

https://www.makingeverycontactcount.co.uk/

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref6

http://refhub.elsevier.com/S0033-3506(19)30090-3/sref7

mailto:publichealth@rsph.org.uk

https://doi.org/10.1016/j.puhe.2019.03.010

Prevention: of what, in whom, and how?
References

The-caries-related-cost-and-effects-of-a-tax-on-sugar-sweete_2019_Public-Hea

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
The caries-related cost and effects of a tax on
sugar-sweetened beverages
M. Jevdjevic a,*, A.-L. Trescherb, M. Rovers c,d, S. Listl a,b

a Department of Quality and Safety of Oral Healthcare, Radboud UMC, Philips van Leydenlaan 25, 6525 EX

Nijmegen, the Netherlands
b Department of Conservative Dentistry, Translational Health Economics Group, Heidelberg University,

Im Neuenheimer Feld 400, 69120 Heidelberg, Germany
c Department of Operating Rooms, Radboud UMC, Geert Grooteplein Zuid 10, 6525 GA Nijmegen, the Netherlands
d Department of Health Evidence, Radboud UMC, Geert Grooteplein Zuid 10, 6525 GA Nijmegen, the Netherlands
a r t i c l e i n f o

Article history:

Received 24 August 2018

Received in revised form

17 January 2019

Accepted 2 February 2019

Available online 16 March 2019

Keywords:

Caries burden

Sugar-sweetened beverages

Taxation

Cost-effectiveness

Markov model
* Corresponding author. Department of Qu
Nijmegen, the Netherlands. Tel.: þ31 645480

E-mail address: milica.jevdjevic@radboud
https://doi.org/10.1016/j.puhe.2019.02.010
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objectives: While taxes on sugar-sweetened beverages (SSBs) have frequently been pro-

posed to reduce non-communicable diseases like obesity and type 2 diabetes, relatively

little is known about the caries-related impacts of SSB taxation. We assessed the effect of a

20% ad valorem tax on SSBs on dental caries and related treatment costs, specifically taking

into account that consumers may switch from SSBs to other (non-taxed) sugar-containing

drinks.

Study design: Cost-effectiveness analysis.

Methods: A tooth-level Markov model was developed to evaluate the cost and effects of SSB

taxation. Tax-related changes in sugar consumption were calculated using available evi-

dence on SSBs price and cross-price elasticities, thereby taking changes in drinks con-

sumption behaviors into account. The model was used to establish lifetime disease-free

tooth years, caries lesions prevented, caries-related treatment costs avoided, tax revenues,

and administrative costs (reference case: the Netherlands). Deterministic and probabilistic

sensitivity analyses were performed to address uncertainties.

Results: A 20% SSB taxation would result in an average of 2.13 (95% uncertainty interval [UI]

2.12e2.13) caries-free tooth years per person and, on population level, prevention of

1,030,163 (95% UI 1,027,903e1,032,423) caries lesions. The intervention was found to save an

aggregate total of V 159.01 (95% UI 158.67e159.35) million in terms of dental care expen-

ditures. The estimated lifetime tax revenues (V3.49billion) were larger than the adminis-

trative costs for taxation (V37.3 million).

Conclusions: Our results show that SSB taxation may substantially improve oral health and

reduce the caries-related economic burden. Benefits would be the greatest for younger age

groups.

© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
ality and Safety of Oral
07.
umc.nl (M. Jevdjevic).

ic Health. Published by E
Healthcare, Radboud UMC, Philips van Leydenlaan 25, 6525 EX

mailto:milica.jevdjevic@radboudumc.nl

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.010&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.02.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2126
Introduction

Dental caries remains one of the largest public health con-

cerns of the 21st century, and it affects more than 2.5 billion

people worldwide.1 Beyond high repercussions on quality of

life and social functioning, it implies a considerable economic

burden to society.2 Excessive intake of sugar is a major risk

factor for caries development.3 For prevention of sugar-

related diseases (including caries), the World Health Organi-

zation (WHO) recommends that sugar intake should not

exceed 10% of total energy intake and should ideally be no

more than 5% of total energy intake.4

Sugar-sweetened beverages (SSBs) are one of the main

sources of added sugars and have become more and more

affordable throughout the last two decades.5 Considering the

adverse effects of SSBs, this raises serious public health con-

cerns, and several interventions have been proposed to reduce

SSB consumption. Fiscal policies, mainly additional ad val-

orem taxation, have increasingly been advocated for and are

already implemented in various settings worldwide.6 Some

studies have demonstrated the positive impacts of such a

health policy strategy on obesity and diabetes in terms of

health benefits, reduced treatment costs, and fewer produc-

tivity losses.7e10 So far, however, little is known about the

caries-related impacts of SSB taxation and about the extent to

which the effects of SSB taxation may be influenced by con-

sumers replacing SSBs by other (non-taxed) sugar-containing

drinks. Therefore, the purpose of the present study was to

assess the potential caries-related effects of introducing a 20%

ad valorem tax on SSBs.
Methods

We assessed, from a societal perspective, the potential health

economic impact of introducing a 20% ad valorem tax on SSBs
Fig. 1 e Conceptual model of the SSB taxation an
with respect to occurrence of dental caries and associated

treatment costs. Using a Markov decision-analytic model, SSB

taxation was compared with the current standard of no SSB

taxation. Markov models are stochastic state-transition

models that enable synthesis and analysis of the published

evidence, considering benefits, harms, and costs to support

decision-making under conditions of uncertainty.11 As shown

in Fig. 1, we assumed that the SSB taxation will result in

increased market prices, which will subsequently result in

changed demands for SSB (price elasticities). This will lead to

changes in sugar consumption and consequent reduction in

caries increment. In addition, tax revenues will be generated.

SSBs were defined to include liquids with added sugar, that is,

carbonated drinks, soft or isotonic drinks, fruit drinks, and

diluted syrups.

Target population and time horizon

We were interested in the effects on permanent dentition;

therefore, those aged younger than 6 years were not

included. The model was designed with reference to the

Dutch population aged 6e79 years in 2016, thereby aiming

at simulations which can be considered relevant for the

context of high-income countries. The composition of the

population, stratified by age and sex, was retrieved from

Statistics Netherlands.12 The demographic characteristics

are comparable to other Western European countries. The

model was run for a lifetime-horizon (i.e. until the 2016

Dutch population reached mean life expectancy). We chose

to model the Dutch population as the Netherlands is the

third country in the world with highest sugar consump-

tion.13 Only 10% of the Dutch children have sugar intake

below the WHO recommended value.14 Furthermore, pre-

ventable oral disorders belong to 10 health problems

responsible for the most disability in the Netherlands.15 In

2015, costs of dental health care in the Netherlands

exceeded V 3.75bn.2
d its effects. SSB, sugar-sweetened beverage.

https://doi.org/10.1016/j.puhe.2019.02.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2 127
SSB consumption in the Netherlands and (cross-) price
elasticities

Beverages consumption data were extracted from the Dutch

National Food Consumption Survey.14 This survey collected

periodical data on food consumption and nutrition status for a

representative sample of people living in the Netherlands. The

data were obtained at an individual level, which allowed us to

derive population-representative consumption estimates for

different age groups (6e8, 9e18, 19e50, and 51e79 years)

(Appendix Fig. A1). Non-alcoholic beverages of interest were

divided into two groups: (1) fruit juices (not subject to SSB

taxation) and (2) SSBs (subject to taxation): diluted syrups,

carbonated, soft, or isotonic drinks. After estimating the

change in volume of beverages consumption due to taxation,

changes in sugar intake were calculated according to bever-

ages sugar content provided by the Dutch Food Composition

Database (NEVO-online version 2016/5.0).16

Because there were no estimates of own and cross-

price elasticities available for SSB consumption in the

Netherlands, we primarily relied on estimates of price elas-

ticities as available from a meta-analysis (Appendix).10

Outcome

We estimated the number of caries lesions prevented and

caries-free tooth years gained. Caries-free tooth year repre-

sents a year spent without caries experience per tooth. Addi-

tionally, we calculated the avoided treatment costs, the

administrative costs of SSB taxation, and tax revenues (see

below for more details).

Modeling

We developed a tooth-level Markov state-transition model to

evaluate the cost and effects of SSB taxation on caries and

oral health care (Fig. 2). Transition patterns were adjusted
Fig. 2 e Conceptual framework of the Markov modeldtooth expos

transitions between Markov states (states of a tooth within the res

readability not all possible options are included. The full transiti
according the Dutch clinical routine and implemented in the

model. We applied a 6-month cycle length, assuming that

dental check-ups would take place every 6 months.17 In case a

need for treatment is determined during diagnostic assess-

ment, we assumed that the required treatments are carried

out immediately. Hence, the required treatment is always

provided at the start of a new state.

Probabilities

Transition probabilities among the different Markov states

were derived from literature (Appendix Table A1). We

assumed that the cohort starts in the caries-free state, and in

case of developing caries within a standard dental visit every 6

months, restoration would be the only possible treatment.

Repairment of the original restoration was defined as partial

replacement involving only one surface. Replacement of

restoration was assumed to generate one additional treated

surface.18 Endodontic re-treatment was not considered within

our model. An individual-level perspective generated through

a simple aggregation of the tooth-level model was adjusted for

the time of permanent teeth eruption and the impact of

declining number of teeth over the life course.19,20

The data on caries incidence were obtained from the

publicly available open-source platform of the Institute for

Health Metrics and Evaluation.15 The relationship between

the amount of consumed sugar and caries was derived from

the 11-years long Finnish longitudinal study.21 We calculated

the 6-month probability of caries development per tooth for

every 10 g of sugar additionally consumed on a daily basis

(Appendix).

Cost

Costs for different treatment modalities were extracted from

the national price list for dental services.22 Additionally, the

dental technician fee was added to the dental costs for crown
ure to caries and dental health care. Arrows indicate possible

torative cycle) represented by the circles. However, for better

on list is provided in Appendix (Table A1).

https://doi.org/10.1016/j.puhe.2019.02.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2128
based on the price list of a large University Policlinic in the

Netherlands (Table 1). Tax revenues of a potential ad valorem

tax on SSBs were calculated based on estimated consump-

tion under taxation and 2015 beverage consumer prices.23

The latter were accumulated to the baseline year by adjust-

ing for inflation rates. Calculations were performed for the

life-time horizon of the current Dutch population. Sensitivity

checks were performed to account for different levels in

price elasticities of demand. According to Dutch average

administrative cost of tax collection, administrative expen-

ditures were assumed to be 1.07% of tax revenues from the

SSB taxation.24

Discounting of future health outcomes and costs

Following the guideline for economic evaluation in health care

in the Netherlands, future costs were discounted at 4%, and

future health effects were discounted at 1.5%.25

Sensitivity and scenario analyses

In our model, we considered uncertainty in several input

parameters, i.e. disease incidence, sugar content in beverages

of interest, transition probabilities, and dental treatment

costs (Appendix). To determine the extent of the uncertainty,

we employed Monte Carlo simulation using 5000 iterations.

This allows estimating the means and 95% uncertainty in-

tervals (UI). In the base case, we used mean values for the

price and cross-price elasticity for SSB and fruit juices re-

ported in the literature. Additionally, deterministic sensi-

tivity analysis was performed to address the uncertainty in

relationship between caries incidence and the amount of

sugar consumed using lower and upper bound for caries

incidence reduction (Appendix).

Owing to high uncertainty around the Dutch (Western

European) preferences for SSB beverages (price elasticity) and

fruit juices (cross-price elasticity), we examined alternative

scenarios, replacing the abovementioned mean values with

lower and upper bounds. Scenario 1 reflects the strongest

possible reaction of the Dutch population to taxation. Own

price elasticity of �3.87 for SSBs and cross-price elasticity for
fruit juices of 0.14 were used in the simulation. The least

elastic demand for SSB was evaluated in Scenario 2, taking the
Table 1 e Costs per dental intervention (V)a.

Course of treatment Cost (V)

Examination 20.44

1-Surface restoration 71.54

2-Surface restoration 84.99

3-Surface restoration 95.74

More than 3-suface restoration 114.57

Repair 1-surface restoration 71.54

Repair 2-surface restoration 71.54

Repair 3-surface restoration 71.54

Repair more than 3-surface restoration 71.54

Replace 1-surface restoration 84.99

a Nederlandse Zorgautoriteit, Tandartstarieven 2017.
b Radboudumc Department of Dentistry.
upper values for price elasticity and cross-price elasticity from

the literature (�0.69 for SSBs and 1.45 for fruit juices) as input
parameters in our model. To compare the impact of un-

certainties in different parameters on the outcomes of interest

(number of caries lesions prevented, amount of caries-free

tooth years gained, and amount of treatment costs avoided),

we created Tornado diagrams as a graphical presentation of

sensitivity of our results.

All analyses were performed using the TreeAge Pro

Healthcare Module 2017 software (TreeAge Software, Inc.,

Williamstown, MA).
Results

In the base case scenario, introducing the 20% SSB taxation

would prevent development of 1,030,163 (95% UI

1,027,903e1,032,423) caries lesions (Table 2), which com-

prises an absolute reduction of 0.55% as compared with the

current situation. On a person level, each individual in the

population will on average benefit with 2.13 (95% UI

2.12e2.13) caries-free tooth years. With a lifetime horizon, a

total of V 159.01 (95% UI 158.67e159.35) million caries-

related treatment cost will be saved on a population level.

For boys aged 6e12 years, the intervention would be most

beneficial with 162,213 (95% UI 161,095e163,330) caries le-

sions prevented and 6.17 (95% UI 6.14e6.20) millions of

caries-free tooth years gained. In girls and women, the

benefits in terms of caries-free tooth years per person are

lower as compared with boys and men, 1.64 and 2.61 caries-

free tooth years, respectively. The estimated lifetime tax

revenues (V3.49bn) were larger than the administrative

costs for tax collection (V37.30million).

The deterministic sensitivity analysis showed that the

model was highly sensitive for caries reduction input values.

Fig. 3 shows the number of caries lesion prevented, ranging

from 313,516 for the lowest reported caries incidence reduc-

tion to 1,549,627 for the highest reported value. Avoided

treatment costs ranged from V49.84 million to V238.83 million

(Fig. 4), and the total amount of caries-free tooth years ranged

from 12.86 million to 48.19 as shown in Fig. 6.

In the alternative scenario with lower values for SSB price

and cross-price elasticity (scenario 1, �3.87 for SSB and 0.14
Course of treatment Cost (V)

Replace 2-surface restoration 95.74

Replace 3-surface restoration 114.57

Replace more than 3-surface restoration 114.57

Endodontic th þ restoration 359.3
Endodontic th þ crownb 948.64
Crownb 632.37

Crown replacementb 659.26

Crown recementation 21.51

Extraction 55.4

https://doi.org/10.1016/j.puhe.2019.02.010

Table 2 e Oral health benefits and treatment costs avoided owing to 20% SSB taxation, lifetime horizon, mean values for
price elasticities, and mean value for caries incidence reduction (95% uncertainty level).

Population Caries lesions prevented
(total)

Caries-free tooth years
(per person)

Caries-free tooth years
(total, million)

Treatment costs avoided
(million V)

Boys & men

Boys aged 6e12

years

162,213 (161,095e163,330) 9.07 (9.03e9.11) 6.17 (6.14e6.20) 23.83 (23.73e23.93)

Men aged 13e18

years

103,270 (102,941e103,598) 5.82 (5.80e5.83) 3.64 (3.63e3.65) 15.90 (15.86e15.95)

Men aged 19e35

years

197,449 (196,819e198,079) 3.42 (3.40e3.44) 6.16 (6.13e6.19) 32.32 (32.20e32.44)

Men aged 36e55

years

134,201 (132,873e135,530) 1.33 (1.32e1.35) 3.16 (3.12e3.20) 20.23 (20.00e20.46)

Men aged 56e79

years

45,929 (45,662e46,195) 0.32 (0.32e0.32) 0.68 (0.67e0.69) 5.46 (5.42e5.51)

Girls & women

Girls aged 6e12

years

101,068 (100,329e101,806) 6.35 (6.31e6.38) 4.12 (4.10e4.14) 16.12 (16.05e16.20)

Women aged 13

e18 years

62,371 (62,173e62,570) 3.70 (3.69e3.71) 2.21 (2.21e2.22) 10.05 (10.01e10.08)

Women aged 19

e35 years

13,622 (123,244e124,001) 2.22 (2.21e2.23) 3.92 (3.90e3.94) 20.47 (20.40e20.55)

Women aged 36

e55 years

77,446 (76,523e78,369) 0.78 (0.77e0.79) 1.84 (1.81e1.87) 11.94 (11.78e12.10)

Women aged 56

e79 years

22,594 (22,464e22,725) 0.15 (0.15e0.15) 0.33 (0.33e0.34) 2.69 (2.67e2.71)

Total 1,030,163 (1,027,903e1,032,423) 2.13 (2.12e2.13) 32.25 (32.18e32.31) 159.01 (158.67e159.35)

SSB, sugar-sweetened beverage.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2 129
for fruit juices), taxation would yield 7.52 (95% UI 7.51e7.54)

caries-free tooth years per person (Fig. 5) and prevent 4.09

(95% UI 4.09e4.10) million caries lesions (Appendix Table A2).

SSB taxation would subsequently result in a saving of

V603.17 (95% UI 601.92e604.43) million dental healthcare

expenditures on a population level. Assuming a high reduc-

tion in demands for SSB as a response to increased taxes, tax

revenues will amount to V1.03 billion with V11.00 million of

administrative costs.

Even in case of a very strong preference for SSBs and a less

elastic demand (�0.69 for SSB and 1.45 for fruit juices), as
depicted in scenario 2, taxation appears to be a cost-saving

intervention resulting in 0.81 (95% UI 0.80e0.81) caries-free

tooth years per person, 350,694 (95% UI 349,658e351,730)

caries lesions prevented, and V59.44 (95% UI 59.27e59.27)

million as averted caries-attributable treatment costs

(Appendix Table A3). With this reaction to SSB taxation, the

tax revenues will be V3.96 billion, whereas the administration

of taxation will be V42.42 million.
Fig. 3 e Tornado diagram. The bars indicate the effect of

different input values for price elasticities and caries

incidence reduction on the number of caries lesions

prevented.
Discussion

The findings of our study show that a total of 1,030,163 (95% UI

1,027,903e1,032,423) caries lesions could be prevented with

SSB taxation, while gaining a total of 32.25 (95% UI

32.18e32.31) million of caries-free tooth years. The introduc-

tion of SSB taxation could potentially reduce caries-related

dental expenditures by V 159.01 (95% UI 158.67e159.35)

million over a lifetime horizon. The estimated lifetime tax

revenues (V3.49 billion) were estimated to be larger than the

administrative costs for taxation (V 37.3 million).

To our knowledge, this is the first study to assess the

potential benefits of SSB taxation on dental care, taking into

account short- and long-term consequences of taxation and

the whole caries treatment cycle until potential tooth loss.

The major strength of this study is the use of country-

specific data for beverages consumption and caries inci-

dence, stratified by age and sex. Moreover, to estimate

taxation-related changes in sugar consumption, we
Fig. 4 e Tornado diagram. The bars indicate the effect of

different input values for price elasticities and caries

incidence reduction on avoided treatment costs.

https://doi.org/10.1016/j.puhe.2019.02.010

Fig. 5 e Tornado diagram. The bars indicate the effect of

different input values for price elasticities and caries

incidence reduction on the amount of caries-free tooth

years gained per person.

Fig. 6 e Tornado diagram. The bars indicate the effect of

different input values for price elasticities and caries

incidence reduction on the total amount of caries-free

tooth years.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2130
accounted for reduced SSB intake and potential substitution

with fruit juice. To arrive at more precise estimates, tooth

eruption time and a specific tooth loss trajectory were

incorporated in the model.

Some potential limitations should also be discussed. First,

country-specific estimates for SSB price and cross-price

elasticities were not available. However, the range of values

reported in the literature was employed and evaluated in

alternative scenarios. Second, owing to lack of more detailed

evidence, possible substitution with sugar-containing foods

was not considered. Therefore, we should be aware that the

magnitude of the effect of SSB taxation is highly dependent

on people’s behaviors affecting absolute sugar intake. Price
elasticities for SSBs and fruit juices were derived from panel-

based data, mostly containing purchases for at-home con-

sumption. Hence, in case of strong consumers’ preferences
for SSBs within away from home markets (e.g. restaurants),

total sugar intake as assumed in our study might be under-

estimated. Another important consideration to be taken into

account are meal deals that could diminish the price differ-

ence and motivate consumers to select high-sugar products,

which they perceive as the most expensive alternative,

maximize the cost-benefit ratio of the purchase.26 This could

particularly affect the amount of health benefits we esti-

mated for children and the population under the age of 18

years as they are the most frequent consumers of combo

meals.27 A SSB tax may potentially induce a reduction in drink

waste, as suggested by Smith et al.28 Yet, this effect has not

been quantified. In our study, we assumed a constant ratio

between the purchased and consumed amount of SSBs.

Further research should address the uncertainty around

people’s preferences, substitution patterns with sugar-
containing beverages and foods. Third, we assumed the

same sensitivity to taxation across the whole population.
With this assumption, potential differences within educa-

tional or income groups are neglected. Low socio-economic

households possibly have higher reductions in purchasing

SSBs reflecting the differential taxation impact, as found in a

recent study by Colchero et al.29 In case the differential

taxation impact is absent, the SSB taxation would be regres-

sive imposing the highest burden on the economically most

vulnerable citizens. Fourth, societal changes will likely result

in availability of alternative drinks with reduced or no added

sugar. It implies that the effect of SSB taxation will not be

constant over the entire life course, as presupposed in our

study. Finally, our results are simulation-based, and all inputs

for beverage consumption and costs are based on Dutch data.

The consumption of SSBs and fruit juices in the Netherlands

is, however, comparable to consumption in United Kingdom,

Ireland, Germany, France, Spain, Portugal, and Australia.30

The price elasticities for SSBs and fruit juices that we used

for our simulation were pooled through a meta-analysis that

was based on several studies performed in different high-

income settings. Therefore, these estimates are also appli-

cable to the abovementioned countries.31 Healthcare prices,

administrative costs of taxation as well as tax revenues may

also differ across countries. In addition, disease-related pa-

rameters (e.g. caries incidence) for the Netherlands are

similar to the rest of the Western European countries or

Australia.15 Taking into account the aforementioned extent of

generalizability of input parameters, simulations for these

countries would most likely yield similar results with regard

to health benefits in Australia and Western Europe. In

contrast, sugary drinks consumption is higher in the United

States and Canada, thus analyses with country-specific pa-

rameters for these settings are likely to produce different

results. However, given the detailed presentation of the

model and its input parameters, those interested can

straightforwardly assess the transferability of the cost esti-

mates to their specific situation.

Our results are in agreement with previous studies that have

shown the reduction in caries incidence and decayed, missing,

or filled teeth (DMFT) increment due to SSB taxation.9,32,33

Though, none of the studies accounted for any further conse-

quences of caries except the restoration placement. Sowa et al.

estimated that SSB taxation in Australia would result in 3.9

million units of DMFT averted and V0.43 billion in cost saved

over 10 years.33 Nevertheless, in their study, they did not

consider potential substitution with sugar-containing bever-

ages not subjected to taxation. Schwendicke et al. reported 0.75

million of caries lesions prevented and avoided treatment costs

of V0.8 billion over a 10-year horizon in Germany.32 A specific

novelty of our study is that the effects on oral health are illus-

trated by the total number of caries lesions prevented and

caries-free tooth years gained. Considering these two out-

comes, we were able to analyze the effect of taxation on oral

health in a more comprehensive way taking into account pat-

terns from clinical practice. By postponing caries onset and

entering the restorative cycle in the later stage in life, more

invasive treatments could be avoided and consequently lead to

prevention of tooth loss.

Clinical prevention and dental treatment severely affect

both the financial capacity of individuals and already limited

healthcare budgets. Moreover, these interventions are not

https://doi.org/10.1016/j.puhe.2019.02.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2 131
affordable to all people in need. Improvements in oral health

are most likely to be achieved through population-based in-

terventions. In our study, we considered only the effects on

oral health and even then SSB taxation appears to be a cost-

saving intervention. That could be a good starting point for

future dialogs with health policy-makers. Thereby, it could

also be highlight thatdindependent of arguable cost savings

due to lower treatment needs in response to SSB tax-

ationdreduced caries burdens would be beneficial in terms

of reducing absence from work and school. To arrive at the

more precise estimation of overall potential impact of taxa-

tion, all diseases with a common risk factor should be

included. However, several implementation challenges

should be considered. Policy-makers might be reluctant to

introduce SSB taxation because of high uncertainties or

absence of empirical evidence. In addition, such an inter-

vention can be seen as an intrusion in individual autonomy

to make choices and contribute to the mistrust of true

intention of taxation.34 Briggs et al. have shown that the in-

dustry response is of substantial importance for the actual

SSB tax effect.9 However, its exact reaction cannot be pre-

cisely predicted. Currently, available evidence from natural

experiments suggests that the industry could opt for refor-

mulation of products by reducing the amount of sugar in

high- and mid-sugar drinks as done by Lucozade Ribena

Suntory and Tesco when tiered taxes are introduced.35,36

Based on the findings from a UK modeling study, this sce-

nario would result in the highest amount of health benefits

gained.9 Similarly, producers could aim for innovation by

developing new versions of sugar-free alternatives and

shifting their marketing strategies toward healthier brands.37

This may eventually reduce sugar consumption even further.

However, recently estimated cross-price elasticities for reg-

ular soft drinks and their diet versions have shown that these

products are rather complements than substitutes.38 Also

note that both regular and diet sodas may imply risks of

dental erosion.39 In addition, the effects of SSB taxation could

be offset by price increases for other drinks. Dependent on

setting, the impact of SSB taxation could potentially also be

diminished by cross-border shopping. For example, a recent

study has shown that after a tax was introduced in Berkeley,

California, SSB purchases increased by 7% in surrounding

regions without this fiscal policy.40 All important de-

terminants should be considered through engagement of

relevant stakeholders to secure that maximum of health

benefits is achieved.

In conclusion, our model shows that SSB taxation may

substantially improve oral health and reduce the caries-

related economic burden.
Authors statements

Ethical approval

None sought.

Funding

None declared.
Competing interest

The authors declare no potential conflicts of interest with

respect to the authorship and publication of this manuscript.

Author contribution

M.J. had full access to all the data used in this study and takes

responsibility for the integrity of the data and accuracy of its

analysis. M.J. and A.T. contributed to design, data acquisition,

analysis, and interpretation, drafted and critically revised the

manuscript; M.R. contributed to interpretation, drafted, and

critically revised the manuscript; S.L. contributed to design,

interpretation, drafted, and critically revised the manuscript.

All authors read and approved the final manuscript.
r e f e r e n c e s

1. Kassebaum NJ, Smith AGC, Bernab�e E, Fleming TD,
Reynolds AE, Vos T, et al. Global, regional, and national
prevalence, incidence, and disability-adjusted life years for
oral conditions for 195 countries, 1990e2015: a systematic
analysis for the global burden of diseases, injuries, and risk
factors. J Dent Res 2017 Apr 1;96(4):380e7.

2. Righolt AJ, Jevdjevic M, Marcenes W, Listl S. Global-, regional-,
and country-level economic impacts of dental diseases in
2015. J Dent Res 2018;97(5):501e7.

3. Moynihan PJ, Kelly SAM. Effect on caries of restricting sugars
intake. J Dent Res 2014;93(1):8e18.

4. World Health Organization. Guideline: sugars intake for adults
and children. Geneva. 2015.

5. Blecher E, Liber AC, Drope JM, Nguyen B, Stoklosa M. Global
trends in the affordability of sugar-sweetened beverages,
1990e2016. Prev Chronic Dis 2017;14:E37.

6. World Health Organization. Fiscal policies for diet and prevention
of noncommunicable diseases: technical meeting report [Internet].
Geneva. 2016 [cited 2018 Nov 20]. Available from: http://apps.
who.int/iris/bitstream/handle/10665/250131/9789241511247-
eng ?sequence¼1.

7. Lal A, Mantilla-Herrera AM, Veerman L, Backholer K, Sacks G,
Moodie M, et al. Modelled health benefits of a sugar-
sweetened beverage tax across different socioeconomic
groups in Australia: a cost-effectiveness and equity analysis.
PLoS Med 2017 Jun;14(6):e1002326.

8. Barrientos-Gutierrez T, Zepeda-Tello R, Rodrigues ER,
Colchero-Aragon�es A, Rojas-Martı́nez R, Lazcano-Ponce E,
et al. Expected population weight and diabetes impact of the
1-peso-per-litre tax to sugar sweetened beverages in Mexico.
PLoS One 2017;12(5):e0176336.

9. Briggs ADM, Mytton OT, Kehlbacher A, Tiffin R, Elhussein A,
Rayner M, et al. Health impact assessment of the UK soft
drinks industry levy: a comparative risk assessment
modelling study. Lancet Public Heal 2017;2(1):e15e22.

10. Long MW, Gortmaker SL, Ward ZJ, Resch SC, Moodie ML,
Sacks G, et al. Cost effectiveness of a sugar-sweetened
beverage excise tax in the U.S. Am J Prev Med
2015;49(1):112e23.

11. Sonnenberg FA, Beck JR. Markov models in medical decision
making. Med Decis Mak 1993;13(4):322e38.

12. Centraal Bureau voor de Statistiek. CBS StatLine Database
[Internet]. 2017 [cited 2017 Oct 23]. Available from: http://
statline.cbs.nl/Statweb/publication/?VW¼T&DM¼SLEN&PA¼
03743ENG&D1¼0&D2¼a&D3¼0&D4¼a&HD¼171023-1136&LA¼
EN&HDR¼T,G3&STB¼G1,G2.

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref1

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref2

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref3

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref4

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref5

http://apps.who.int/iris/bitstream/handle/10665/250131/9789241511247-eng ?sequence=1

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref7

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref8

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref9

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref10

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref11

http://statline.cbs.nl/Statweb/publication/?VW=T&DM=SLEN&PA=03743ENG&D1=0&D2=a&D3=0&D4=a&HD=171023-1136&LA=EN&HDR=T,G3&STB=G1,G2

https://doi.org/10.1016/j.puhe.2019.02.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 2 5 e1 3 2132
13. Pariona A. Top sugar consuming nations in the world [Internet].
2017 [cited 2018 Jun 13]. Available from: https://www.
worldatlas.com/articles/top-sugar-consuming-nations-in-
the-world.html.

14. van Rossum C, Buurma-Rethans E, Vennemann F, Beukers M,
Brants H, de Boer E, et al. The diet of the Dutch: results of the first
two years of the Dutch National Food Consumption Survey
2012e2016 [Internet]. 2016 [cited 2018 Jun 13]. Available from:
https://www.rivm.nl/dsresource?objectid¼20ff527e-10dd-
4d5d-a4a1-a43815be2b5f&type¼pdf&disposition¼inline.

15. Institute for Health Metrics and Evaluation. Global health
data exchange: GBD results tool [Internet]. [cited 2018 Jun 13].
Available from: http://ghdx.healthdata.org/gbd-results-tool.

16. Dutch Food Consumption Database (NEVO online version
2016/5.0) [Internet]. [cited 2017 Oct 19]. Available from:
https://nevo-online.rivm.nl/.

17. OECD. OECD data (Online-Database) [Internet]. 2018 [cited 2018
Jun 15]. Available from: https://stats.oecd.org/viewhtml.aspx?
datasetcode¼HEALTH_PROC&lang¼en#.

18. Brantley CF, Bader JD, Shugars DA, Nesbit SP. Does the cycle
of rerestoration lead to larger restorations? J Am Dent Assoc
1995;126(10):1407e13.

19. American Dental Association. Permanent tooth development
[Internet]. 2012 [cited 2017 Oct 15]. Available from: https://
www.mouthhealthy.org/~/media/MouthHealthy/Files/Kids_
Section/ADAPermanentTeethDev_Eng ?la¼en.

20. Stock C, Jürges H, Shen J, Bozorgmehr K, Listl S. A comparison
of tooth retention and replacement across 15 countries in the
over-50s. Community Dent Oral Epidemiol 2015;44(3):223e31.

21. Bernab�e E, Vehkalahti MM, Sheiham A, Lundqvist A,
Suominen AL. The shape of the dose-response relationship
between sugars and caries in adults. J Dent Res
2016;95(2):167e72.

22. Zorgautoriteit Nederlandse. Tariefbeschikking: Tandheelkundige
zorg [Internet]. 2017 [cited 2018 Jun 13]. Available from: https://
puc.overheid.nl/nza/doc/PUC_6078_22/1/.

23. Eurostat. Detailed average prices report: an analysis into
measurement of detailed average prices for consumer products
[Internet]. Brussels. 2015. Available from: http://ec.europa.eu/
eurostat/documents/272892/272992/Consumer-price-
research-2015/.

24. OECD. Tax administration 2015: comparative information on OECD
and other advanced and emerging economies [Internet]. Paris:
OECD Publishing; 2015 (Tax Administration). Available from:
http://www.oecd-ilibrary.org/taxation/tax-administration-
2015_tax_admin-2015-en.

25. Zoorginstituut Nederland. Guideline for economic evaluations in
healthcare [Internet]. 2016 [cited 2018 Jun 13]. Available from:
https://english.zorginstituutnederland.nl/publications/
reports/2016/06/16/guideline-for-economic-evaluations-in-
healthcare.

26. Williams C. Meal deals could undo the benefits of the sugar tax.
The Conversation [Internet]. 2018 Feb 9 [cited 2019 Jan 15];
Available from: https://theconversation.com/meal-deals-
could-undo-the-benefits-of-the-sugar-tax-91136.

27. Cantor J, Breck A, Elbel B. Correlates of sugar-sweetened
beverages purchased for children at fast-food restaurants. Am
J Public Health 2016;106(11):2038e41.

28. Smith E, Scarborough P, Rayner M, Briggs ADM. Should we tax
unhealthy food and drink? Proc Nutr Soc 2018;77:314e20.
29. Colchero MA, Popkin BM, Rivera JA, Ng SW. Beverage
purchases from stores in Mexico under the excise tax on
sugar sweetened beverages: observational study. BMJ
2016;352:h6704.

30. Singh GM, Micha R, Khatibzadeh S, Shi P, Lim S, Andrews KG,
et al. Global, regional, and national consumption of sugar-
sweetened beverages, fruit juices, and milk: a systematic
assessment of beverage intake in 187 countries. PLoS One 2015
Aug;10(8):e0124845. Müller M, editor.

31. Seale J, Regmi A, Bernstein J. International evidence on food
consumption patterns [Internet]. 2003 [cited 2018 Jul 6]. Available
from: www.ers.usda.gov.

32. Schwendicke F, Thomson WM, Broadbent JM, Stolpe M.
Effects of taxing sugar-sweetened beverages on caries and
treatment costs. J Dent Res 2016;95(12):1327e32.

33. Sowa PM, Keller E, Stormon N, Lalloo R, Ford PJ. The
impact of a sugar-sweetened beverages tax on oral
health and costs of dental care in Australia. Eur J Public
Health 2019 Feb 1;29(1):173e7. https://doi.org/10.1093/
eurpub/cky087.

34. Thomas-Meyer M, Mytton O, Adams J. Public responses to
proposals for a tax on sugar-sweetened beverages: a thematic
analysis of online reader comments posted on major UK
news websites. PLoS One 2017;12(11):e0186750.

35. Tesco PLC. Tesco reduces sugar content in all own brand soft
drinks [News release Nov 7, 2016] [Internet]. 2016 [cited 2019 Jan
15]. Available from: https://www.tescoplc.com/news/news-
releases/2016/tesco-reduces-sugar-content-in-all-own-
brand-soft-drinks.

36. Lucozade Ribena Suntory. Sugar reduction and health &
wellbeing [Internet]. 2019 [cited 2019 Jan 15]. Available from:
https://www.lrsuntory.com/health-and-wellbeing/sugar-
reduction/.

37. The Coca-Cola Company. Our way forward 2017 update [press
release Apr 25, 2018] [Internet]. 2018 [cited 2019 Jan 15].
Available from: https://www.coca-colacompany.com/stories/
our-way-forward-2017-update.

38. Dharmasena S, Capps O. Intended and unintended
consequences of a proposed national tax on sugar-sweetened
beverages to combat the U.S. obesity problem. Health Econ
2012 Jun 1;21(6):669e94.

39. Ehlen LA, Marshall TA, Qian F, Wefel JS, Warren JJ. Acidic
beverages increase the risk of in vitro tooth erosion. Nutr Res
[Internet] 2008 May;28(5):299e303 [cited 2019 Jan 17].
Available from: http://www.ncbi.nlm.nih.gov/pubmed/
19083423.

40. Silver LD, Ng SW, Ryan-Ibarra S, Taillie LS, Induni M,
Miles DR, et al. Changes in prices, sales, consumer spending,
and beverage consumption one year after a tax on sugar-
sweetened beverages in Berkeley, California, US: a before-
and-after study. PLoS Med 2017 Apr 18;14(4):e1002283.
Langenberg C, editor.
Appendix A. Supplementary data

Supplementary data to this article can be found online at

https://doi.org/10.1016/j.puhe.2019.02.010.

https://www.worldatlas.com/articles/top-sugar-consuming-nations-in-the-world.html

https://www.rivm.nl/dsresource?objectid=20ff527e-10dd-4d5d-a4a1-a43815be2b5f&type=pdf&disposition=inline

http://ghdx.healthdata.org/gbd-results-tool

https://nevo-online.rivm.nl/

https://stats.oecd.org/viewhtml.aspx?datasetcode=HEALTH_PROC&lang=en#

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref18

https://www.mouthhealthy.org/%7E/media/MouthHealthy/Files/Kids_Section/ADAPermanentTeethDev_Eng ?la=en

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref20

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref21

https://puc.overheid.nl/nza/doc/PUC_6078_22/1/

http://ec.europa.eu/eurostat/documents/272892/272992/Consumer-price-research-2015/

http://www.oecd-ilibrary.org/taxation/tax-administration-2015_tax_admin-2015-en

https://english.zorginstituutnederland.nl/publications/reports/2016/06/16/guideline-for-economic-evaluations-in-healthcare

https://theconversation.com/meal-deals-could-undo-the-benefits-of-the-sugar-tax-91136

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref27

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref28

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref29

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref30

http://www.ers.usda.gov

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref32

https://doi.org/10.1093/eurpub/cky087

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref34

https://www.tescoplc.com/news/news-releases/2016/tesco-reduces-sugar-content-in-all-own-brand-soft-drinks

https://www.lrsuntory.com/health-and-wellbeing/sugar-reduction/

https://www.coca-colacompany.com/stories/our-way-forward-2017-update

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref38

http://www.ncbi.nlm.nih.gov/pubmed/19083423

http://refhub.elsevier.com/S0033-3506(19)30034-4/sref40

https://doi.org/10.1016/j.puhe.2019.02.010

The caries-related cost and effects of a tax on sugar-sweetened beverages
Introduction
Methods
Target population and time horizon
SSB consumption in the Netherlands and (cross-) price elasticities
Outcome
Modeling
Probabilities
Cost
Discounting of future health outcomes and costs
Sensitivity and scenario analyses

Results
Discussion
Authors statements
Ethical approval
Funding
Competing interest
Author contribution

References
Appendix A. Supplementary data

Do-anxiety-or-determination-of-life-differ-based-on-the-perceiv_2019_Public-

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Do anxiety or determination of life differ based on
the perceived financial status to cope with severe
diseases?
E. Lee a, J. Cho b,*

a Department of Nursing, Hoseo University, Asan, South Korea
b Department of Nursing, College of Medicine, Inje University, Busan, South Korea
a r t i c l e i n f o

Article history:

Received 21 September 2018

Received in revised form

16 December 2018

Accepted 15 January 2019

Available online 20 March 2019

Keywords:

Perceived financial status

Severe disease

Anxiety

Determination of life

Household income
* Corresponding author. Department of Nur
Korea. Tel.: þ82 51 890 6233; fax: þ82 51 896

E-mail address: jhcho@inje.ac.kr (J. Cho).
https://doi.org/10.1016/j.puhe.2019.01.010
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objective: The objective of this study was to identify the relationships among people’s

perceived financial status to cope with severe disease, levels of anxiety and determinations

of life.

Study design: This is a secondary analysis of population-based cross-sectional surveys.

Methods: The 2016 Social Integration Survey of 8000 Korean participants aged 19 years or

older was used. Data were analysed using correlation, correspondence and covariate

analyses.

Results: Of all the participants, 84.6% responded that they had insufficient perceptions of

financial status; decision-making power was found to have a stronger correlation with

perceived financial stability than with real income. In addition, the perceived ability, based

on financial status, to cope with severe disease was correlated with anxiety.

Conclusions: The study proposes that when developing health and medical treatment policy

and intervention programmes, perceptions of personal financial status and stability should

be considered concurrently.

© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
Approximately 1.6 million Korean people suffer from

Introduction

With the increased average lifespan and ratio of senescence in

the life cycle, improved quality of life is becoming increasingly

important. The burden of health expenditures may result in

lower quality of life and poverty.1,2 Particularly, the sudden

increase in health expenditure associated with severe dis-

eases not only affects household finances but also decreases

the quality of life.3
sing, College of Medicine
9840.

ic Health. Published by E
severe diseases, and 8.6 trillion of the total 54 trillion KRW

health expenditure is used for the four major severe dis-

eases, namely, cancer, cardiovascular disease, cerebrovas-

cular disease and rare diseases. Furthermore, the average

health expenditure of 4e10 million KRW4 per patient is

indicative of the significant health expenditure burden

of severe diseases. In a study by Lee,5 72.2% and 47.2%

of patients with severe diseases felt burdened by and
, Inje University, 75, Bokji-ro, Busanjin-gu, Busan, 47392, South

mailto:jhcho@inje.ac.kr

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.010&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9134
experienced excessive pressure from health expenditure,

respectively; only 27.8% of patients could cover their own

health expenditure. Previous studies on catastrophic health

expenditure claimed that an excessive out-of-pocket pay-

ment in the use of medical services may decrease normal

consumption or lead to poverty,6,7 which generates the

need for a policy alternative for financial protection of

households, especially for the economic burden of severe

diseases.8

Financial status is usually measured by variables such as

income and socio-economic status; however, income is not

an appropriate measure of financial well-being because it

cannot sufficiently explain perceived financial status.9 People

tend to base their attitudes and behaviours on subjective

perception.10 According to Jackman and Jackman,11 the

meaning or substance of the stratum cannot be evaluated

without paying attention to subjective perception. Perceived

financial status is a notion distinct from income in that it

takes into account expenditure relative to income. Perceived

financial status refers to an individual’s perception of his or
her ability to meet expenses and a tendency to worry about

debt, among other factors.12 Han et al.13 emphasised that the

subjective perception of financial status is a major factor that

affects the planning of economic life in later years. In view of

this, there is a need to determine the extent of an individual’s
perception of his or her financial ability to meet expenses in

case of severe disease, and people’s perception of their own
financial status should be reflected in policy making

regarding severe disease.

Various studies indicate that most households have

already faced catastrophic health expenditure due to severe

diseases5,7,14 or have discussed the equity of health expen-

diture based on categorisation of income brackets.15 In

situations of decreasing productivity and increasing

health expenditure due to severe disease, it is necessary

to examine the extent of people’s perception of their
financial ability to meet expenses in case of severe disease

and the relationship between the objective ‘household

income’ and the subjective ‘perceived financial status’. On a

similar note, perceived financial status is reported to be

associated with emotional well-being and quality of life. In

this regard, there is a need to study how the variables of

income, financial stability and perceived financial status are

associated with subjective well-being and life-determining

power.

Against this background, this study pursues the following

objectives based on data from the 2016 Korea Social Integra-

tion Survey: 1) to determine the association between not only

household income as an objective indicator but also perceived

financial status to meet expenses in case of serious disease as

a subjective indicator and decision-making over life; 2) to

check whether pecuniary anxiety and decision-making over

life are influenced by the degree of perceived financial status

to meet expenses in case of severe disease. This analysis will

enable determination of whether the perceived financial sta-

tus affects coping, which is difficult to assess from absolute

figures of household income or health expenditure alone.

Furthermore, it will establish baseline data for policy devel-

opment to promote determination of, and positive feedback

in, life.
Methods

Sampling and data collection

Data of the 2016 Social Integration Survey of the Korea Insti-

tute of Public Administration were used. This survey aimed to

provide baseline data for national policy that minimises social

conflict and contributes to national integration; it did so by

determining trends in perceptions and attitudes among citi-

zens regarding unity levels and by determining the percep-

tions of social unity experienced by citizens in each social

domain. The survey, conducted from September 1 to October

31, 2016, comprised a sample of 8000 participants from the

entire Korean population aged 19 years and above. The sam-

pling units were first classified by city/province, second by the

households in the study area and third by individuals. The

sample’s composition was controlled by comparing the sam-
ple size with the population by city/province, gender and age

group.

Measurements

Thefollowingquestionwas usedtoevaluateperceivedfinancial

status to cope with severe disease: ‘How much financial ability

do you think you have to cope with severe disease?’ Responses

were rated on a 4-point scale (highly insufficient, somewhat

insufficient, somewhat sufficient and highly sufficient). The

level of anxiety was determined using the following question:

‘How anxious were you yesterday?’, and the determination of

life was assessed using thefollowing question: ‘How free do you

think you are in determining your life?’ The perception of

financialstability was determined usingthefollowingquestion:

‘How stable do you think your current financial situation is?’

Thesubjectivelevelwasratedonan11-pointscaleranging from

‘not at all’ (0 points) to ‘very much’ (10 points). Income was

measured in 12 sections at an interval of 1 million KRW, from

0 KRW to 10 million KRW and higher; it was recorded as quan-

titative variables, represented by the median value of each

section, and as qualitative variables in seven classes ranging

from below 1 million KRW to 6 million KRW and higher.

Statistical analysis

A frequency analysis, independent sample t-test and analysis

of variance were used to determine the difference in the level

of anxiety and determination of life according to general

characteristics. Correlation and correspondence analyses

were conducted to examine the correlation among household

income, perceived financial stability and perceived financial

status to cope with severe disease. Simple regression analysis

was conducted to examine whether there was a difference in

the level of anxiety and determination of life according to

perceived financial stability and household income. Gender,

age, marital status and education level were used as control

variables to identify the relationships among perceived

financial status to cope with severe diseases and the level of

anxiety and determination of life, and covariate analysis

(linear model) was conducted using perception of financial

stability as a covariate.

https://doi.org/10.1016/j.puhe.2019.01.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9 135
Results

Differences between variables based on general
characteristics

Differences in the level of anxiety according to age, marital

status, education level, household income and perception of

financial stability, as well as differences in determination of

life according to age, marital status, and perception on

financial stability, were statistically significant (Table 1).

The level of anxiety was higher when age, education level,

household income and perception of financial stability were

lower; it was highest among married participants without

spouses, followed by single and married participants with

spouses. Older participants were freer in determination of life,

and participants in their 20s were relatively freer than those in

their 30s and 40s. Married people without spouses were freer

than singles, and a higher perception of financial stability led

to greater determination of life.

Household income and perceptions of financial stability and
financial status

Perceived financial status was considered to be insufficient by

84.6% of participants. The differences in perceived financial

status to cope with severe disease were statistically significant

according to age, marital status, education level and percep-

tion of financial stability (Table 2). In most cases, participants
Table 1 e General characteristics of participants and the differ
general characteristics (N ¼ 8000).
Characteristics Categories n

Gender Male 4067 5

Female 3933 4

Age in years 20s (including 19) 1582 1

30s 1610 2

40s 1875 2

50s 1792 2

60s or older 1141 1

Marital status Single 2204 2

Have a spouse (married) 5219 6

Do not have a spouse (married) 577

Education level Elementary-school graduate or lower 329

Middle-school graduate 570

High-school graduate 3806 4

College graduate or higher 3295 4

Household income Below 1 million KRW 5302 6

1e2 million KRW 2698 3

2e3 million KRW 388

3e4 million KRW 774

4e5 million KRW 1347 1

5e6 million KRW 1781 2

6 million KRW or higher 1404 1

Perception of financial

stabilitya
8e10 1101 1

4e7 1205 1

0e3 386

SD ¼ standard deviation.
a 11-point scale ranging from ‘not at all’ (0 points) to ‘very much’ (10 poi
in their 20s, singles, married people without spouses and

those with low education level considered their financial

status as highly insufficient or somewhat insufficient; the

latter response did not differ according to household income.

Even for participants with at least 6 million KRW in household

income, 15.5% perceived their financial status to be highly

insufficient.

A correlation analysis of perceived financial status to cope

with severe disease yielded correlation coefficients for house-

hold income and perception of financial stability of 0.189 and

0.505, respectively, indicating a closer relationship to subjec-

tive perception of financial stability than actual income.

A correspondence analysis to determine the correlation

between the class interval of income and perceived financial

status to cope with severe disease (Fig. 1) yielded the following

high correlations: between below 2 million KRW and ‘highly

insufficient’; 2e6 million KRW and ‘somewhat insufficient’;

and 6 million KRW or higher and ‘somewhat sufficient’. No

income class interval was correlated with ‘highly sufficient’.

Perceived financial status and the level of anxiety and
determination of life

A higher perception of financial stability was found to lead to a

lower level of anxiety and a higher determination of life. There

was no statistically significant difference in the level of anxi-

ety and determination of life according to income (Table 3).

Therefore, the perception of financial stability was selected as

a covariate.
ences in anxiety and determination of life according to

% Anxiety Determination of life

Mean SD t or F (p) Mean SD t or F (p)

0.8 4.06 2.27 �0.302 (0.762) 6.27 1.91 1.103 (0.270)
9.2 4.07 2.32 6.23 1.90

9.8 4.11 2.31 4.591 (0.001) 6.22 1.95 15.459 (<0.001) 0.1 4.18 2.33 6.09 1.93 3.4 4.13 2.24 6.09 1.89 2.4 4.01 2.27 6.41 1.87 4.3 3.83 2.35 6.52 1.81 7.5 4.15 2.29 7.030 (0.001) 6.26 1.95 2.929 (0.054) 5.2 4.00 2.27 6.23 1.87 7.2 4.33 2.50 6.43 1.99 4.1 4.25 2.61 9.578 (<0.001) 6.41 1.95 1.230 (0.297) 7.1 4.27 2.39 6.30 1.72 7.6 3.93 2.28 6.22 1.95 1.2 4.18 2.25 6.26 1.87 6.3 4.45 2.49 6.354 (<0.001) 6.20 2.06 0.657 (0.684) 3.7 4.27 2.38 6.33 2.01 4.9 4.10 2.38 6.23 1.97 9.7 3.92 2.36 6.24 1.93 6.8 4.19 2.15 6.20 1.80 2.3 3.85 2.13 6.30 1.77 7.6 4.04 2.26 6.27 1.91 3.8 3.22 2.38 104.99 (<0.001) 7.20 1.82 177.72 (<0.001) 5.1 3.86 2.17 6.38 1.74 4.8 4.97 2.44 5.58 2.23 nts). https://doi.org/10.1016/j.puhe.2019.01.010 https://doi.org/10.1016/j.puhe.2019.01.010 Table 2 e Perceived financial status to cope with severe disease according to general characteristics (N ¼ 8000). Characteristics Categories Highly insufficient Somewhat insufficient Somewhat sufficient Very sufficient Х2 or r p n (%) or mean (SD) Gender Male 948 (23.3) 2502 (61.5) 568 (14.0) 49 (1.2) 2.431 0.488 Female 908 (23.1) 2411 (61.3) 579 (14.7) 36 (0.9) Age in years 20s (including 19) 651 (41.2) 811 (51.3) 107 (6.8) 12 (0.8) 403.124 <0.001 30s 303 (18.8) 1061 (65.9) 228 (14.2) 18 (1.1) 40s 340 (17.9) 1241 (65.3) 296 (15.6) 24 (1.3) 50s 358 (20.0) 1116 (62.2) 301 (16.8) 18 (1.0) 60s or older 203 (17.8) 711 (62.4) 213 (18.7) 13 (1.1) Marital status Single 795 (36.1) 1218 (55.3) 172 (7.8) 19 (0.9) 396.704 <0.001 Have a spouse (married) 878 (16.8) 3375 (64.7) 908 (17.4) 59 (1.1) Do not have a spouse (married) 183 (31.7) 320 (55.5) 67 (11.6) 7 (1.2) Education level Elementary-school graduate or lower 106 (32.2) 193 (58.7) 28 (8.5) 2 (0.6) 81.892 <0.001 Middle-school graduate 165 (28.9) 340 (59.6) 60 (10.5) 5 (0.9) High-school graduate 947 (24.9) 2321 (61.0) 501 (13.2) 37 (1.0) College graduate or higher 637 (19.3) 2060 (62.5) 558 (16.9) 41 (1.2) Household income 355.82 (207.57) 409.69 (200.69) 486.21 (244.21) 529.97 (287.62) 0.189 <0.001 Perception of financial stability 3.40 (1.83) 4.98 (1.42) 6.21 (1.34) 7.10 (2.38) 0.505 <0.001 SD ¼ standard deviation. Fig. 1 e Correspondence analysis between perceived financial status to cope with severe disease and household income. Table 3 e Relationships among perception of financial stability, household income, the level of anxiety and determination of life (N ¼ 8000). Variables Level of anxiety Determination of life Beta t (p) Beta t (p) Perception of financial stability �0.239 �22.021 (<0.001) 0.249 22.984 (<0.001) Household income �0.017 �1.537 (0.124) 0.006 0.535 (0.593) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9136 To identify the relationships among perceived financial status to cope with severe disease, the level of anxiety and determination of life, models using the perception of financial stability as covariate (model 1) and covariate and control variables (model 2) (Table 4) were used. Results indicate that the level of anxiety had a significant relationship, whereas determination of life did not. Furthermore, the level of anxiety was higher for participants who considered their power to cope as somewhat sufficient than for participants who considered it as somewhat or highly insufficient. https://doi.org/10.1016/j.puhe.2019.01.010 https://doi.org/10.1016/j.puhe.2019.01.010 Table 4 e Relationships among perceived financial status to cope with severe disease, the level of anxiety and determination of life (N ¼ 8000). Variables Level of anxiety Determination of life F (p) F (p) Model 1a 12.059 (<0.001) Somewhat sufficient>somewhat
insufficient and highly insufficient

2.261 (0.079)

Model 2b 12.462 (<0.001) Somewhat sufficient>somewhat
insufficient and highly insufficient

2.112 (0.096)

a Linear model using perception of financial stability as covariate

(covariate analysis).
b Gender, age, marital status and education level as control

variables.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9 137
Discussion

Perceived financial status to cope with severe disease does not

have a significant relationship with determination of life but

has a significant relationship with anxiety. Because severe

disease increases health expenditure and worsens household

finances, more substantial policies are required to reduce

anxiety. Furthermore, as perception of power to cope is more

closely related to the subjective perception of financial sta-

bility than to actual income, interventions are required that

focus on subjective perception of economic feasibility, rather

than on actual income, to promote quality of life.

The level of anxiety was lower when the income and

perception of financial stability were higher; this may support

the result that financial strain was associated with having an

anxiety disorder.16 Anxiety may increase because lower in-

come leads to less opportunity to meet needs.17 Previous

studies on the relationship between income and anxiety have

indicated that income may lower the level of sadness;18

although higher income does not affect the positive emotion

of happiness, it may reduce the negative emotion of anxiety or

worry.19 Low-income earners perceive a lack of efficacy to

control their environment more than high-income earners;20

furthermore, the absence of efficacy to change one’s situa-
tion may cause negative emotions such as helplessness, sor-

row and anxiety.21 Income, or perception thereof, therefore

has a correlation with negative than with positive emotions.

Negative emotions can further decrease satisfaction with life

compared with positive emotions.22 It is thus necessary to

focus on the results that indicate a significant relationship

between financial perception and negative emotions.

Determination of life was related more to perception of

financial stability than to actual income, which is aligned with

the result that perceived financial status had a positive effect

on evaluation of life, whereas income did not.19 Previous

studies found that psychological or cognitive factors play a

crucial role in the relationship between income and well-

being.23,24 The theory of desire states that happiness depends

on the gap between desired and actual income, and not on the

actual income itself.24 Therefore, if people want more income

but do not consider themselves to be economically stable,
their determination of life may still be low, even if their in-

come increases.

Study participants perceived their financial status to cope

with severe disease as somewhat or highly insufficient,

depending on their current earnings. Only 8.6% of paid Korean

workers have a monthly income of 6.5 million KRW or higher;

the monthly average income is 3.29 million KRW.25 Accord-

ingly, many people (except those with a monthly income of 6.5

million KRW or higher) feel a sense of crisis in life when faced

with severe disease. Thus, it is difficult to reduce social

stratification and conflicts due to severe diseases and health

expenditure without increasing health insurance coverage for

severe diseases, even though there is national health

insurance.

After controlling for the perception of financial stability,

the perceived financial status to cope had a significant rela-

tionship with anxiety, but not with the determination of life.

People who perceive themselves as economically stable can

afford medical expenses, even when a significant expenditure

such as a severe disease arises. However, people who perceive

themselves as not economically stable despite high income

refrain from spending money for necessary treatment or

sacrifice other parts of life. According to Li et al.,19 higher in-

come did not affect positive well-being but reduced negative

well-being. Furthermore, perceived financial status plays a

mediating role in the relationship between actual income and

subjective well-being,19 thereby supporting the study’s results
of a significant relationship between perceived financial sta-

tus to cope with severe disease and anxiety. The perceived

financial status to cope appears to have a significant rela-

tionship with negative emotions, due to the reality that

suffering from a severe disease leads to loss of the ability to

work and rapidly deteriorates the household economy.

Therefore, it is necessary to increase health insurance

coverage by considering not only income brackets but also

subjective stratum consciousness and to develop systems for

health and medical services, such as medical security policies

and delivery systems.

Health was identified as the biggest concern of middle- and

prime-aged Korean people regarding their later years.26 This

indicates that anxiety about health comprises a particularly

significant part of anxiety about life, supporting this study’s
results that the perceived financial status to cope with severe

disease has a relationship with anxiety. Korea has imple-

mented the nationwide National Health Insurance System

and is constantly reducing the health expenditure burden of

the four major severe diseases through the copayment

decreasing policy. However, people with severe disease face a

rapid income decline and are forced to stop working within a

short period of time.5 Moreover, a Korean study found that

82.3% of people applying for commercial health insurance

prepare for severe diseases.27 This indicates a high risk of

income loss due to severe diseases, which increases people’s
anxiety. Higher income indicates a higher ratio of application

for commercial health insurance,27 and preparation for dis-

eases in old age is also concentrated on those with a ‘high’

socio-economic status.27 Thus, a higher risk of income loss

due to severe diseases among the low-income group exists in

reality. Because financial response to health expenditure

varies according to the economic level and perception of

https://doi.org/10.1016/j.puhe.2019.01.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9138
households, it is necessary to define groups that require

preferential protection from health expenditure, such as the

low-income group, households that lack assets or social cap-

ital and people with severe or chronic diseases, and to

implement policies to reinforce their medical safety net.

The study results should be interpreted with caution due to

some methodological limitations. First, the survey’s cross-
sectional design implies that causality between anxiety or

determination of life and other variables could not be identi-

fied. Second, although the inclusion of many nationwide

samples renders this secondary analysis of data collected by

the Korea Health Panel generalisable, the content and scope of

data analysis were limited and failed to include different

variables that may affect anxiety or determination of life.

Conclusion and recommendation

The present study found that perceived financial status to

cope with serious illness has a relationship with anxiety;

however, it did not have a relationship with decision-making

power in life, which in turn had a stronger correlation with

perceived financial stability than it did with real income.

When establishing a health and medical treatment index se-

lection or payment standards for catastrophic health expen-

ditures, decision-making is primarily done based on real

income brackets; however, this study suggests that the anxi-

ety level experienced by participants has a stronger founda-

tion in financial stability or the perceived financial status to

cope than in real income. Accordingly, when developing

health and medical treatment policy and intervention pro-

grammes related to serious illnesses among the subjects,

one’s perceived financial status or stability should be consid-
ered concurrently.
Author statements

Ethical approval

This article is based on secondary analysis of extant data. The

study was exempt from ethics review and was confirmed by

the university’s institutional review board (No. 2018-02-009).

Funding

None declared.

Competing interests

None declared.
r e f e r e n c e s

1. Lee WY, Shin YJ. Catastrophic health expenditures among
income groups in urban households. Korean Soc Sec Stud
2005;21:105e33.

2. Song EC, Shin YJ. The comprehensive health expenditure
ceiling system to prevent catastrophic health expenditure:
focusing on applicability using cost estimation. Health and Soc
Welfare Rev 2015;35:429e56.

3. Kim DH, Kang SH. Effects of critical illnesses on losing income
and their policy implications. J Korean Acad Ins
2015;102:39e57.

4. National Health Insurance Service & Health Insurance Review
and Assessment Service of Korea. National health insurance
statistical yearbook. 2016. September 2017, http://www.nhis.or.
kr/menu/retriveMenuSet.xx?menuId¼F3321. [Accessed 15
April 2018].

5. Lee SH. Implications for medical utilization and medical
expenditure of severely ill patients: focusing on the socio-
economic status. Health Welfare Iss Focus 2014;235:1e8.

6. Lee HJ, Lee TJ. Factors associated with incidence and
recurrence of household catastrophic health expenditure in
South Korea. Korean Soc Sec Stud 2012;28:39e62.

7. Song EC, Shin YJ. The effect of catastrophic health
expenditure on the transition to and persistence of poverty in
South Korea: analysis of the Korea Welfare Panel Study Data,
2007e2012. Health Policy Manag 2014;24:242e53.

8. Woo KS, Shin YJ. The effect of catastrophic health
expenditure on household economy: focusing on financial
coping and poverty. Health Soc Welfare Rev 2014;35:166e98.

9. Joo S. Personal financial wellness. In: Xiao JJ, editor. Handbook
of consumer research. New York: Springer; 2008.

10. Crompton R. Class and stratification: widening inequalities and
debates on ‘class’. Cambridge, UK: Polity Press; 2008.

11. Jackman MR, Jackman RW. Class awareness in the United States.
Berkeley, CA: University of California Press; 1983.

12. Kim J, Garman ET. Financial stress and absenteeism: an
empirically derived research model. Financ Counsel Plan
2003;14:31e42.

13. Han HJ, Oh EJ, Joung SH. An analysis on variables related to
the financial satisfaction of one-person households. Fin Plan
Rev 2014;7:173e98.

14. Kim Y, Jung K, Park J. Study on state dependence of
catastrophic health expenditure occurrence. Soc Sec Res
2016;32:1e37.

15. Lim SJ, Kim SH, Baek JH, Kim NY. Improvement for enhancing
coverage of health insurance in low income class. Seoul, South
Korea: Health Insurance Policy Research Institute, National
Health Insurance Service; 2013.

16. Dijkstra-Kersten SM, Biesheuvel-Leliefeld KE, van der
Wouden JC, et al. Associations of financial strain and income
with depressive and anxiety disorders. J Epidemiol Community
Health 2015;69:660e5.

17. Moser RP, McCaul K, Peters E, Nelson W, Marcus SE.
Associations of perceived risk and worry with cancer health-
protective actions: data from the Health Information National
Trends Survey (HINTS). J Health Psychol 2007;12:53e65.

18. Kushlev K, Dunn EW, Lucas RE. Higher income is associated
with less daily sadness but not more daily happiness. Soc
Psychol Personal Sci 2015;6:1e7.

19. Li B, Li A, Wang X, Hou Y. The money buffer effect in China: a
higher income cannot make you much happier but might
allow you to worry less. Front Psychol 2016;7:234.

20. Johnson W, Krueger RF. How money buys happiness: genetic
and environmental processes linking finances and life
satisfaction. J Pers Soc Psychol 2006;90:680e91.

21. Roseman IJ, Antonious AA, Jose PE. Appraisal determinants of
emotions: constructing a more accurate and comprehensive
theory. Cognit Emot 1996;10:241e78.

22. Ryff CD, Keyes CLM. The structure of psychological well-being
revisited. J Pers Soc Psychol 1995;69:719e27.

23. Brown KW, Kasser T, Ryan RM, Linley AP, Orzech K. When
what one has is enough: mindfulness, financial desire
discrepancy, and subjective well-being. J Res Pers
2009;43:727e36.

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref1

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref2

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref3

http://www.nhis.or.kr/menu/retriveMenuSet.xx?menuId=F3321

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref5

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref6

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref7

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref8

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref9

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref10

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref11

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref12

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref13

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref14

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref15

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref16

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref17

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref18

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref19

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref20

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref21

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref22

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref23

https://doi.org/10.1016/j.puhe.2019.01.010

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 3 3 e1 3 9 139
24. Solberg EC, Diener E, Wirtz D, Lucas RE, Oishi S. Wanting,
having, and satisfaction: examining the role of desire
discrepancies in satisfaction with income. J Pers Soc Psychol
2002;83:725e34.

25. Statistics Korea. Income (wage) distribution analysis by paid
job. 2017. http://kostat.go.kr/portal/korea/kor_nw/2/3/1/
index.board?bmode¼read&aSeq¼361207. [Accessed 15 April
2018].
26. Lee SJ. The characteristics of old age preparation of Korean
adults and policy implications. Health Welfare Pol Forum
2009;147:1e9.

27. Yoon TH, Hwang IK, Sohn HS, Koh KW, Jeong BK. The
determinants of private health insurance purchasing
decisions under national health insurance system in Korea:
the expanding of private health insurance market, for the
better or worse. Korean J Health Pol Admin 2005;15:161e75.

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref24

http://kostat.go.kr/portal/korea/kor_nw/2/3/1/index.board?bmode=read&aSeq=361207

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref26

http://refhub.elsevier.com/S0033-3506(19)30010-1/sref27

https://doi.org/10.1016/j.puhe.2019.01.010

Do anxiety or determination of life differ based on the perceived financial status to cope with severe diseases?
Introduction
Methods
Sampling and data collection
Measurements
Statistical analysis

Results
Differences between variables based on general characteristics
Household income and perceptions of financial stability and financial status
Perceived financial status and the level of anxiety and determination of life

Discussion
Conclusion and recommendation

Author statements
Ethical approval
Funding
Competing interests
References

Expanding-the-injury-definition–evidence-for-the-need-to-incl_2019_Public-H

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 5
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Expanding the injury definition: evidence for the
need to include musculoskeletal conditions
A. Schuh-Renner*, M. Canham-Chervak, T.L. Grier, V.D. Hauschild,
B.H. Jones

U.S. Army Public Health Center, Injury Prevention Division, E-1570 8977 Sibert Road, Aberdeen Proving Ground, MD

21010, USA
a r t i c l e i n f o

Article history:

Received 16 July 2018

Received in revised form

12 December 2018

Accepted 3 January 2019

Available online 26 February 2019

Keywords:

Epidemiology

Surveillance

Military

Sports

Occupational injury

Surveys
* Corresponding author.
E-mail address: usarmy.apg.medcom-phc

https://doi.org/10.1016/j.puhe.2019.01.002
0033-3506/Published by Elsevier Ltd on beha
a b s t r a c t

Objectives: The objectives of the study are to quantify the proportion of cumulative

microtraumatic overuse injuries in a physically active population, evaluate their impact in

terms of lost work time, and link them to precipitating activities to inform prevention

initiatives.

Study design: The study design is retrospective cohort study.

Methods: For a population of U.S. Army Soldiers, diagnoses from medical records (Inter-

national Classification of Diseases [ICD]-9 800e999 and selected ICD-9 710e739) were

matched with self-reported injury information. Common diagnoses, limited duty days, and

activities and mechanisms associated with the injuries were summarized.

Results: Most self-reported injuries (65%) were classified by providers with diagnoses that

described cumulative microtraumatic tissue damage, and these injuries led to a higher

incidence of limited duty (85%) than acute traumatic injury diagnoses. Reported mecha-

nisms and activities often indicated repetitive physical training-related onset.

Conclusions: Because many diagnoses for cumulative microtraumatic musculoskeletal

tissue damage are categorized as diseases to the musculoskeletal system in the In-

ternational Classification of Diseases, they are often not included in definitions of

injury. However, reported injury activities and mechanisms in this population provide

evidence that cumulative microtraumatic injuries often arise from identifiable and

preventable events. This finding confirms that these diagnoses should be classified as

injuries in epidemiologic evaluations and surveillance to accurately represent injury

burden.

Published by Elsevier Ltd on behalf of The Royal Society for Public Health.
.mbx.injuryprevention@mail.mil (A. Schuh-Renner).

lf of The Royal Society for Public Health.

mailto:usarmy.apg.medcom-phc.mbx.injuryprevention@mail.mil

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.002&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.002

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 570
Introduction

An injury is defined as tissue damage due to a transfer of

external energy to the body.1e5 It is widely accepted in military

and athletic communities that overuse injuries such as stress

fractures, tendinitis, and joint pain result from cumulative

microtrauma from regular participation in exercise and phys-

ical training6e11 and should be included in injury surveillance

and research.4,7,12,13 Unfortunately, the International Classifi-

cation of Diseases (ICD) considers diagnoses for cumulative

microtraumatic tissue damage as Diseases of the Musculo-

skeletal System and Connective Tissue (Chapter 13, ICD-9 710-

739 series codes and ICD-10 M codes). As a result, this catego-

rization groups overuse and pain-related conditions together

with other disorders and diseases that do not have a specific

external cause or mechanism (e.g. infectious arthropathies,

rheumatoid arthritis, systemic connective tissue conditions,

and age-related degenerative conditions). Therefore, there is

often a failure to recognize these conditions as injuries.

Prior military studies have indicated that 30%e70% of mili-

tary injuries are musculoskeletal overuse injuries (selected

710e739 codes in the ICD, 9th Revision, Clinical Modification

(ICD-9-CM) code series; selected M codes in the ICD, 10th Revi-

sion, Clinical Modification (ICD-10-CM) code series).7,14 But, in

accordance with the ICD classifications, many epidemiologic

studies and national medical surveillance systems only include

diagnoses for acute traumatic injuries that appear in the Injury

and Poisoning chapter of the ICD (Chapter 17, ICD-9-CM

800e999;15 Chapter 19, ICD-10-CM S and T codes).16,17 Because

this ICD classification fails to recognize numerous musculo-

skeletal injury diagnoses resulting from cumulative micro-

traumatic mechanisms as injuries, the burden of injuries on our

medical systems, communities, and workplaces is

underestimated.

This is especially problematic for physically active pop-

ulations, such as athletes and military members. Cumulative

microtraumatic injuries (often referred to as ‘overuse in-

juries’) have been widely accepted in injury definitions among

the sports medicine community.11,18 The exclusion of these

diagnoses in injury surveillance and epidemiological studies

leads to missed injury prevention opportunities in physically

active populations. Assignment of external cause codes was

not recommended for diagnoses of diseases to the musculo-

skeletal system (i.e. ICD-9-CM 710e739 codes19), so associated

activities and mechanisms for many cumulative micro-

traumatic injuries are not reliably captured.

To address this gap, the purpose of this investigation was

to quantify the proportion of self-reported injuries that

received clinical diagnoses from the Diseases of the Muscu-

loskeletal System and Connective Tissue ICD chapter, but

which meet the definition of injury. External causes and

limited duty days resulting from these cumulative micro-

traumatic injuries are evaluated.
Methods

Soldiers in two U.S. Army Infantry Brigade Combat Teams

completed surveys as part of initial baseline data collection for
injury prevention program evaluations. These projects were

reviewed and approved by the Army Public Health Center

Public Health Review Board as public health practice. Surveys

were administered in 2010 and 2011 in coordination with unit

leadership. Injury risks for these populations have been pre-

viously reported.20e22

The survey collected detailed information about re-

spondents’ most recent injury. Injury was defined as any acci-
dental or intentional force applied to the body. For their most

recent injury, collected details included the following: injury

date, injured body part, injury type, activity during which the

injury occurred, mechanism of injury, and how many days of

limited days resulted (if applicable). Survey respondents were

asked to describe their injured body region and injury type.23

The following options were provided for respondents to

report the mechanism of their most recent injury: fall, jump,

trip, or slip; struck against or struck by an object or person; cut

by a sharp instrument, tool, or object; overexertion, strenuous

or repetitive movements; fire, hot substance or object, or

steam; environmental factors such as heat or cold; breathing or

swallowing dust, particles, liquid vapors, or fumes; and other

(further specification requested). Likewise, respondents could

choose from the following options when asked about the ac-

tivity associated with their injury: riding or driving in a

motorized vehicle; exercising (further specification requested);

sports (further specification requested); walking, hiking, or

road marching; stepping or climbing; lifting or moving heavy

objects; repairing or maintaining equipment or vehicles; and

other (further specification requested).

For the same surveyed brigades, electronic medical record

data for all soldiers on the rosters were obtained from the

Defense Medical Surveillance System with visit dates

extending back six months from the survey date. Injury di-

agnoses were identified using an index of ICD-9-CM medical

diagnoses codes25 that included the ICD-9-CM Injury and

Poisonings code group (800e999), and diagnoses for selected

overuse and pain-related musculoskeletal conditions in the

Musculoskeletal Conditions chapter (predominantly selected

710e739 codes, based on a previously defined Overuse Injury

Index25 and other definitions of common injury-related

musculoskeletal conditions in the military7,26). Codes for

poisonings, toxins, and complications related to medical

procedures were not included, consistent with recommen-

dations for military injury surveillance and prior military field

investigations.25e27

Self-reported injuries were previously matched with

medical records based on a report date within three months

and an identical (primary) or proximate (secondary) matching

body part.23 Descriptive statistics were calculated for the

leading ICD-9-CM diagnoses associated with the matched in-

juries using the Statistical Package for the Social Sciences

(SPSS®), version 19.0. For the five self-reported injuries most

frequently matched with a medical diagnosis, the following

information was reported: leading diagnoses from the medi-

cal record, number of injuries resulting in lost duty time, the

average number of self-reported limited duty days resulting

from the diagnosis, leading mechanisms of injury, and leading

activities associated with injury. OpenEpi (www.openepi.com)

was used to calculate the risk ratio of experiencing limited

duty from an injury coded as a musculoskeletal condition

http://www.openepi.com

https://doi.org/10.1016/j.puhe.2019.01.002

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 5 71
compared with experiencing limited duty from an injury

coded as an acute traumatic injury. This site was also used to

conduct an independent sample t-test to compare the average

number of limited duty days experienced for each group of

diagnosis codes.
Results

Over 5000 soldiers from two Army Infantry Brigades (n ¼ 5492)
were surveyed. Most respondents were enlisted ranks (91%),

men (92%), and the mean age was 31 ± 6 years. Many of the
survey respondents (n ¼ 2332, 42%) had at least one medical
encounter with an acute traumatic injury diagnosis code (ICD-

9-CM 800e999) or cumulative microtraumatic overuse and

pain-related musculoskeletal injury diagnoses (selected ICD-

9-CM 710e739). Among the soldiers with injury medical re-

cords, a majority (n ¼ 3185, 62%) had diagnoses for overuse
injuries (selected ICD-9-CM 710e739).

A total of 1336 soldiers both reported an injury on the

survey and had a medical record diagnosis for injury within 6

months of survey administration. Among them, 996 (75%) of

the self-reported injuries matched a medical record by date

(±3 months) and had an identical or proximate injured body
region.23 As shown in Table 1, the top five self-reported in-

juries that matched a medical diagnosis were ankle sprain/

strains (10%), knee sprain/strains (9%), lower back sprain/

strain (4%), shoulder sprain/strains (3%), and lower back pain

(3%). Self-reported limited duty days were associated with 83%

of these injuries (n ¼ 823 of 996), for an average of 58 reported
days of limited duty per injury.

Diagnoses for injury-related musculoskeletal conditions

(selected ICD-9-CM 710e739 codes) accounted for 65% of the

matched injuries (n ¼ 646 of 996). Of these, 85% (n ¼ 546 of 646)
were associated with self-reported limited duty for an average

of 68 days per injury. In comparison, 80% of acute traumatic

matched injuries (ICD-9-CM 800e899 codes) resulted in limited

duty (n ¼ 274 of 345), with an average of 39 days per injury.
Therefore, the relative risk of experiencing limited duty for an

overuse injury was significantly higher (1.06, P ¼ 0.04) than the
risk of experiencing limited duty for an injury that was diag-

nosed as an acute traumatic injury. Analysis with a two-sample

independent t-test indicated that the number of limited duty

days associated with overuse injuries was also significantly

higher than those associated with acute traumatic injuries

(P < 0.001). The existence of an associated causal mechanism and/or activity identifies incidents as preventable injuries, rather than underlying disease conditions, especially for those injury encounters diagnosed as musculoskeletal conditions (selected ICD-9-CM 710e739; selected ICD-10-CM M codes). Nearly all respondents with a self-reported injury and matching diag- nosis in this population reported an associated activity (96%) and mechanism (94%). The top activities associated with the matched self-reported injuries were running (29%), other ex- ercise (13%), and lifting or moving heavy objects (11%). Ac- tivities commonly reported as ‘other exercise’ were weight training, martial arts, and unspecified military physical training. The mechanisms most frequently reported for matched injuries were overexertion (39%), falls/trips/slips (29%), and struck by or against an object or person (12%). Discussion The contribution of cumulative microtraumatic tissue damage diagnoses to athletic injuries The matched diagnoses from electronic medical records were predominantly (65%) selected ICD-9-CM 710e739 codes which reflect overuse and overexertion-related injuries. These in- juries have a significant impact on military readiness28 and have been consistently identified as common injuries in physically active populations.7,11,29e31 For example, one pre- vious investigation found that the most common outpatient injury diagnoses among military members were lower ex- tremity overuse injuries, accounting for 43% of injuries and 15% of limited duty days.30 Among soldiers self-reporting an injury in this study, the most common injuries matching a medical record were ankle sprains, knee sprains, lower back strains, shoulder sprains, and lower back pain. These musculoskeletal injuries are consistent with the types of injuries to the lower extremities and lower back that are most commonly associated with physical activities.6,7,9e11,18,32e35 As shown by this study, many of these injuries are diagnosed by medical providers with ICD codes that are not formally classified as injuries (selected ICD-9-CM 710e739; selected ICD-10-CM M codes). Sports and physical training have often been cited as ac- tivities associated with military injuries,6e8,30,31 even during deployments.36 Knowledge of the specific activities and mechanisms associated with injuries are actionable details that can inform injury prevention planning,18 but this infor- mation cannot reliably be obtained directly from medical re- cords because cause-coding of outpatient military injuries is not required and is, therefore, only captured in one-tenth of outpatient records.37 Acknowledging this limitation, a past analysis of Army soldier medical records showed that 72% of injury diagnoses with an external cause code for overexertion (ICD-9-CM E927) were sprains or strains.32 Most of those sprains or strains were to the lower extremities (59%), and 32% also had an activity code associated with running (ICD-9-CM E001.1). Because clinical documentation of the mechanisms and activities leading to injuries is lacking,32 especially for cumu- lative microtraumatic injuries, the use of self-reported survey data is integral to identifying injury causes. In one survey of over 10,000 military service members, 52% of all self-reported injuries were associated with sports or exercise.6 In fact, running is often the most frequently cited activity leading to injury in military populations.30 Identifying associated activities and mechanisms of injury is an integral aspect of the public health process to injury prevention.7 This study provides further verification and evi- dence of the relationship between common military injuries and their precipitating activities and mechanisms. For example, it was observed from these survey responses that many injuries to the ankle and knee were commonly associ- ated with falls/slips/trips during running, whereas back and https://doi.org/10.1016/j.puhe.2019.01.002 https://doi.org/10.1016/j.puhe.2019.01.002 Table 1 e Leading self-reported injuries (n ¼ 996 soldiers with injuries matched to medical records). Top five injury type/ body part responses on surveya Soldiers with matched diagnosis n (%) Top three medical diagnoses from ICD records Soldiers with resulting limited duty daysa n (%) Average number of limited duty days per soldiera Top injury activitiesa Top injury mechanismsa Diagnoses n (%) Activity n (%) Mechanism n (%) Ankle sprain/strain 98 (10) 845.00 Sprain of ankle 42 (43) 83 (85) 41b Running 47 (48) Fall/trip/slip 73 (74) 719.47 Joint pain, ankle 28 (29) Walking/hiking 12 (12) Overexertion 15 (15) 719.46 Joint pain, lower leg 8 (8) Sports 11 (11) Struck by 2 (2) Knee sprain/strain 89 (9) 719.46 Joint pain, lower leg 55 (62) 77 (87) 62b Running 33 (37) Fall/trip/slip 40 (45) 844.90 Sprain or strain, knee 8 (9) Sports 13 (22) Overexertion 34 (38) 726.64 Patellar tendinitis 6 (7) Other exercise 13 (22) Struck by 7 (8) Lower back sprain/strain 44 (4) 724.20 Lumbago 24 (59) 33 (75) 46b Lifting 16 (36) Overexertion 29 (66) 847.20 Sprain lumbar region 4 (10) Running 6 (14) Fall/trip/slip 4 (8) 724.50 Backache 3 (7) Road marching 5 (11) Struck by 3 (5) Shoulder sprain/strain 34 (3) 719.41 Joint pain, shoulder 15 (41) 29 (85) 54b Other exercise 14 (41) Overexertion 18 (53) 840.40 Rotator cuff sprain 4 (12) Lifting 8 (23) Fall/trip/jump 8 (24) 840.80 Sprain of shoulder 4 (12) Running 2 (6) Struck by 4 (12) Stepping/climbing 2 (6) Motor vehicle 2 (6) Lower back pain 32 (3) 724.20 Lumbago 23 (72) 28 (88) 63b Lifting 8 (25) Overexertion 15 (47) 724.10 Pain in thoracic spine 2 (6) Running 7 (21) Fall/trip/slip 5 (16) 724.50 Backache 2 (6) Road marching 3 (9) Struck by 3 (9) a Self-reported survey response for most recent reported injury. b 78 of 83 who said they had an ankle sprain/strain injury with limited duty reported the number of days; 69 of 77 (knee sprain/strains); 23 of 33 (lower back sprain/strains); 27 of 29 (shoulder sprain/ strains); 24 of 28 (lower back pain). p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e 7 5 7 2 https://doi.org/10.1016/j.puhe.2019.01.002 https://doi.org/10.1016/j.puhe.2019.01.002 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 5 73 shoulder injuries were likely to be attributed to overexertion while exercising or lifting heavy objects. Even though medical providers used codes from the Diseases of the Musculoskeletal System and Connective Tissues ICD chapter (selected ICD-9- CM 710e739; selected ICD-10-CM M codes) to diagnose 65% of these injuries, they can all be attributed to a transfer of energy because specific activities and mechanisms were identified. Findings from this investigation provide evidence that reporting only acute traumatic injuries identified by the ICD Injury and Poisonings chapter (ICD-9-CM 800e999; ICD-10-CM S and T codes) may fail to capture up to two-thirds of all in- juries, especially in physically active populations. Further- more, the impact of these missed injuries may often be greater than that of the acute traumatic injuries. Specifically, musculoskeletal overuse and overexertion injuries in the current population led to a higher incidence of reported limited duty days compared to acute traumatic injuries (85% compared to 80%, P ¼ 0.03), and the limited duty lasted longer on average (68 days vs 39 days, P < 0.001). Information relevant to the prevention of musculoskeletal conditions While aerobic and muscular endurance are protective against injuries during sports and physical training,8,38,39 the cumu- lative effects of repetitive microtrauma experienced during physical activities have been shown to lead to overuse injuries.9e11,20,40,41 Program-induced cumulative overload, or overtraining that increases injury risk, has been identified as a concern for Army soldiers.42 Other studies have also shown that an increased volume of physical training among soldiers can lead to increased injury risk,28 especially higher running mileage,20,41 or more intense road marching training.43 Stra- tegies that have previously been suggested to minimize the risk of injury from overtraining include balanced physical training which incorporates a variety of activities and exercise types,44,45 reducing weight-bearing activities,46 reducing running mileage,20,41 gradually increasing road marching training intensity,43,47 and leading strategic initiatives to reduce injuries in smaller, targeted populations to address differences in mission and physical training needs.28 Although it has previously been suggested that injuries resulting from cumulative microtraumatic tissue damage should be included in a broader injury surveillance definition to accurately capture the burden of injuries,2,4,7,12,13 they are not currently included in the Injury and Poisonings chapter of the ICD and are often excluded from medical injury surveillance and epidemiological studies. The sports medicine and military health communities, however, do acknowledge musculoskel- etal overuse injuries as a substantial injury problem in their physically active populations,4,11,18,26 and a matrix has been created to categorize, monitor, and focus prevention strategies for these cumulative microtraumatic injuries.7 A consistent injury definition is needed to understand the distribution of injuries among populations of interest, which can be used to inform injury prevention planning.4,29,48e50 This definition should include both acute traumatic injuries and cumulative microtraumatic injuries, regardless of their ICD classification code. To support this need, a recent effort has systematically defined injury, identified all ICD diagnoses meeting the definition, and categorized them.4 Categories specify causal energy mechanisms (e.g. mechanical, thermal, chemical, or electrical energy), injury types, and injured body regions. This comprehensive methodology allows for all in- juries, including overuse injuries, to be captured in the injury distribution for a population, which can provide data-driven motivation to focus prevention initiatives on certain sub- categories of injuries (i.e. acute mechanical musculoskeletal injuries, heat injuries, or lower extremity injuries). This taxo- nomic structure allows specific types of injuries to be easily monitored separately, as desired. The results of the present study support the need for organizations and surveillance sys- tems to embrace this broader, all-inclusive definition of injury. Limitations Although the proposed inclusion of overuse injuries best represents a population’s burden of injury, the application of a broader injury definition may result in larger injury datasets which could require additional resources to process and analyze. As with any analytical methodology transition, ob- servations using the revised definition may not be comparable to past analyses. This study used medical data that predated the ICD-10-CM coding system that is now required for U.S. medical practice.17 Most of the ICD-9-CM 710e739 codes for injury-related musculoskeletal conditions map to ICD-10-CM M-series codes, and ICD-9CM 800e999 codes for acute injuries map to ICD-10-CM S-series and T-series codes.16 As many more codes exist in ICD-10-CM than were in ICD-9-CM,17,51 additional specificity may be gained by considering ICD-10-CM coding in future studies. The ICD-10-CM update also classifies cumula- tive microtraumatic injuries with other non-injuryerelated musculoskeletal conditions, separate from acute injuries. In addition, the data acquired for this study represent predominantly male soldiers, as only 8% of this survey pop- ulation was women. While this is an approximately current representation of women in the Army,52 musculoskeletal conditions are typically more prevalent among women9,11 and female soldiers may require different prevention strategies than men.53 Conclusions Injury is often defined as the transfer of external energy to the body, and injuries resulting from cumulative microtrauma are common in physically active populations. However, because the ICD classifies overuse and cumulative microtraumatic injuries as musculoskeletal diseases and disorders, these di- agnoses are grouped inappropriately with other disease- related chronic conditions and often go unrecognized as in- juries, even though they have an external cause. Although injury-related musculoskeletal conditions are sometimes included in definitions of injury in certain populations of athletes and military members, this practice should be extended to injury definitions more broadly. This investigation was the first to demonstrate that a majority of self-reported injuries are diagnosed as musculo- skeletal overuse injuries in a large, physically active popula- tion. Most reported injuries (65%) were associated with https://doi.org/10.1016/j.puhe.2019.01.002 https://doi.org/10.1016/j.puhe.2019.01.002 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 574 musculoskeletal overuse injury diagnoses, and these di- agnoses were significantly more likely to be associated with self-reported limited duty for significantly more days. Asso- ciated activities and mechanisms were reported, suggesting that strategic injury mitigation strategies could have pre- vented their occurrence. These findings support the inclusion of musculoskeletal overuse injuries resulting from cumulative microtraumatic tissue damage during physical training in injury definitions. Injury prevention initiatives used to reduce overtraining in athletes should be adapted for other pop- ulations with high rates of cumulative microtraumatic in- juries. The results of this study can be used to specifically emphasize the need for continued investigation into and development of interventions for the prevention of overuse injuries that occur during running, sports, and lifting activities. Author statements Acknowledgments The views expressed in this document are those of the authors and do not necessarily reflect the official policy of the Department of Defense, Department of Army, US Army Med- ical Department, or the US Government. Ethical approval This work was approved by the authors' institution Public Health Review Board as public health practice. Funding None declared. Competing interests None declared. r e f e r e n c e s 1. Haas E. Handbook of injury and violence prevention. Springer Science & Business Media; 2007. 2. Langley J, Brenner R. What is an injury? Inj Prev 2004;10:69e71. 3. Rivara FP, Cummings P, Koepsell TD, Grossman DC, Maier RV. Injury control: a guide to research and program evaluation. Cambridge University Press; 2009. 4. U.S. Army Public Health Center. A taxonomy of injuries for public health monitoring and reporting. Public Health Report No. 12-01-0717. Prepared by Hauschild V, Hauret K, Richardson M, Jones BH, and Lee T. Available at: http://www. dtic.mil/docs/citations/AD10394812017. 5. The National Committee for Injury Prevention and Control (US). Injury prevention: meeting the challenge. New York: Oxford University Press; 1989. 6. Hauret KG, Bedno S, Loringer K, Kao T-C, Mallon T, Jones BH. Epidemiology of exercise-and sports-related injuries in a population of young, physically active adults: a survey of military service members. Am J Sports Med 2015;43:2645e53. 7. Hauret KG, Jones BH, Bullock SH, Canham-Chervak M, Canada S. Musculoskeletal injuries: description of an under- recognized injury problem among military personnel. Am J Prev Med 2010;38:S61e70. 8. Jones BH, Hauschild VD. Physical training, fitness, and injuries: lessons learned from military studies. J Strength Condit Res 2015;29:S57e64. 9. Yang J, Tibbetts AS, Covassin T, Cheng G, Nayar S, Heiden E. Epidemiology of overuse and acute injuries among competitive collegiate athletes. J Athl Train 2012;47:198e204. 10. Junge A, Engebretsen L, Mountjoy ML, Alonso JM, Renstr€om PA, Aubry MJ, et al. Sports injuries during the summer Olympic games 2008. Am J Sports Med 2009;37:2165e72. 11. Roos KG, Marshall SW, Kerr ZY, Golightly YM, Kucera KL, Myers JB, et al. Epidemiology of overuse injuries in collegiate and high school athletics in the United States. Am J Sports Med 2015;43:1790e7. 12. Timpka T, Alonso J-M, Jacobsson J, Junge A, Branco P, Clarsen B, et al. Injury and illness definitions and data collection procedures for use in epidemiological studies in Athletics (track and field): consensus statement. Br J Sports Med 2014;48:483e90. 13. Noyes FR, Lindenfeld TN, Marshall MT. What determines an athletic injury (definition)? Who determines an injury (occurrence)? Am J Sports Med 1988;16. S-65-S-8. 14. Reynolds K, Cosio-Lima L, Bovill M, Tharion W, Williams J, Hodges T. A comparison of injuries, limited-duty days, and injury risk factors in infantry, artillery, construction engineers, and special forces soldiers. Mil Med 2009;174:702. 15. Centers for Disease Control and Prevention. International classification of diseases, clinical modification. 2009. 16. World Health Organization. International statistical classification of diseases and related health problems: 10th revision. World Health Organization; 2004. 17. Hedegaard H, Johnson R, Warner M, Chen L, Annest J. Proposed framework for presenting injury data using the international classification of diseases, tenth revision, clinical modification (ICD-10-CM) diagnosis codes. Natl Health Stat Rep 2016:1e20. 18. Brown MW, Brown RC. Athletic injuries. Trauma 1999;1:271e8. 19. National Center for Health Statistics. ICD-9-CM official guidelines for coding and reporting. 2006. 20. Grier TL, Canham-Chervak M, Anderson MK, Bushman TT, Jones BH. Effects of physical training and fitness on running injuries in physically active young men. J Strength Condit Res 2017;31:207e16. 21. Anderson MK, Grier T, Canham-Chervak M, Bushman TT, Nindl BC, Jones BH. Physical activity effect of mandatory unit and individual physical training on fitness in military men and women. Am J Health Promot 2017;31:371e4. 22. U.S. Army Public Health Command. In: Grier T, Canham- Chervak M, Anderson MK, Bushman TT, Jones BH, editors. Evaluation of the Iron Horse Performance Optimization Physical Training Program (IHPOP) in a Light Infantry Brigade, October 2010eApril 2011; 2014. 23. Schuh-Renner A, Canham-Chervak M, Grier TL, Jones BH. Accuracy of self-reported injuries compared to medical record data. Musculoskel Sci Pract 2019;39:39e44. 25. U.S. Army Center for Health Promotion and Preventive Medicine. Technical report no. 12-HF-5772B-04. Evaluation of two army fitness programs: the TRADOC standardized physical training program for basic combat training and the fitness assessment program. Prepared by Knapik JJ, Darakjy S, Scott S, Hauret KG, Canada S, Marin R, Palkoska F, VanCamp http://refhub.elsevier.com/S0033-3506(19)30002-2/sref1 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref1 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref1 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref2 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref2 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref2 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref3 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref3 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref3 http://www.dtic.mil/docs/citations/AD10394812017 http://www.dtic.mil/docs/citations/AD10394812017 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref5 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref5 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref5 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref6 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref7 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref7 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref7 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref7 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref7 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref8 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref8 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref8 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref8 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref9 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref9 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref9 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref9 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref9 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref10 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref11 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref11 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref11 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref11 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref11 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref12 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref13 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref13 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref13 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref14 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref14 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref14 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref14 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref15 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref15 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref16 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref16 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref16 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref17 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref18 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref18 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref19 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref19 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref20 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref20 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref20 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref20 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref20 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref21 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref21 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref21 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref21 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref21 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref22 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref23 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref23 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref23 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref23 https://doi.org/10.1016/j.puhe.2019.01.002 https://doi.org/10.1016/j.puhe.2019.01.002 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 6 9 e7 5 75 S, Piskator E, Rieger W, and Jones BH. Available at: http:// www.dtic.mil/docs/citations/ADA420942.2004. 26. DoD Military Injury Metrics Working Group. DoD military injury metrics working group white paper. US Department of Defense; 2002. Available from: org/ewgweb/SubPages/ProgramTools/ Metrics/MilitaryInjuryMetricsWhitepaperNov02rev . 27. Defense Health Agency Armed Forces Health Surveillance Branch. Installation injury report documentation. 2004. 28. Nindl BC, Castellani JW, Warr BJ, Sharp MA, Henning PC, Spiering BA, et al. Physiological employment standards iII: physiological challenges and consequences encountered during international military deployments. Eur J Appl Physiol 2013;113:2655e72. 29. Bahr R. No injuries, but plenty of pain? On the methodology for recording overuse symptoms in sports. Br J Sports Med 2009;43:966e72. 30. Ruscio BA, Jones BH, Bullock SH, Burnham BR, Canham- Chervak M, Rennix CP, et al. A process to identify military injury prevention priorities based on injury type and limited duty days. Am J Prev Med 2010;38:S19e33. 31. Smith L, Westrick R, Sauers S, Cooper A, Scofield D, Claro P, et al. Underreporting of musculoskeletal injuries in the US Army: findings from an infantry brigade combat team survey study. Sports Health 2016;8:507e13. 32. Canham-Chervak M, Steelman RA, Schuh A, Jones BH. Importance of external cause coding for injury surveillance: lessons from assessment of overexertion injuries among US Army soldiers in 2014. MSMR 2016;23:10e5. 33. Wilder RP, Sethi S. Overuse injuries: tendinopathies, stress fractures, compartment syndrome, and shin splints. Clin Sports Med 2004;23:55e81. 34. Parkkari J, Kujala UM, Kannus P. Is it possible to prevent sports injuries? Sports Med 2001;31:985e95. 35. Marshall SW, Canham-Chervak M, Dada EO, Jones BH. The burden of musculoskeletal diseases in the United States: military injuries. 2014. Available from: http://www.boneandjointburden. org/2013-report/military-injuries/vi5. 36. Hauret KG, Taylor BJ, Clemmons NS, Block SR, Jones BH. Frequency and causes of nonbattle injuries air evacuated from operations Iraqi freedom and enduring freedom, US Army, 2001e2006. Am J Prev Med 2010;38:S94e107. 37. Gunlicks JB, Patton JT, Miller SF, Atkins MG. Public health and risk management: a hybridized approach to military injury prevention. Am J Prev Med 2010;38:S214e6. 38. Grier T, Canham-Chervak M, McNulty V, Jones BH. Extreme conditioning programs and injury risk in a US Army Brigade Combat Team. US Army Medical Department Journal; 2013. 39. Knapik JJ, East WB. History of United States army physical fitness and physical readiness testing. US Army Med Dept J 2014. 40. Kaufman KR, Brodine S, Shaffer R. Military training-related injuries: surveillance, research, and prevention. Am J Prev Med 2000;18:54e63. 41. Jones BH, Cowan DN, Knapik JJ. Exercise, training and injuries. Sports Med 1994;18:202e14. 42. Orr R, Knapik J, Pope R. Avoiding program-induced cumulative overload (PICO). J Spec Oper Med 2017;16:91e5. 43. Schuh-Renner A, Grier TL, Canham-Chervak M, Hauschild VD, Roy TC, Fletcher J, et al. Risk factors for injury associated with low, moderate, and high mileage road marching in a US Army infantry brigade. J Sci Med Sport 2017;20:S28e33. 44. Bullock SH, Jones BH, Gilchrist J, Marshall SW. Prevention of physical trainingerelated injuries: recommendations for the military and other active populations based on expedited systematic reviews. Am J Prev Med 2010;38:S156e81. 45. Department of Defense. Field manual 7e22 army physical readiness training. Washington, DC: US Government Printing Office; 2012. 46. Jones BH, Thacker SB, Gilchrist J, Kimsey Jr CD, Sosin DM. Prevention of lower extremity stress fractures in athletes and soldiers: a systematic review. Epidemiol Rev 2002;24:228e47. 47. Army Public Health Center (Provisional). In: Hauschild VD, Roy T, Grier T, Schuh A, Jones BH, editors. USAPHC technical information paper (TIP) No. 12-054-0616: foot marching, load carriage, and injury risk; May 2016. Available at: http://www. dtic.mil/get-tr-doc/pdf?AD¼AD1010939 2016. 48. Brooks JH, Fuller CW. The influence of methodological issues on the results and conclusions from epidemiological studies of sports injuries. Sports Med 2006;36:459e72. 49. Van Mechelen W, Hlobil H, Kemper HC. Incidence, severity, aetiology and prevention of sports injuries. Sports Med 1992;14:82e99. 50. Cryer C, Langley JD. Studies need to make explicit the theoretical and case definitions of injury. Inj Prev 2008;14:74e7. 51. Manchikanti L, Hammer MJ, Boswell MV, Kaye AD, Hirsch JA. Survival strategies for tsunami of ICD-10-CM for interventionalists: pursue or perish! Pain Physician 2015;18:E685e712. 52. Defense medical surveillance system. Department of Defense; 2016. Available from: http://afhsc.army.mil/Home/DMSS. 53. Nindl BC, Jones BH, Van Arsdale SJ, Kelly K, Kraemer WJ. Operational physical performance and fitness in military women: physiological, musculoskeletal injury, and optimized physical training considerations for successfully integrating women into combat-centric military occupations. Mil Med 2016;181:50e62. http://www.dtic.mil/docs/citations/ADA420942.2004 http://www.dtic.mil/docs/citations/ADA420942.2004 http://org/ewgweb/SubPages/ProgramTools/Metrics/MilitaryInjuryMetricsWhitepaperNov02rev http://org/ewgweb/SubPages/ProgramTools/Metrics/MilitaryInjuryMetricsWhitepaperNov02rev http://refhub.elsevier.com/S0033-3506(19)30002-2/sref27 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref27 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref28 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref29 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref29 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref29 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref29 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref30 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref30 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref30 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref30 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref30 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref31 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref31 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref31 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref31 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref31 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref32 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref32 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref32 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref32 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref32 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref33 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref33 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref33 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref33 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref34 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref34 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref34 http://www.boneandjointburden.org/2013-report/military-injuries/vi5 http://www.boneandjointburden.org/2013-report/military-injuries/vi5 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref36 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref37 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref37 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref37 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref37 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref38 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref38 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref38 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref39 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref39 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref40 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref40 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref40 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref40 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref41 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref41 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref41 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref42 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref42 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref42 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref43 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref44 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref45 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref45 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref45 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref45 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref46 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref46 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref46 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref46 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref46 http://www.dtic.mil/get-tr-doc/pdf?AD=AD1010939%202016 http://www.dtic.mil/get-tr-doc/pdf?AD=AD1010939%202016 http://www.dtic.mil/get-tr-doc/pdf?AD=AD1010939%202016 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref48 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref48 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref48 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref48 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref49 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref49 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref49 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref49 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref50 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref50 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref50 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref51 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref51 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref51 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref51 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref51 http://afhsc.army.mil/Home/DMSS http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 http://refhub.elsevier.com/S0033-3506(19)30002-2/sref53 https://doi.org/10.1016/j.puhe.2019.01.002 https://doi.org/10.1016/j.puhe.2019.01.002 Expanding the injury definition: evidence for the need to include musculoskeletal conditions Introduction Methods Results Discussion The contribution of cumulative microtraumatic tissue damage diagnoses to athletic injuries Information relevant to the prevention of musculoskeletal conditions Limitations Conclusions Author statements Acknowledgments Ethical approval Funding Competing interests References Fiscal-measures-to-promote-healthier-choices--an-economic-pers_2019_Public-H ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7 Available online at w Public Health journal homepage: www.elsevier.com/puhe Themed Paper e Original Research Fiscal measures to promote healthier choices: an economic perspective on price-based interventions A. Ludbrook* Health Economics Research Unit, Institute of Applied Health Sciences, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, UK a r t i c l e i n f o Article history: Received 15 April 2018 Received in revised form 21 November 2018 Accepted 4 February 2019 Available online 21 March 2019 Keywords: Economics Taxation Subsidies Nudge Tobacco Alcohol Diet Physical activity * Tel.: þ44 (0) 1224 437168. E-mail address: a.ludbrook@abdn.ac.uk. https://doi.org/10.1016/j.puhe.2019.02.008 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Introduction: Non-communicable diseases strongly linked to lifestyle factors create an increasing burden of disease. Fiscal interventions (tax and subsidy) are one approach to improving lifestyles, but their effective design might be improved. Economic framework: Conventional economic theory suggests that fiscal interventions are only used to correct prices for externalities (costs or benefits imposed on others). These can be difficult to calculate accurately. Fiscal interventions operate by altering the prices that consumers face. Price increases are predicted to reduce demand, and the size of the effect is measured by the price elasticity. Tax changes may not translate directly into price changes, however. Evidence for the effect of taxes, subsidies and prices: There is strong evidence for the effec- tiveness of taxation in relation to reducing tobacco and alcohol consumption and resulting harms. There has been less evaluation of taxation in relation to other unhealthy behaviors or of subsidies to promote healthy behaviors. Discussion: Fiscal levers have been used as interventions to improve health rather than for market correction. Taking account of behavioral insights may improve the design of fiscal interventions and combining interventions may increase effectiveness. Conclusion: Both types of intervention have a role in improving health, but there may be challenges in promoting uptake of healthy behaviors. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. Introduction Non-communicable diseases (NCDs) such as cancer, heart disease, stroke, and diabetes are amongst the leading causes of morbidity and mortality in higher income countries and increasing in low- and middle-income countries,1 and their prevalence is strongly linked to lifestyle factors including diet, ic Health. Published by E sedentary behavior, smoking, and alcohol. In the United Kingdom, the mortality rate from some NCDs, particularly heart disease, has been falling, but they remain the leading cause of death, and there is an increasing burden of disease as people live longer in ill health. There are many factors which contribute to unhealthy lifestyles; individuals make choices about what to consume, lsevier Ltd. All rights reserved. mailto:a.ludbrook@abdn.ac.uk http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.008&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7 181 but these choices are shaped and constrained by their life circumstances and environmental factors. There is a general consensus that action is required across a broad range of causes and settings that contribute to unhealthy lifestyles. However, this article focuses principally on an economic perspective of the role of fiscal interventions (taxes and sub- sidies) and other price-based policies in promoting healthier choices. This is not intended to imply that this is the best or the only way to address health behaviors or that individual health behavior is the only factor that should be addressed. In the discussion, it is argued that the formation of policy should consider a broad range of effective interventions, including smarter fiscal interventions and other price-based policies and take account of the relative costs and benefits of indi- vidual interventions as well as the combined efficiency of multiple interventions. The next section outlines the economic framework which underpins the application of taxes, subsidies, or other price- based interventions. This is followed by a focused review of the evidence relating to health behaviors which is intended to provide an insight into the different challenges posed by the varying characteristics of the behaviors and the different na- ture of the economic decisions that are involved. The discus- sion considers how the effectiveness of fiscal interventions and price-based policies might be improved as well as their role in relation to other potential economic interventions. Economic framework Consumer perspective The underlying model for the economics of consumer choice assumes that individuals behave as if they are allocating limited resources (time and money) across a range of goods and services such that the benefit they receive is maximized, based on the cost to them of acquiring the goods and services. Individual preferences are taken as a given, and these pref- erences are complete and stable over time. Consumers are assumed to act rationally and to know what is in their best interest, making decisions accordingly. This is the principle of consumer sovereignty. Clearly, the real world does not conform exactly to this model, and this is explicitly recog- nized in the phrase ‘as if’ above. The strength of this or any model lies in its ability to make reasonably accurate and testable predictions of what will happen when conditions change. Within the given parameters, the economic model can predict how the use of any price-based intervention, including taxes or subsidies, will affect demand for a particular product or service. The direct effect of a price increase (decrease) on any good is to reduce (increase) purchases of that good (and any complementary products) and increase (reduce) pur- chases of goods which are substitutes. One practical difficulty in relation to the goods or activities relating to health behavior is that these are typically classes of goods or activities, and consumers can maintain a higher level of consumption by substituting within a product category; for example, switching to a cheaper brand of alcohol or tobacco. This phenomenon is known as trading down and implies that from a public health policy perspective, the lowest prices in any particular market are important. The economic measure of the effect of a tax or other price change can be expressed through the price elasticity; this is the ratio of the change in quantity to the change in price. If a 10% increase in price reduces the quantity bought by 5%, the price elasticity is �0.5. Price elasticities are almost always negative (price increases reduce the quantity purchased). If the price elasticity is between 0 and �1, then demand is described as inelastic; that is, the quantity purchased falls by a smaller percentage than the price increase. All other goods with a negative price elasticity have demand that is elastic. Price elasticities vary considerably across goods related to health behaviors, across time, and across countries and depending on the estimation methods and data sources used. Estimates will also vary when there are differences or changes in the use of other policies to promote healthy behavior. In some cases, these could reinforce the price effect, e.g. a media campaign coinciding with a price change. In other cases, there may be reverse effect; for example, the use of licensing re- strictions to regulate access to alcohol or tobacco can create a higher time cost, making money price relatively less impor- tant than in settings where access is less restricted. The consistent finding, however, is that price elasticities are negative and therefore a tax (subsidy) on unhealthy (healthy) goods, leading to a price increase (reduction), will reduce (increase) the quantity purchased, all other things being equal. One of the things that may not remain equal is household income; if incomes increase more rapidly than the price of unhealthy goods, then this will make them more affordable and dilute any effect of a tax increase. Thus, it is the real terms price change (adjusted for income) that determines the impact of a tax or subsidy. Note that a change in tax or subsidy also has an indirect effect on real income. For a health behavior such as smoking, where quitting confers more health benefits than cutting down, it is also possible to estimate the effect of tax changes (the tax elasticity of quitting) with appropriate data.2 Similarly, the effect on smoking initiation can also be estimated. Producer perspective Although the consumer or purchaser of goods and services is most often seen as the focus for government fiscal in- terventions and is generally given the most attention in the discussion of taxes or subsidies, these can also be directed toward changing producer behavior. Determining the target or subject of a tax or subsidy is one feature of designing a fiscal intervention,3 and in some contexts, intervening with pro- ducers may be a more efficient approach. A brief consider- ation of the producer or supply side of the market and their potential role in determining what is supplied and at what price follows. In the basic economic model, consumers are price-takers; i.e. they respond to whatever price is set by the market. This model also assumes that there are many firms competing to supply a particular good or service, which has identical characteristics, and this process of competition will deter- mine the market price. As with the consumer side of the https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7182 market, this model abstracts from reality to provide a starting point for analyzing producer behavior. Just as maximizing utility is taken to be the aim of consumer behavior, profit- maximizing is the aim of producer behavior. To achieve this, producers have to be cost-minimizers, in the sense that they produce a specified good efficiently. If costs are reduced by changing quality, then this is, in effect, a different good. Pro- ducers will, therefore, respond to changes in the costs that they face, just as consumers respond to price changes. Thus, fiscal interventions on the producer side of the market have the potential to change what is consumed. Intervening in the market Within the economic model outlined, intervention can be justified where there is deviation from the conditions of per- fect competition, and this occurs rather frequently. External- ities provide one such justification. External costs (or benefits) arise when costs (or benefits) accrue to others in society and are not taken into account in the market. This provides the main economic justification for taxing harmful goods or sub- sidizing beneficial ones. In strict economic terms, the tax or subsidy should be set at a level that equates the private cost (benefit) with the social cost (benefit) at the margin that is to say measured on the last unit purchased, sometimes referred to as a Pigouvian tax.4 This approach faces a number of difficulties, not least of which is calculating the external cost where there can be considerable debate about what counts as an external cost as well as how to measure it.5 Even if the external cost can be agreed, the relationship between consumption of an un- healthy product and the resulting harms is not necessarily linear. Taking alcohol as an example, health risks are not linear, and other harms, including antisocial behavior, may only arise with excessive consumption. The optimal tax might be one which increases with consumption, and this seems impractical. The focus of these debates has been almost exclusively based on considering the externalities as arising from con- sumer decisions rather than producer decisions. While pro- ducers will argue that they respond to consumer demand, production decisions such as whether to produce a more or less healthy version of a particular good will not take account of potential external costs or benefits. They will take account of whether there are different input costs and what the effect will be on demand for the product. From an economic perspective, it is just as valid to tax or subsidize inputs to the production process that generate the externalities that arise from the consumption of the product. While the increase or reduction in production cost will be reflected in the final price to consumers, this approach provides an incentive for pro- ducers to reformulate products. If the existence of external costs (or benefits) provide an economic rationale for intervening in markets, other major challenges to the conventional economic model come from the recognition that individuals are not always fully informed, particularly when faced with many alternatives, and prefer- ences may not be complete or consistent, particularly over time. Imperfections also exist on the producer side of the market. Competition may be limited by the concentration of production or retailing in large companies. Various forms of product promotion may be used to increase sales. Lower pri- ces for larger sizes (volume discounts) may be justified by economies of scale (lower costs) in production but will in- crease purchasing and, potentially, the amount consumed per occasion. This is particularly true in relation to low price offers to upgrade portion size or add side dishes to a meal. The very complexity of real world decision-making may lead to con- sumers using shortcuts (decision heuristics) when taking de- cisions, and the implications of this for fiscal interventions will be picked up in the discussion. Evidence for the effect of taxes, subsidies, and prices The relevance and impact of fiscal instruments in advancing public health goals has been varied and depends to a certain extent on characteristics of the health behavior being addressed and the object (good or activity) to which the fiscal instrument is applied. In this section, an overview of evidence is presented for different health behaviors which exemplify some of the different challenges they pose. Smoking There has been widespread use of taxation over an extended period of time to address smoking behavior. Substantial levels of taxation have been applied to tobacco in the United Kingdom, partly as a revenue raising activity but increasingly as a public health measure, as the evidence of health and social harms has accumulated. Currently, the tax on a typical packet of 20 cigarettes can account for around 80% of the selling price and comprises a flat duty of £4.57 plus 16.5% of the retail price plus value added tax (VAT) of 20%. As a result, there has been considerable impact on affordability, and this has contributed to reduction in consumption. It should also be noted that the flat rate element of the tax ensures that the percentage tax is higher on cheaper products, which means that the affordability of the cheapest product is affected most by the tax. Tobacco has become 40% less affordable in the United Kingdom since 1980; much of this effect has been concen- trated in the more recent period between 2006 and 2016, when tobacco became 27% less affordable.6 Increasing tax rates may not in itself ensure that tobacco products become less affordable as this may be affected by the response of pro- ducers or retailers. While evidence from the United States7 suggests that cigarette taxes are overshifted (prices to con- sumers increase by more than the tax), in the United Kingdom, part of recent tax increases have been absorbed by producers or retailers resulting in under-shifting, particularly for cheaper products.8 Smoking prevalence has fallen steadily in the United Kingdom from around 40% in 1980 to 16% in 2016.9 Accurately measuring the impact of tax policy on health behaviors is confounded by the coterminous use of several policy in- struments. Thus, tobacco control policy in the United Kingdom has not relied solely on tax measures but an array of interventions including public health information, individual https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7 183 support for smoking cessation (including the development of nicotine replacement therapy (NRT) and other products to reduce craving), regulation of advertising including at point of sale, and most recently plain packaging. However, Forster and Jones2 estimated a tax elasticity of quitting of �0.6 for men and �0.46 women using British data from 1984, which would predate much of the increase in use of other interventions. These figures are interpreted as years of smoking avoided as a result of quitting; thus, a 10% real increase in tax would, on average, produce a 6-year reduction in years of smoking by men. Furthermore, it is estimated that doubling the price of tobacco in real terms worldwide would reduce prevalence of smoking by a third.10 Alcohol consumption Alcohol consumption shares some characteristics with smoking but also some differences. Both behaviors can be considered non-essential; i.e. it is not necessary for anyone to smoke or drink alcohol. However, while there are clear health risks associated with any smoking, there is less consensus about health risks associated with any consumption of alcohol as opposed to excessive consumption of alcohol. A large ma- jority of the adult population consume some alcohol, whereas smokers are a decreasing minority. However, clear associa- tions have been shown between tax levels and health harms from alcohol.11,12 There is a long history of taxing alcohol as a revenue raising mechanism, but the focus on tax as a public health measure is more recent. The UK taxation of alcohol is com- plex; while VAT is levied uniformly at 20%, alcohol duty rates are applied differently to different products and are not directly based on alcoholic strength, resulting in a tax rate per unit of alcohol that varies both between products and, in some cases, within products. Spirits are the only product where duty is proportional to alcoholic strength and is currently 28.74p per unit of alcohol. However, the duty per unit on the lowest strength of still cider would be nearly the same (27p per unit), whereas duty on strong cider would be only 7-8p; this is because cider is taxed on the volume of product, within bands, and not on the volume of alcohol. In contrast with the situation with smoking, alcohol has become 60% more affordable since 1980, despite tax in- creases,13 although the increase in affordability has reversed more recently. Alcohol consumption has begun to fall more recently; the volume of alcohol released for sale per adult in the United Kingdom peaked in 2004/5,14 whereas the increase in affordability peaked a little later in 2007.13 This more recent reduction in the affordability of alcohol in the United Kingdom owes more to falling real incomes than to higher taxation. One reason for the increase in affordability, particularly with respect to alcohol sold for home consumption (off-sales), has been the failure to pass on tax increases,15 with supermarkets often promoting alcohol products as loss leaders, with the aim of increasing footfall and overall sales. Legislation has been introducedto prevent‘below cost’ selling;i.e. alcohol cannotbe sold for less than the combined duty and VAT payable. How- ever, products can still be sold for less than their true cost. It was in response to the specific problem of very cheap alcohol fueling problem drinking that the Scottish Government passed legislation in 2012 to allow the introduc- tion of a different type of price intervention, minimum unit pricing (MUP) per unit alcohol. This establishes a ‘floor’ price across all types of alcohol product below which alcohol cannot be sold. The relative effectiveness of increasing the lowest prices in the alcohol market, compared with increasing the average price, had been demonstrated by Gruenewald et al.16 It was shown that a price increase targeting lower cost prod- ucts would reduce the volume of alcohol sold by 4.2% compared with a reduction in sales of 1.7% when prices increased across all products. Some Canadian provinces operate a similar policy of Social Reference Pricing; increasing this minimum price has been shown to be associated with reductions in alcohol consumption and alcohol-related health harms.17,18 Legal challenges delayed the introduction of MUP until 2018, and the effects are currently being evaluated. However, modeling of the policy's effect on problem drinking suggests it will be more effective than large tax increases in reducing alcohol consumption by the heaviest drinkers.19 MUP cannot be described as a fiscal intervention, but it is a price-based intervention that affects the supply of alcohol products. Scotland had previously introduced other re- strictions on ‘irresponsible’ price promotions, which were defined as those which would potentially encourage increased alcohol consumption over a period of time. This was initially targeted at ‘happy hour’ offers and other short-term price reductions within licensed premises but was later extended to all volume-based price discounts. Again, this is aimed at restricting some aspects of price setting on the supply side of the market where these have been associated with adverse drinking behavior. Food taxes and subsidies Moving into the area of food and healthy diets provides different challenges for policymakers when compared with smoking and alcohol. Food is an essential commodity, and the challenge is to implement taxes, subsidies, or other price in- terventions that will promote healthier choices without causing potential hardship to low income groups. The aim of fiscal interventions should be to achieve a healthy diet through increasing the consumption of healthy foods, reducing the consumption of unhealthy foods, and reducing excessive calorie intake. Healthy foods include those which are less processed and are high in fiber and important nutri- ents such as fish oils. Unhealthy foods are those which are high in fat, sugar, and salt and usually highly processed and energy dense. In this context, it is particularly important to consider what is to be taxed or subsidized but also how the tax is to be applied. Taxes or subsidies could be applied to particular foods and to particular nutrients or could be applied on the basis of calorie content or energy density.3 Most evidence in this area relates to the effect of in- terventions aimed at directly increasing prices paid by con- sumers for specific products. Recently, however, attention has been shifting to other uses of taxation which may be more suitable for the complexities of improving diets. Taxes which target nutrients rather than products may be more effective,20 and if these are taxed as inputs to food production, they may induce behavior change in producers. The soft drink industry https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7184 levy, which was introduced in the United Kingdom from April 2018, targets the sugar content of certain sugar-sweetened beverages (SSBs). The tax is levied on producers of soft drinks based on their production of soft drinks within two bands of sugar content, with a zero levy for drinks containing less than 5 g of sugar per 100 mL. A stated aim of this policy was to encourage refor- mulation of soft drinks with a clear incentive for producers to reduce sugar content to avoid or reduce the tax. In this respect, the levy has been successful with the government reporting that more than 50% of manufacturers had refor- mulated drinks.21 However, the decision on whether to pass on the levy to customers, in part or in full, by charging higher prices for drinks with a higher sugar content is left in the hands of the suppliers (producers and retailers). The most obvious marker of the health issues relating to food intake is the increased prevalence of obesity,22 which almost doubled from 15% of adults in 1993 to 26% in 2016. This is, of course, also related to increasingly sedentary lifestyles (see Physical activity subsection below). Considering general price trends, until relatively recently food was becoming more affordable in the United Kingdom. Food prices fell in real terms by nearly 30% from 1980 to 2007.23 This was followed by a sharp increase in prices as a fall in real incomes combined with increasing prices; real prices began to fall again from 2014. The effect of this natural change in the affordability of food reveals interesting evidence about consumer responses, with consumption of relatively unhealthy food categories being protected, partly by trading down to cheaper alterna- tives, and consumption of fruit and vegetables falling.23 There have been relatively few evaluations of imple- mented tax initiatives relating to unhealthy foods, although there have been many modeling studies based on price elas- ticity data. A recent review by Marron et al. in 2015, specifically relating to taxation of unhealthy food and drink, suggests that well-designed taxes may be helpful, but the evidence to date is rather limited.24 In particular, more evidence is required both on substitution effects and on the response of food industry.25 Substitution effects will not necessarily undermine specific food taxes26 but need to be carefully considered. The most frequent target for taxation has been SSBs, and mixed results have been reported. Very small taxes, mainly as a revenue raising device, produce limited effects. Collins et al.27 estimated that a 10% price increase would reduce SSB consumption in the United Kingdom by 4.6%, and a 20% price increase would reduce consumption by 9.1%. Evaluation of an 8% tax on non-essential foods in Mexico28 indicates that the effect of the tax in reducing purchases increased over time (year two compared to year 1 post tax) and that the effect was concentrated in households defined as unhealthy purchasers pretax. In Hungary, taxing unhealthy foods produced a shift from processed to unprocessed foods, with the lowest income groups being the most responsive.29 Although the tax on fat in Denmark was short-lived, it did impact on household purchases.30,31 Rather less evidence is available on the potential effect of subsidies for healthy eating, although these are included in some modeling studies. Specific interventions to promote healthier eating have tended to take the form of vouchers for fruit and vegetable purchases usually targeted at low-income households. The UK Healthy Start scheme, for example, increased expenditure on fruit and vegetables by 15%.32 These are not strictly a fiscal measure as they are a form of income supplement, reducing the price of voucher purchases to zero and potentially freeing income to spend on other things. However, insights from behavioral economics suggest that there is a tendency for consumers to earmark money for specific purposes (known as mental accounting) such that food vouchers will increase total food expenditure.33 In the Healthy Start scheme, expenditure only increased among re- cipients who previously spent less than the voucher value.32 Physical activity Using taxes, subsidies or price-based interventions to increase physical activity presents a different set of challenges when compared with the other health behaviors considered previ- ously. Apart from some extreme exceptions, everyone spends some time in some level of physical activity and is sedentary for some time. While shifting the balance toward greater physical activity may involve financial costs for some activ- ities, the main cost to individuals may be time costs, which have been shown to be a greater barrier to physical activity than monetary costs.34 Physical activity required in the workplace, which could be considered as physically active time which was being paid for, has declined over time. There is very limited evidence relating to the use of taxes, subsidies, or price-based interventions relating to physical activity and sedentary behavior. A systematic review35 iden- tified just 13 papers, and this included evaluations of transfer payments and the effect of congestion charging, where there would be an indirect effect of improving the environment for active travel. Subsidies to public transport could also be seen as pro- moting active travel but are primarily aimed at meeting other policy objectives. In Canada, schemes to provide tax credits or refunds for enrolling children in physical activity programs and exemptions from sales tax for sports-related goods have been introduced at federal and provincial levels36 but do not appear to have been evaluated in terms of increased physical activity. Discussion From the perspective of economics, the most controversial aspect of any intervention in the market is the potential violation of consumer sovereignty. The economic principle underlying fiscal interventions is that they should be market correcting; i.e. they should ensure that the prices consumers face reflect the social cost of their choices. In practice, the optimal level of tax or subsidy can be difficult to determine, and fiscal interventions are more often judged on their effectiveness in reducing health and social harms and pro- moting positive health and social outcomes. There is a strong argument that the choice of interventions to promote health should be based on the costs and benefits to society, including the individuals whose choices are affected. The best in- terventions, or combination of interventions, would have the highest ratio of benefits to costs. https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7 185 The available evidence relating to fiscal levers is variable across different health behaviors. Far more use has been made of taxation for tobacco and alcohol than for diet and physical activity, partly reflecting the greater complexity of these latter behaviors and partly the fact that they are so embedded in daily activity. Levels of taxation for tobacco have been kept sufficiently high to reduce affordability and make a contribution to reducing smoking prevalence. However, alcohol has until recently become more affordable as tax in- creases have not kept pace with real incomes. Different ap- proaches to targeting the cheapest alcohol products are now being evaluated. Fiscal levers aim to alter the prices that consumers face, but cost includes other factors particularly time. Convenience foods, which free up time from food preparation but may be energy dense, have played a part in fueling the increase in obesity. So too has the increasing availability of food outside the home. While the underlying economic principle that higher prices will reduce demand will hold the scale of effect is less predictable. Other interventions apart from fiscal policy can increase the non-monetary cost of unhealthy behaviors, for example, regulations that impact on availability of un- healthy products increase the time and effort involved in acquiring them. Smarter design of fiscal instruments could provide more effective interventions. Ensuring that affordability is taken into account is one of the principles suggested for ‘smart’ taxation.37 In relation to food, taxes (or subsidies) that give manufacturers incentives to reformulate products may have greater impact than taxes applied directly to the products. The design of fiscal interventions also needs to be informed by consideration of potential behavioral responses; particu- larly, the likelihood that some consumers will switch to cheaper versions of the same foods, which may be even less healthy. Behavioral economics uses insights from psychology to investigate the systematic biases in decision-making that cannot be addressed satisfactorily within the conventional model. Within this model, the failure of individuals to follow their own best interests has been described as an ‘internality’ or within-person externality,38,39 but there is considerable debate within economics over who decides what the better choice is when assisting individuals to make better choices and the value judgments this implies. However, there is po- tential for behavioral economics to contribute to the design of better interventions. Perhaps, one of the most important insights from behavioral economics is that consumers apply shortcuts when facing complex decisions, and these shortcuts pro- duce systematic bias. For example, status quo bias suggests consumers are more likely to choose the same foods or familiar foods rather than consider all the alternatives. Introducing a fiscal intervention may then be less effective unless the consumer is aware of the price change. If the consumer is unaware of the price change, then the inter- vention is not salient to their decision-making. Decision- making can also be affected by context or feelings (the affect heuristic); for example, consumers make a different, less healthy, food choice when they are feeling hungry. It has been suggested that this bias could be addressed by per- commitment devices such as pre-ordering lunch at school or in work places. Interventions to address these and other potential biases in decision-making have been popularized as ‘nudges’.40 How- ever, the most successful examples of applications tend to be those where there is an obvious ‘default’ choice which can be reset.41 There is, however, potential for a broader range of applications of behavioral interventions that go beyond nudging. Galizzi42 proposes clusters of policy formulation in- struments where tax and subsidy and nudges are seen, respectively, as purely conventional and purely behavioral. However, the use of information and incentives can be con- ventional or behaviorally inspired. These clusters can also be debated, but the principle that there is potential overlap or symbiosis between conventional and behavioral approaches appears sound. This awareness may lead to better design of interventions. Equity considerations are also a factor when considering the role of fiscal instruments, particularly as attention moves beyond the so called ‘sin taxes’ on tobacco and alcohol. Food taxation might be limited to discretionary or treat items but would limit their scope considerably while still having potentially adverse effects on low-income households. Sub- sidizing healthy foods could offset the higher costs of taxed foods, but if applied across the board could be more beneficial to higher income households. Conclusion The use of tax (or subsidy) to affect individual choice has sometimes been described as a shove, rather than a nudge, ‘but though nudges certainly have their place, occasionally a good shove advances individual and social welfare considerably more.’43 Debate around conventional approaches and behav- ioral approaches to health behaviors will no doubt continue, both from the perspective of adherence to economic principles and in terms of effectiveness in reducing health harms and other social costs. There is clearly scope for a middle-ground based on what works in a particular context and for designing better interventions by incorporating conventional and behavioral perspective. The greatest challenges may lie in developing effective interventions where the emphasis is on increasing healthy choices through subsidies or nudges. Author statements Ethical approval None sought. Funding HERU is core funded by the Chief Scientist Office, Scottish Government Health and Social Care Directorates, and the University of Aberdeen. Competing interest None declared. https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7186 r e f e r e n c e s 1. World Health Organisation. The top 10 causes of death. http:// www.who.int/mediacentre/factsheets/fs310/en/ [Accessed 19 February 2018]. 2. Forster M, Jones AM. The role of tobacco taxes in starting and quitting smoking: duration analysis of British data. J Roy Stat Soc 2001;164(3):517e47. 3. Jensen J, Smed S. State-of-the-art for food taxes to promote public health. Proc Nutr Soc 2018;77(2):100e5. https://doi.org/ 10.1017/S0029665117004050. 4. Pigou AC. The economics of welfare. London: Macmillan; 1920. http://lf-oll.s3.amazonaws.com/titles/1410/0316_Bk . [Accessed 21 March 2018]. 5. Bhattacharya A. Which cost of alcohol? What should we compare it against? Addiction 2016. https://doi.org/10.1111/ add.13335. 6. NHS Digital. Statistics on smoking, England. 2017. https://digital. nhs.uk/catalogue/PUB24228. [Accessed 13 March 2018]. 7. Sullivan Ryan S, Dutkowsky Donald H. The effect of cigarette taxation on prices: an empirical analysis using local-level data. Publ Finance Rev 2012;40(6):687e711. 8. Hiscock R, Branston JR, McNeill A, Hitchman SC, Partos TR, Gilmore AB. Tobacco industry strategies undermine government tax policy: evidence from commercial data. Tob Contr 09 October 2017. https://doi.org/10.1136/tobaccocontrol- 2017-053891. 9. ONS. Adult smoking habits in the UK: 2016. 2017. https://www. ons.gov.uk/peoplepopulationandcommunity/ healthandsocialcare/healthandlifeexpectancies/bulletins/ adultsmokinghabitsingreatbritain/2016. [Accessed 7 March 2018]. 10. Jha P, Peto R. Global effects of smoking, of quitting, and of taxing tobacco. N Engl J Med 2014;370(1):60e8. 11. Elder RW, Lawrence B, Ferguson A, Naimi TS, Brewer RD, Chattopadhyay SK, Toomey TL, Fielding JE. The effectiveness of tax policy interventions for reducing excessive alcohol consumption and related harms. Am J Prev Med 2010;38(2):217e29. 12. Wagenaar AC, Tobler AL, Komro KA. Effects of alcohol tax and price policies on morbidity and mortality: a systematic review. Am J Public Health 2010;100:2270e8. 13. NHS Digital. Statistics on alcohol, England e 2017. 2017. https:// digital.nhs.uk/catalogue/PUB23940. [Accessed 13 March 2018]. 14. HMRC. Alcohol factsheet 2012e13. 2013. https://www. uktradeinfo.com/Statistics/StatisticalFactsheet/Pages/ FactsheetArchive.aspx. [Accessed 27 March 2018]. 15. Ally AK, Meng Y, Chakraborty R, Dobson PW, Seaton JS, Holmes J, et al. Alcohol tax pass-through across the product and price range: do retailers treat cheap alcohol differently? Addiction 2014;109:1994e2002. 16. Gruenewald PJ, Ponicki WR, Holder HD, Romelsj€o A. Alcohol prices, beverage quality, and the demand for alcohol: quality substitutions and price elasticities. Alcohol Clin Exp Res 2006;30(1):96e105. 17. Stockwell T, Auld MC, Zhao J, Martin G. Does minimum pricing reduce alcohol consumption? The experience of a Canadian province. Addiction 2012;107(5):912e20. 18. Stockwell T, Zhao J, Martin G, et al. Minimum alcohol prices and outlet densities in British Columbia, Canada: estimated impacts on alcohol-attributable hospital admissions. Am J Public Health 2013;103:2014e20. https://doi.org/10.2105/ AJPH.2013.301289. 19. Angus C, Holmes J, Pryce R, Meier P, Brennan A. Model-based appraisal of the comparative impact of minimum unit pricing and taxation policies in Scotland: an adaptation of the Sheffield alcohol policy model version 3, ScHARR. University of Sheffield; 2016. 20. Harding M, Lovenheim M. The effect of prices on nutrition: comparing the impact of product- and nutrient-specific taxes. J Health Econ 2017;53:53e71. 21. https://www.gov.uk/government/news/soft-drinks-industry- levy-comes-into-effect published 5 April 2018 accessed 20th November 2018. 22. NHS Digital. Statistics on obesity, physical activity and diet, England. 2018. https://files.digital.nhs.uk/publication/0/0/ obes-phys-acti-diet-eng-2018-rep . 23. DEFRA. Food Statistics Pocketbook. 2017. https://www.gov.uk/ government/uploads/system/uploads/attachment_data/file/ 608426/foodpocketbook-2016report-rev-12apr17 . 24. Marron Donald, Gearing Maeve, Iselin John. Should we tax unhealthy foods and drinks? Washington, DC: Urban Institute; 2015. https://www.urban.org/sites/default/files/alfresco/ publication-pdfs/2000553-Should-We-Tax-Unhealthy-Foods- and-Drinks . 25. Cornelsen L, Green R, Dangour A, Smith R. Why fat taxes won't make us thin. J Public Health 2014;37(1):18e23. 26. Finkelstein EA, Zhen C, Bilger M, Nonnemaker J, Farooqui AM, Todd JE. Implications of a sugar-sweetened beverage (SSB) tax when substitutions to non-beverage items are considered. J Health Econ 2013;32:219e39. 27. Collins B, Capewell S, O'Flaherty M, Timpson H, Razzaq A, Cheater S, Ireland R, Bromley H. Modelling the health impact of an English sugary drinks duty at national and local levels. PLoS One 2015;10(6):e0130770 [Electronic Resource]. 28. Taillie LS1, Rivera JR, Popkin BM, Batis C. Do high vs. low purchasers respond differently to a nonessential energy- dense food tax? Two-year evaluation of Mexico's 8% nonessential food tax. Prev Med 2017;105:S37e42. https:// doi.org/10.1016/j.ypmed.2017.07.009. 29. Biro A. Did the junk food tax make the Hungarians eat healthier? Food Policy 2015;54:107e15. 30. Jensen JB, Smed S. The Danish tax on saturated fat e short run effects on consumption, substitution patterns and consumer prices on fats. Food Policy 2013;42:18e31. 31. Smed S, Scarborough P, Rayner M, Jensen JD. The effects of the Danish saturated fat tax on food and nutrient intake and modelled health outcomes: an econometric and comparative risk assessment evaluation. Eur J Clin Nutr 2016;70(6):681e6. 32. Griffith R, von Hinke S, Smith S. Getting a healthy start: the effectiveness of targeted benefits for improving dietary choices. Health Econometrics and Data Group Working Paper 15/10 University of York; 2015. 33. Just DR, Mancino L, Wansink B. Could behavioral economics help improve diet quality for nutrition assistance program participants?. 2007. United States Department of Agriculture Economic Research Report Number 43. 34. Anokye NK, Pokhrel S, Fox-Rushby J. Economic analysis of participation in physical activity in England: implications for health policy. Int J Behav Nutr Phys Act 2014;11(117). http:// www.ijbnpa.org/content/11/1/117. 35. Shemilt I, Hollands GJ, Marteau TM, Nakamura R, Jebb SA, Kelly MP, et al. Economic instruments for population diet and physical activity behaviour change: a systematic scoping review. PLoS One 2013;8(9):e75070. https://doi.org/10.1371/ journal.pone.0075070. 36. von Tigerstrom B, Larre T, Sauder J. Using the tax system to promote physical activity: critical analysis of Canadian initiatives. Am J Public Health 2011;101(8):e10e6. https:// doi.org/10.2105/AJPH.2011.300201. 37. Jha P, MacLennan M, Chaloupka FJ, Yurekli A, Ramasundarahettige C, Palipudi K, Zaton�ksi W, Asma S, Gupta PC. Global hazards of tobacco and the benefits of smoking cessation and tobacco taxes. In: Gelband H, Jha P, Sankaranarayanan R, Horton S, editors. Disease control priorities. 3rd ed.vol. 3. World Bank Group; 2015. https:// http://www.who.int/mediacentre/factsheets/fs310/en/ http://www.who.int/mediacentre/factsheets/fs310/en/ http://refhub.elsevier.com/S0033-3506(19)30032-0/sref2 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref2 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref2 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref2 https://doi.org/10.1017/S0029665117004050 https://doi.org/10.1017/S0029665117004050 http://lf-oll.s3.amazonaws.com/titles/1410/0316_Bk https://doi.org/10.1111/add.13335 https://doi.org/10.1111/add.13335 https://digital.nhs.uk/catalogue/PUB24228 https://digital.nhs.uk/catalogue/PUB24228 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref7 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref7 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref7 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref7 https://doi.org/10.1136/tobaccocontrol-2017-053891 https://doi.org/10.1136/tobaccocontrol-2017-053891 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/bulletins/adultsmokinghabitsingreatbritain/2016 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/bulletins/adultsmokinghabitsingreatbritain/2016 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/bulletins/adultsmokinghabitsingreatbritain/2016 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/bulletins/adultsmokinghabitsingreatbritain/2016 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref12 https://digital.nhs.uk/catalogue/PUB23940 https://digital.nhs.uk/catalogue/PUB23940 https://www.uktradeinfo.com/Statistics/StatisticalFactsheet/Pages/FactsheetArchive.aspx https://www.uktradeinfo.com/Statistics/StatisticalFactsheet/Pages/FactsheetArchive.aspx https://www.uktradeinfo.com/Statistics/StatisticalFactsheet/Pages/FactsheetArchive.aspx http://refhub.elsevier.com/S0033-3506(19)30032-0/sref15 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref15 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref15 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref15 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref15 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref17 https://doi.org/10.2105/AJPH.2013.301289 https://doi.org/10.2105/AJPH.2013.301289 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref19 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref19 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref19 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref19 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref20 https://www.gov.uk/government/news/soft-drinks-industry-levy-comes-into-effect%20published%205%20April%202018%20accessed%2020th%20November%202018 https://www.gov.uk/government/news/soft-drinks-industry-levy-comes-into-effect%20published%205%20April%202018%20accessed%2020th%20November%202018 https://www.gov.uk/government/news/soft-drinks-industry-levy-comes-into-effect%20published%205%20April%202018%20accessed%2020th%20November%202018 https://files.digital.nhs.uk/publication/0/0/obes-phys-acti-diet-eng-2018-rep https://files.digital.nhs.uk/publication/0/0/obes-phys-acti-diet-eng-2018-rep https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/608426/foodpocketbook-2016report-rev-12apr17 https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/608426/foodpocketbook-2016report-rev-12apr17 https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/608426/foodpocketbook-2016report-rev-12apr17 https://www.urban.org/sites/default/files/alfresco/publication-pdfs/2000553-Should-We-Tax-Unhealthy-Foods-and-Drinks https://www.urban.org/sites/default/files/alfresco/publication-pdfs/2000553-Should-We-Tax-Unhealthy-Foods-and-Drinks https://www.urban.org/sites/default/files/alfresco/publication-pdfs/2000553-Should-We-Tax-Unhealthy-Foods-and-Drinks http://refhub.elsevier.com/S0033-3506(19)30032-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref27 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref27 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref27 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref27 https://doi.org/10.1016/j.ypmed.2017.07.009 https://doi.org/10.1016/j.ypmed.2017.07.009 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref33 http://www.ijbnpa.org/content/11/1/117 http://www.ijbnpa.org/content/11/1/117 https://doi.org/10.1371/journal.pone.0075070 https://doi.org/10.1371/journal.pone.0075070 https://doi.org/10.2105/AJPH.2011.300201 https://doi.org/10.2105/AJPH.2011.300201 https://openknowledge.worldbank.org/bitstream/handle/10986/22552/9781464803499 https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 0 e1 8 7 187 openknowledge.worldbank.org/bitstream/handle/10986/ 22552/9781464803499 . 38. Herrnstein R, Loewenstein G, Prelec D, Vaughan Jr W. Utility maximisation and melioration: internalities in individual choice. J Behav Dec Mak 1993;6(3):149e85. 39. Bhargava S, Loewenstein G. Behavioral economics and public policy 102: beyond nudging American economic review. Pap Proc 2015;105(5):396e401. 40. Thaler RH, Sunstein CR. Nudge: improving decisions about health. Wealth, and Happiness New Haven CT: Yale University Press; 2008. 41. French R, Oreopoulos P. Applying behavioural economics to public policy in Canada. Can J Econ 2017;50:599e635. https:// doi.org/10.1111/caje.12272. 42. Galizzi Matteo M. Behavioral aspects of policy formulation: experiments, behavioral insights, nudges. In: Howlett Michael, Mukherjee Ishani, Fraser Simon, editors. Handbook of policy formulation. Handbooks of research on public policy. Cheltenham, UK: Edward Elgar Publishing; 2017. ISBN 9781784719319. 43. Loewenstein G, Asch DA, Friedman JY, Melichar LA, Volpp KG. Can behavioural economics make us healthier? BMJ 2012;344. https://doi.org/10.1136/bmj.e3482. https://openknowledge.worldbank.org/bitstream/handle/10986/22552/9781464803499 https://openknowledge.worldbank.org/bitstream/handle/10986/22552/9781464803499 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref40 https://doi.org/10.1111/caje.12272 https://doi.org/10.1111/caje.12272 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30032-0/sref42 https://doi.org/10.1136/bmj.e3482 https://doi.org/10.1016/j.puhe.2019.02.008 https://doi.org/10.1016/j.puhe.2019.02.008 Fiscal measures to promote healthier choices: an economic perspective on price-based interventions Introduction Economic framework Consumer perspective Producer perspective Intervening in the market Evidence for the effect of taxes, subsidies, and prices Smoking Alcohol consumption Food taxes and subsidies Physical activity Discussion Conclusion Author statements Ethical approval Funding Competing interest References Toward-patient-centered-care-and-inclusive-health-care-governan_2019_Public- ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4 Available online at w Public Health journal homepage: www.elsevier.com/puhe Review Paper Toward patient-centered care and inclusive health- care governance: a review of patient empowerment in the UAE V. Bodolica a,*, M. Spraggon b,1 a American University of Sharjah, School of Business Administration, P.O. Box 26666, Sharjah, United Arab Emirates b Mohammed Bin Rashid School of Government (MBRSG), Convention Tower, Level 7, P.O. Box 72229, Dubai, United Arab Emirates a r t i c l e i n f o Article history: Received 10 May 2018 Received in revised form 14 January 2019 Accepted 31 January 2019 Available online 14 March 2019 Keywords: Patient empowerment Patient-centered care Health-care governance Policy making United Arab Emirates * Corresponding author. Tel.: þ(971) 56 10186 E-mail addresses: virginia.bodolica@hec.c 1 Tel.: þ(971) 56 6069447. https://doi.org/10.1016/j.puhe.2019.01.017 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: The purpose of this article was twofold. We aimed to both clarify the multidi- mensional notion of patient empowerment (PE) and conduct a comprehensive survey of PE-related literature in the specific context of the United Arab Emirates (UAE). Study design: The study objectives were achieved by means of a two-phased systematic review of the literature on PE and associated dimensions. Methods: The first phase consisted in the database search for recent review articles on the construct of PE that were published in the past five years. The second phase focused on the identification of extant empirical research on PE and related concepts in UAE settings. In total, 13 review articles and 17 empirical studies were eligible and included in our analysis. Results: The retained PE review articles pointed to two major themes and four topics on ‘conceptual clarification’ and ‘contextual embeddedness’, where PE was tackled in relation to national health-care system, health-care governance, information technology, and therapeutic continuum. Our analysis of UAE-based PE studies unveiled three themes on ‘chronic disease care’ (with three topics of ‘general inquiries’, ‘diabetes management’, and ‘diabetic complications’), ‘self-medication with drugs’, and ‘non-therapeutic in- terventions’. By juxtaposing the identified PE themes and topics, we derived three prom- ising opportunities for researchers, practitioners, and policymakers to consolidate, expand, and initiate relevant PE interventions in the UAE. Conclusion: This review article found that PE represents an emergent and underexplored notion in the UAE health-care system. As UAE ambitions to become a sought-after medical hub in the global arena, the design and implementation of adequate PE strategies and reforms play a critical role in the development of a world-class patient-centered health care in the country. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. 88. a (V. Bodolica), martin.sp ic Health. Published by E raggon-hernandez@hec.ca (M. Spraggon). lsevier Ltd. All rights reserved. mailto:virginia.bodolica@hec.ca mailto:martin.spraggon-hernandez@hec.ca http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.017&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4 115 Introduction Giving patients a greater voice and more power in their interaction with suppliers of health care is a national priority in many countries around the globe.1e4 Patient empower- ment (PE) emerged as a focal element in governmental ef- forts to achieve patient-centeredness for inducing enhanced self-care efficacy, increased medication adherence, opti- mized resource usage, and improved citizens’ well-being.5e7 This trend was accompanied by a growth in studies on PE along the therapeutic continuum,8,9 resulting in many con- ceptualizations and measurements to operationalize this notion.10,11 PE is defined as the perceived ability of the pa- tient to self-manage own health by getting involved in de- cisions and assuming responsibility for choices affecting personal health.12 Recognizing the challenges of applying the multidimen- sional PE concept in empirical settings, scholars synthesized extant knowledge on PE and related dimensions to generate a unified measurement scale.13 Although much research was conducted on PE in Western nations, little is known about PE initiatives in the emerging market context of the United Arab Emirates (UAE). The UAE government pursues continuous improvement across an integrated health-care system, including institutional and service quality, resource usage and cost control, and actual health outcomes for its pop- ulation.14e16 To achieve this strategic priority, while adapting to the needs of a changing demographic profile and account- ing for the surge of several chronic diseases, the UAE embarked on a program of medical reforms and renova- tions.15 Determined to secure a leading position in interna- tional rankings, the country aspires to develop a world-class health-care system and transform itself into a sought-after medical hub in the global arena.17e20 The notable progress that has been achieved in the UAE health-care sector has been well documented in academic/ practitioner publications.21e23 Yet, the extent to which members of the public have been empowered to take control of their health and participate in the design, delivery, and governance of health care remains unclear. No systematic effort has been deployed to analyze the UAE-based evidence on PE, and we bridge this gap via a two-phased literature re- view. This article offers a critical assessment of the current state of empirical PE research and related constructs in the cultural/regulatory framework of the UAE. Methods All the review procedures were conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines13,15,24 (Fig. 1). The pur- pose of the first phase was to clarify the notion of PE and related dimensions. A thorough database search was made to detect refereed articles published in the past five years that reviewed PE research. We focused deliberately on recent review articles as they offer a holistic analysis of the diversified PE-related literature cumulated over the years. Our search was made using ProQuest Health and Medical Complete, PubMed Central, and PsycInfo databases. To generate review-only PE articles, we used the keyword technique searching for ‘PE’ and ‘review’ words simulta- neously in ’ title/abstract. After screening the generated records, we removed duplicates and excluded non-reviews and non-PE papers. Five articles were identified by holding discussions with colleagues, but only two were retained as the other three were conceptual papers. The assessment of articles’ full texts for securing compliance with eligibility criteria was performed by two peers in parallel to increase confidence and eliminate bias. The eligibility considerations were as follows: articles had to review PE and related literature; be published in peer- reviewed English-language journals; and appear in press from 2013 onwards. Empirical and conceptual articles, tech- nical papers/doctoral theses, books/reports, policy briefings, and non-refereed contributions were excluded. The individ- ual results of each peer were compared, and one discrepancy emerged regarding a review on patient involvement.4 This discrepancy was addressed in a meeting, where the decision to drop this paper was made because of its lack of conceptual PE focus. A final check for sample comprehensiveness was per- formed using the ancestry approach of articles’ identifica- tion.25,26 It allows examining the reference lists of the most recent articles to determine whether new entries could be generated that were missed through database searching. We screened the references of the six retained articles published in 2016e2017 in search for non-identified reviews. Because this procedure did not yield additional eligible articles, our final sample of 13 PE-related reviews remained unchanged. These review articles were examined to derive an accurate definition and understanding of PE and associated dimensions. Islamic scholars highlighted ethical concerns regarding the relevance/timeliness of the principle of individual autonomy and right to self-determination in local settings on the basis of historic, religious, and sociocultural considerations of the Muslim population.22,27 These concerns could have delayed the widespread acceptance of PE in the UAE, affecting the amount of scholarly production on this topic. Because by focusing on PE exclusively, without considering closely related terms, we could have missed relevant literature, we kept our search as open/comprehensive as possible. Our analysis revealed that patient autonomy/self- determination, patient choice/voice, patient engagement/ participation, patient involvement/activism, patients’ rights, self-care/self-management, coping with disease, patient knowledge/information empowerment, and shared decision- making are used interchangeably to infer PE. These terms, separated by the Boolean operator ‘or’, were entered as key- words when performing the database search. We used the operator ‘and’ to limit the search to only UAE-based PE studies. To capture locally oriented articles, we alternated between the country name and its three-letter acronym and used the names of Abu Dhabi, Dubai, and Sharjah emirates. Because PE topicality is a recent occurrence in the Arab region, we focused on articles published over the past decade to generate an updated account of PE in the UAE. To complete the second phase of our survey, we followed the aforementioned procedures. To decide about the retention https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 Fig. 1 e Methods of the two-phased systematic review of the literature on PE in the UAE. PE, patient empowerment; UAE, the United Arab Emirates. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4116 of papers, we defined a slightly different set of eligibility criteria. The following conditions for inclusion should have been met: empirical articles on PE and associated constructs in the UAE, published in refereed English-language journals from 2008 onwards. Non-academic literature, conceptual pa- pers, and empirical studies that offered an aggregated anal- ysis of PE-related issues in several Arab states were excluded. This was the case of a breast cancer control strategies’ study in four geographical regions, where PE emerged as a consti- tutive dimension of ‘promoting advocacy’ theme.28 This article was excluded because its findings were presented indiscriminately for a group of 12 Arab nations, without separating the effects for each included country. The full texts of 75 generated articles were thoroughly examined to verify their eligibility. Because most of them did not pertain to any dimension of PE, only 16 papers were retained. By screening the references of a recently published survey on diabetes self-management in the Gulf,24 we iden- tified one pertinent entry. Each researcher from our team analyzed the contents of the retrieved studies and confirmed their inclusion in the review. The second phase of the pro- cess yielded a sample of 17 empirical articles on UAE-based PE issues. Results PE-related review articles Two themes emerged from the content analysis of 13 PE re- views: ‘conceptual clarification’ and ‘contextual embedded- ness’ (Table 1). Four first-theme articles are dedicated exclusively to elucidating the notion of PE and associated di- mensions. Each survey advances a different conceptual model to map PE-related constructs, highlighting the difficulty of making sense of this complex literature. Some authors focus on PE indicators and behaviors,29 whereas others uncover the antecedents, attributes, and consequences of PE.30 The concept is interpreted through the lens of an enabling process that leads to various outcomes10 or presented as a combina- tion of ability, motivation, and power.11 Among the most cited PE elements are patient's medical knowledge, coping skills, health literacy/education, information-seeking behavior, gaining control, sense of meaning, shared decision-making, self-care/self-manage- ment, and self-efficacy. PE is mapped in relationship with patient activation, enablement, engagement, involvement, https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 Table 1 e Recent literature review articles related to the concept of PE. Reference Context Emphasis PE elements and associated dimensions Findings and contribution Theme 1: conceptual clarification (4 articles) Bravo et al.29 Acontextual(concept focused) Conceptual model of PE (at patient, professional, and system level) Indicators (self-efficacy, knowledge, skills, personal control, health literacy, sense of meaning, feeling respected); behaviors (self-management, shared decision making, take part in groups, use Internet to search/share info) PE is a state ranging from low to high level; responsibilities of patients, providers and health-care system to use PE interventions to enhance clinical outcomes Cerezo et al.10 Acontextual(concept focused) Analysis of dimensions and measures of PE concept PE as enabling process (knowledge acquisition, coping skills); PE as outcome (participation in decision making, gaining control); other PE dimensions (self-care, capacity building, trust, motivation, sense of meaning, positive attitude) challenge of incorporating PE in health-care practice through motivational strategies to induce behavioral change Fumagalli et al.11 Acontextual(concept focused) Conceptual map for PE and 5 related concepts PE as a process, emergent state, and behavior (participation and involvement); PE as a combination of ability (enablement), motivation (engagement), and power (activation) PE mapped in close relationship with patient activation, enablement, engagement, involvement, and participation Castro et al.30 Acontextual(concept focused) Conceptual analysis of PE, patient participation, and patient-centeredness Antecedents (patient education, knowledge, control, participation); attributes (enablement, activation); consequences (self-efficacy, control, self-management); is a much broader concept than patient participation and patient-centeredness process model in health care to improve quality of care/ life: strategy of patient participation facilitates patient- centeredness, which leads to PE Theme 2: contextual embeddedness (9 articles): Topic 2.1: PE and national health-care system Boudioni et al.31 National health-care system Role of citizenship, culture, voluntary community organization in PE in Greece and England Patient/social/community participation; public involvement; patient informed choice and voice; patients’ rights; expert patients; patient ability to control own care PE shaped by stronger (weaker) citizenship and longer (shorter) tradition of voluntary action in England (Greece) Boudioni et al.32 Health-care policies Comparison of national policies, systems, structures of PE in Greece and England legislation-driven; patients’ rights (access to health care, quality of care, approval of treatment, respect, consent, confidentiality, information, informed choice, involvement in own health care, right of redress); patient-focused services policies emphasize: patient-centered services, public involvement and PE, in England; patient rights, responsibilities, and quality of services, in Greece Topic 2.2: PE and health-care governance Bodolica and Spraggon33 Health-care governance Divide between macrogovernance and microgovernance PE (patient choice, autonomy, medical literacy) as a component of micro-level governance (in the patient ephysician relationship) advocate the integration of macro and micro governance devices in health-care settings Tofan et al.34 Relational governance PE as governance mechanism in physiciane epatient relationship PE as distrust-based governance tool (patient autonomy, assertive control, info empowerment, choice, involvement, decision-making authority, eHealth, system distrust, use of Internet for info) conceptual framework integrating both trust- (doctor- focused) and distrust-based (patient-led) governance Topic 2.3: PE and information technology Risling et al.35 Electronic health Analysis of PE construct; relationship between PE and eHealth portal usage involvement in decisions; ability to find mistakes; preparedness; personal control; understanding of provider instructions; patient engagement (self-control, self-management, self-efficacy) and activation (skill, knowledge, confidence for self-care); use of eHealth tech huge variety in conceptual operationalization of PE; need to attain definitional consensus and standardized measure of PE to assess its association with the uptake of eHealth solutions (continued on next page) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e 1 2 4 1 1 7 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 T a b le 1 e (c o n ti n u ed ) R e fe re n c e C o n te x t E m p h a s is P E e le m e n ts a n d a s s o c ia te d d im e n s io n s F in d in g s a n d c o n tr ib u ti o n G ro e n e t a l. 3 6 In fo rm a ti o n te c h n o lo g y (I T ) C o n c e p tu a l c o m p o n e n ts o f P E ; c o n tr ib u ti o n o f IT to e m p o w e rm e n t o f c a n c e r s u rv iv o rs b e in g a u to n o m o u s a n d re s p e c te d ; h a v in g k n o w le d g e ; h a v in g p s y c h o s o c ia l a n d b e h a v io ra l s k il ls ; p e rc e iv in g s u p p o rt fr o m c o m m u n it y , fa m il y a n d fr ie n d s ; p e rc e iv in g o n e s e lf to b e u s e fu l IT (e d u c a ti o n a l, e le c tr o n ic , p a ti e n t- to -p a ti e n t, p o rt a l s e rv ic e s a n d m u lt ic o m p o n e n t) c o n tr ib u te s to P E b y e n h a n c in g a u to n o m y , k n o w le d g e a n d s k il ls C a lv il lo e t a l. 3 7 In fo rm a ti o n te c h n o lo g y A n a ly s is o f h o w v a ri o u s IT to o ls c o n tr ib u te to P E p a ti e n t e d u c a ti o n ; h e a lt h li te ra c y ; re m o te a c c e s s to h e a lt h s e rv ic e s ; a c c e s s c o n tr o l; s e lf -c a re ; p a ti e n t a s in fo s o u rc e ; d e c is io n m a k in g ; p ri v a c y a n d c o n fi d e n ti a li ty IT c o n tr ib u te s to P E (m o re p ro a c ti v e b e h a v io r) a n d a ll o w s c it iz e n s to a c t a s in fo p ro v id e rs T o p ic 2 .4 : P E a n d th e ra p e u ti c c o n ti n u u m te B o v e ld t e t a l. 3 8 C a n c e r p a in m a n a g e m e n t C o n c e p tu a l m o d e l a n d a n a ly s is o f P E in c o n tr o ll in g p a in s e lf -e ffi c a c y ; in c re a s in g a b il it ie s ; lo c u s o f c o n tr o l; a c ti v e c o p in g ; p a rt ic ip a ti o n in d e c is io n m a k in g ; re s o u rc e s (i n d u c e d b y c a re g iv e r, s k il ls ) fo c u s o n p a in tr e a tm e n t g iv e n b y c li n ic ia n , in v o lv e m e n t o f p a ti e n t a n d in te ra c ti o n o f b o th N a fr a d i e t a l. 3 9 M e d ic a ti o n a d h e re n c e E ff e c t o f 2 P E d im e n s io n s o n a d h e re n c e P E c o m p o n e n ts : in te rn a l h e a lt h lo c u s o f c o n tr o l (b e li e f o f b e in g in c o n tr o l o f o w n h e a lt h ) a n d s e lf -e ffi c a c y (d is e a s e m a n a g e m e n t, g e n e ra l) P E p ro m o te s a d h e re n c e ; n e e d fo r jo in t e m p o w e rm e n t (p a ti e n ts w h o s h a re c o n tr o l w it h d o c to rs ) P E , p a ti e n t e m p o w e rm e n t. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4118 participation,11 and patient-centeredness, indicating that PE is a broad concept that embraces multiple components.30 To improve clinical outcomes and quality of life, PE interventions should be integrated into medical practice through jointly deployed efforts of patients, physicians, and health-care or- ganizations to induce a long-lasting behavioral change.10,29 Nine ‘contextual embeddedness’ articles conceptualize PE in relation to ‘health-care system’, ‘health-care governance’, ‘information technology’ (IT), and ‘therapeutic continuum’. Citizenship, culture, and tradition of voluntary action play a critical role in shaping PE at the national level,31 giving rise to health-care reforms and PE policies and systems.32 PE is viewed as a distrust-based (patient-driven) governance attri- bute in the physicianepatient relationship34 and a microlevel governance component.33 Acknowledging the macroemicro divide in health-care governance, scholars promote the inte- gration of macrolevel (policy-making) and microlevel (doc- torepatient interaction) governance initiatives in medical settings. IT-based studies conclude that patients’ utilization of eHealth tools contributes to their empowerment35 through enhanced autonomy, knowledge, proactive behavior, and self-management,36 allowing people to act as seekers and suppliers of medical information.37 To achieve better out- comes in cancer pain management38 and medication adher- ence,39 ‘therapeutic continuum’ researchers advocate the principle of shared control and joint empowerment of pa- tients and physicians. PE is a complex multidimensional construct, which in- corporates many tightly intertwined elements (autonomy, self- efficacy, health literacy, information search/use, personal control, coping with illness), is relevant at multiple levels (micro, meso, macro), is analyzed from the standpoint of several stakeholders (patient, clinician, health-care system), is interpreted in many ways (process, state, behavior, feeling, intervention, outcome), emerges at different stages (anteced- ents, attributes, consequences), is seen as a intrapersonal disposition (patient power/control) or relational concept (power in clinicianepatient relationship), is linked to many disciplines (IT, governance, public policy/administration), oscillates along a continuum (low to high), and is influenced by various mod- erators (culture, citizenship, legislation, socio-economic con- ditions, professional goals, health status, education). PE-related issues in the UAE Our analysis suggests that most of the 17 UAE-based PE in- quiries represent the outcome of repeated efforts of the same researcher teams from local medical colleges/institutions (Table 2). The majority uses cross-sectional surveys and sta- tistical analyses, with only three articles making use of qual- itative methodologies to analyze specific case/interview data.40e42 While adult patients represent the common study subjects, one paper uses clinical professionals43 and four use college students,41,44 of which two focus on expatriate ado- lescents.45,46 Only one investigation has explicitly mentioned PE (as a perception of being knowledgeable47), with most ar- ticles referring to various PE dimensions. The two closely related PE constructs that were tackled in UAE studies are patient involvement48 and local community engagement.41 Eleven papers concentrate on patient capacity for self-care/ https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 Table 2 e Reviewed empirical studies related to PE in the UAE health-care settings. Reference Method Subjects Field PE/related constructs use Main findings and implications Theme 1: chronic disease care (11 studies): Topic 1.1: general inquiries Hashim et al.43 Cross-sectional survey, regression 38 nurses and physicians attending a workshop Chronic disease care n/a/patient self-management support services’ usage by clinicians clinicians’ non-use of proactive patient self-management tools; outreach programs and patient education needed to improve self-care and make chronic disease care systematic (not episodic) Sayiner et al.49 Cross-sectional survey, comparison 27 subjects (out of 1392 across 11 MENA states) Chronic obstructive pulmonary disease n/a/knowledge/being informed about disease, info seeking behavior from various sources feeling of being informed about respiratory condition is suboptimal (higher in the UAE than MENA); info obtained from doctors, TV and Internet; need for more patient education Topic 1.2: diabetes management Baynouna et al.47 Surveys, regressions 442 patients, 7 centers in Al Ain Hypertension, diabetes mellitus PE as perception of being knowledgeable; ability for self- management behavior assessment needed to design effective interventions to increase PE and adherence to healthy lifestyle behavior (via self-management) Hashim et al.50 Cross-sectional survey, regression 165 patients, 2 clinics in Al Ain Type 2 diabetes mellitus n/a/disease-related knowledge of patients, self-management patients’ knowledge about diabetes remained low over the 2001 e2014 period; education efforts need to focus on behavioral strategies to enable and encourage patients to adopt self-care Abduelkarem and Sackville51 Before-after study, 24 months 59 patients, 3 pharmacies in Sharjah Type 2 diabetes mellitus n/a/self-care and self- management (achieved via information reminders sent through pharmacists) poor disease knowledge, diet and exercise; info programs improve self-management; continuous long-term info/ education initiatives needed to induce behavioral change to adopt self-care Sulaiman et al.40 Qualitative, interviews 41 patients, Sharjah Diabetes n/a/patients’ disease-related knowledge knowledge varied; disease attributed to lifestyle, contextual and cultural factors; need for culturally-sensitive strategies to educate about illness Al-Maskari et al.73 Cross-sectional survey 575 patients, 2 hospitals in Al Ain Diabetes mellitus n/a/patient self-management of their chronic disease low patient awareness; poor knowledge/skills to self-manage the condition; awareness programs critical to improve coping, adherence and self-care Topic 1.3: diabetic complications Al-Kaabi et al.58 Cross-sectional study 409 patients, clinics in Al Ain Diabetes and dietary practice n/a/self-monitoring or self- management of disease poor self-monitoring and dietary practice; patient-tailored dietary counseling needed to empower patients to self-manage their chronic disease Al-Kaabi et al.52 Cross-sectional survey 390 patients, 6 clinics in Al Ain Diabetes and physical activity n/a/self-monitoring or self- management of disease low level of self-monitoring and physical activity; patient- tailored counseling needed to empower patients to self-manage their chronic disease Al-Kaabi et al.53 Experimental design, survey 221 illiterate patients in Al Ain Diabetic foot problems n/a/illiteracy of patients as predictor of poor foot-related self- care illiteracy induces poor knowledge of diabetes and its foot complications; education programs for illiterate patients needed to enhance self-care Sulaiman et al.59 Cross-sectional survey 347 patients, clinics in Sharjah Diabetes, depression, anxiety n/a/patient self-care (as correlate of depression) depressed diabetic patients have poor self-care and adherence; need for self-management initiatives to improve coping with chronic illness Theme 2: self-medication with drugs (3 studies) Shehnaz et al.45 Cross sectional survey 324 expatriate students, 4 schools Self-medication with drugs n/a/self-care attitude or autonomous health behavior high prevalence of self-medication as evidence of taking responsibility for own health but also risk of misuse; education programs needed for making the transition to self-care successful (continued on next page) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e 1 2 4 1 1 9 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 T a b le 2 e (c o n ti n u ed ) R e fe re n c e M e th o d S u b je c ts F ie ld P E /r e la te d c o n s tr u c ts u s e M a in fi n d in g s a n d im p li c a ti o n s S h e h n a z e t a l. 4 6 C ro s s -s e c ti o n a l s u rv e y 3 2 4 e x p a tr ia te a d o le s c e n t s tu d e n ts S e lf -m e d ic a ti o n w it h d ru g s n /a /d ru g k n o w le d g e ; in fo s e e k in g b e h a v io r fr o m v a ri o u s s o u rc e s lo w d ru g k n o w le d g e s c o re s , h ig h in c li n a ti o n fo r s e lf - m e d ic a ti o n ; p a re n ts , p h a rm a c is ts a n d m e d ia a s s o u rc e s o f in fo ; n e e d o f e d u c a ti o n c a m p a ig n s H a s a n e t a l. 4 2 Q u a li ta ti v e , c o n te n t a n a ly s is 3 0 s u b je c ts , p u rp o s iv e s a m p le S e lf -m e d ic a ti o n w it h d ru g s n /a /s e lf -m e d ic a ti o n a s a n a s p e c t o f s e lf -c a re a n d s o u rc e s o f d ru g s ’ in fo s e lf -m e d ic a ti o n is c o m m o n ; p h a rm a c is ts a n d fa m il y a s m a in s o u rc e o f in fo ; p h a rm a c is t p la y s a k e y ro le in p a ti e n t e d u c a ti o n a b o u t s e lf -c a re T h e m e 3 : n o n -t h e ra p e u ti c in te rv e n ti o n s (3 s tu d ie s ) L a u ra n c e e t a l. 4 1 C a s e s tu d y , p ro je c t 2 1 c o ll e g e s tu d e n ts , c o m m u n it y G e n e ti c d is e a s e s c re e n in g n /a /p a ti e n t a n d lo c a l c o m m u n it y e n g a g e m e n t (t o e d u c a te a n d s p re a d a w a re n e s s ) p ro je c t b e n e fi ts : la w fo r p re m a ri ta l s c re e n in g (d u e to h ig h c o n s a n g u in it y ); h ig h e r a w a re n e s s a n d li fe e x p e c ta n c y , lo w e r in c id e n c e o f d is e a s e a n d c o s t, b e tt e r h e a lt h ; n e e d to a d d re s s c u lt u ra l s e n s it iv it ie s a n d b u il d p a rt n e rs h ip s a c ro s s le v e ls M c L e a n e t a l. 4 8 In te rv ie w s , c li n ic a l s c e n a ri o s , re g re s s io n s 2 1 8 fe m a le E m ir a ti s (M u s li m s ) in A l A in c li n ic s O b s te tr ic s , s to m a c h , fa c e , c h il d s c e n a ri o s n /a /p a ti e n t in v o lv e m e n t (c o n s e n t) in s tu d e n ts ’ m e d ic a l e d u c a ti o n (f o r e x a m in a ti o n b y a s tu d e n t) R e fu s a l o f c ro s s -g e n d e r e x a m in a ti o n (o b s te tr ic s a n d s to m a c h ); n e e d to a c c o u n t fo r re li g io u s a n d c u lt u ra l is s u e s ; re m in d p a ti e n ts o f th e ir re li g io u s d u ty to c o n tr ib u te to w a rd s d o c to rs ’ tr a in in g A l- Y a te e m a n d R o s s it e r4 4 C ro s s -s e c ti o n a l s u rv e y 3 0 0 a d o le s c e n ts , 4 s c h o o ls in S h a rj a h N u tr it io n a n d d ie ta ry h a b it s n /a /k n o w le d g e o f h e a lt h y n u tr it io n a n d d ie t A d o le s c e n ts ’ la c k o f k n o w le d g e o f h e a lt h y e a ti n g a n d n u tr it io n ; to re d u c e ri s k o f o b e s it y , n e e d to d e s ig n m u lt if a c e te d e d u c a ti o n p ro g ra m s to in c re a s e k n o w le d g e o f h e a lt h y n u tr it io n P E , p a ti e n t e m p o w e rm e n t; U A E , th e U n it e d A ra b E m ir a te s ; n /a , n o t a p p li c a b le ; M E N A , M id d le E a s t a n d N o rt h A fr ic a . p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4120 self-management/self-monitoring, with patient’s disease- related knowledge and information-seeking behavior being examined on six and three occasions, respectively. We identified three themes on PE-related issues in the UAE: ‘chronic disease care’, ‘self-medication with drugs’, and ‘non- therapeutic interventions’. The contents of 11 first-theme ar- ticles point to three topics: ‘general inquiries’, ‘diabetes management’, and ‘diabetic complications’. Because public awareness about chronic conditions is suboptimal, the adop- tion of patient education programs and system interventions is recommended to depart from episodic to more systematic chronic disease care.43,49 The five ‘diabetes management’ studies suggest that educational initiatives in the UAE should be deployed continuously to induce sustainable behavioral change for patient espousal of self-care attitudes.50,51 Because diabetic complications are associated with poor dietary prac- tice, foot problems, low physical activity, depression, and anxiety, diabetes counseling should be tailored to patients’ needs to improve their coping with chronic illness.52,53 The second theme includes three studies on self- medication, treated as an aspect of patient self-care. Self- medication, which refers to situations when people admin- ister drugs to treat self-recognized symptoms/sicknesses without professional consultation, represents a means for empowering patients to take control of their health.54 Yet, the high inclination for self-medication by UAE adults and ado- lescents42 is accompanied by low levels of public knowledge and understanding of medicines and antibiotics.46 The prac- tice of self-medication raises ethical concerns and risks of drugs’ misuse, indicating that the population might not be well equipped to take higher responsibility for personal health. To make the transition to self-care a successful un- dertaking in the UAE, patient education programs are needed with the active participation of clinicians, pharmacists, par- ents, media, and other stakeholders.42,45 The three ‘non-therapeutic interventions’ articles address well-being issues that consider UAE's cultural/religious spec- ificities and cut across many stakeholder groups (adolescents, female Muslims, young Emirati couples). Given the prevalence of obesity in the UAE, researchers promote multifaceted educational interventions to enhance adolescents’ knowledge of healthy nutrition and encourage adopting beneficial dietary habits.44 Because consanguineous marriages are practiced by Muslim couples, local community engagement is critical for overcoming culturally induced resistance and spreading awareness about genetic disease screening to drive the implementation of a premarital screening law.41 In a study of patient involvement in students’ medical education, Emirati women with gynecological problems refused to submit to cross-gender examinations.48 While addressing cultural sen- sitivities is important, female patients need to be reminded of their social/religious duty to contribute towards doctors’ training in the country. Discussion By juxtaposing themes and topics from PE reviews and UAE- based studies, we delineate three opportunities for future in- quiry and policy intervention on PE in the UAE (Fig. 2). https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 Fig. 2 e Opportunities derived from juxtaposing themes/topics from Tables 1 and 2. PE, patient empowerment; UAE, the United Arab Emirates. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4 121 Opportunity (1)dconsolidate Most UAE studies connect PE with therapeutic continuum aspects of chronic disease care and self-medication. This provides opportunities for consolidation by conducting confirmatory research on larger samples across different Emirates to build a foundation for making generalizations. Unsurprisingly, most sampled studies relate to diabetes and its complications, because 20% of the total UAE population is battling this chronic condition.55 PE plays a critical role in the successful management of chronic illness, where the onus is on the patient to embrace the logic of active coping with dis- ease through disciplined self-monitoring.49,56,57 Yet, the diabetes-related knowledge, information-seeking behavior, and self-management attitudes of UAE patients remain weak.40,52,58,59 The UAE Government ambitions to reduce the percentage of diabetes by 2021, but the attainment of this goal depends on the effectiveness of PE programs to encourage optimal levels of patient self-care.60 In an analysis of health status in the UAE, cardiovascular diseases, injury, cancers, and respiratory disorders were identified as public health priorities to be addressed at the national level.61 Further assessments are needed on the contribution of PE strategies to the enhancement of health outcomes of patients with these chronic conditions through higher self-efficacy, pain management,38 and medication adherence.39 Extant studies on self-medication in the UAE focus on its risks and negative health implications due to the gap between patients’ state of ‘feeling’ and ‘being’ informed about drugs and medical principles.45,46 Authors noted that the prevalence of antibiotics’ self-medication in Abu Dhabi may reflect both the lack of punitive legislation for pharmacies dispensing drugs without prescription and the demographic aspect of the Emirate where its expatriate majority relies on home country sources of medicines.54 Regulatory in- terventions and educational programs aimed at reducing the incidence of drugs’ misuse will help refocusing researchers’ attention on the study of beneficial aspects of self-medication as a manifestation of autonomous health behavior. Opportunity (2)dexpand The sporadic and inconsistent use of PE suggests that the topic requires deeper exploration in the UAE. The propensity of decision-makers, clinical professionals, and patient advocates to discuss PE is the lowest in Arab countries, compared with the CanadaeAustralia cluster and even Latin American and Asian nations.28 Given the embryonic stage of development and scarcity of relevant UAE-based studies, many opportu- nities for expanding inquiry exist by combining insights from ‘conceptual clarification’ and ‘non-therapeutic interventions’ themes. To design viable interventions,41,48 we recommend delving deeper into PE and its contextual application by considering social, cultural, and religious characteristics of the UAE.62 The difficulty of achieving definitional consensus on PE10,30 is acknowledged because of the variability of national settings where the concept is used, inferring asymmetric levels of lit- eracy and access to information, concerns about digital divide and availability of Internet, and confidentiality issues.35 We call for contextualizing conceptual clarity efforts through a measurement scale that would allow operationalizing PE within the locally relevant value sets of the UAE. A hindrance for PE interventions may be the low literacy rates among older Emiratis and unskilled expatriate workers who lack formal education.15 Developing a reliable health-literacy screening instrument that would be culturally specific to target idio- syncrasies of the UAE socio-economic fabric represents a step forward.63 There is more scope for expanding research/prac- tice on the effectiveness of PE methods directed to the youth to inculcate a mentality of healthy nutrition,44 as obesity reduction among children represents another target of the 2021 UAE National Agenda.60 Although PE is gaining traction in global markets, some cultures might not be ready to embrace the trend toward increased autonomy and self-determination. In Muslim countries, patients may prefer to rely on professionals’ expert opinions or concede their individual decision-making power to their (male) family members.27 From the perspective of https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4122 Islam, although people enjoy the freedom of self-governance, the principles of beneficence and non-maleficence are given priority in medical decision-making, especially when patients are uninformed and possess limited understanding of their disease.64 To decide about a treatment while accounting for patients’ lack of competence, Islamic teachings reserve a central place to physicians, due to their professional and religious duty to do good and ward off harm.65 Faith-based, community-empowered, and family-centered participatory approaches may be appealing to the Muslim UAE majority, where Islamic principles and religious obligations form part of daily life. The role that community leaders, places of worship, and the family can play in spreading awareness about healthy lifestyles and empowering the UAE population to self-manage their health is worthy of further exploration.22 Opportunity (3)dinitiate Our analysis unveils a major decoupling between PE topics in review articles and those examined in the UAE context. The dearth of UAE-embedded inquiries on PE in relation to ‘na- tional health-care system’, ‘health-care governance’, and ‘IT’ provides opportunities for initiating research, reforms, and practice in these areas. Because PE is viewed as a fundamental pillar in the development of a sustainable health-care ecosystem,5 decision-makers ought to craft initiatives that would boost patient participation across levels of an inte- grated health-care system. The future makeup of medical practice and the implementation of a patient-centered approach to care depend on PE strategies that are formu- lated today.30e32 UAE legislators and practitioners should revisit institutional/clinical arrangements in health service provision to offer more room for residents to get involved in the design and delivery of care and play a heightened role in health-care governance.33 A greater sense of health ownership could be developed through educational policies and supporting infrastructures that would empower patients in medical encounters. People should engage in health-related initiatives in their community and voice their opinions regarding national priorities for pol- icymaking. Scholars could assess the effectiveness of patient- directed interventions in transforming people into value cre- ators and active participants in health-care markets.3 Federal public health frameworks should be revised periodically to secure compliance with international best practices and alignment with dominant health concerns. The diversity of the UAE population (age/gender distribution, educational/ economic backgrounds, social/cultural characteristics) poses challenges for the design of adequate public health reforms.61 PE education and intervention methods should be culturally sensitive, embedded in the nation's social fabric,66 and tailored to the needs of a specific group.41 Digital era technologies represent valuable platforms for drawing on citizens’ insights to transform public institutions and policymaking.35 Although health websites allow empowering Saudi Arabia patients,67,68 studies on how the adoption of electronic health systems drive PE in the UAE are lacking. Only one inquiry examined clinicians’ viewpoints about e-health development challenges in the UAE compared with other Arab states, but no associations were made with PE-related consequences.69 In technology-savvy nations, a tighter integration of IT into the medical sector may offer benefits in terms of health outcomes and general well- being.22 In the 2016 Global IT Report, UAE is ranked 26th worldwide and 1st in the Middle East on the Networked Readiness Index, unveiling a high level of government usage and social impact of IT.70 Considering the ever-expanding role of digital technologies and electronic portals in the UAE medical landscape, more research is needed on how e- health contributes to PE. UAE residents are becoming increasingly active on social media, rely on cellphone apps to make decisions, participate in online forums and support groups, and use media channels to access health-related data.22 Although social media usage for health information is an indicator of PE, this technology is associated with data inaccuracies, limited usability, misin- formation, and privacy/security issues.68 If information- seeking behavior is deployed as a tool for ‘empowering’ rather than ‘misleading’ patients,71 health-care organizations have to ensure the readability of data available online to improve people's health literacy. Clinicians should fulfill their moral obligation of facilitating PE by directing patients to health websites that are reliable and trustworthy.72 Studies on the role of social, educational and economic factors in the information-seeking behavior of empowered patients could be insightful for disseminating digitally the medical infor- mation that people can comprehend and act upon to solve their health-related concerns. Author statements Ethical approval None sought. Ethical approval was not required as this is a review article that relies on publicly available published work. Funding None declared. Competing interests None declared. r e f e r e n c e s 1. Moretta Tartaglione A, Cavacece Y, Cassia F, Russo G. The excellence of patient-centered healthcare: investigating the links between empowerment, co-creation and satisfaction. TQM J 2018;30(2):153e67. 2. Tofan G, Spraggon M, Bodolica V. Agency problems, ethical challenges and governance attributes in different models of physicianepatient interaction within the assisted reproduction setting. Publ Health 2013a;127(6):597e600. 3. Dent M, Pahor M. Patient involvement in Europe e a comparative framework. J Health Organisat Manag 2015;29(5):546e55. http://refhub.elsevier.com/S0033-3506(19)30023-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref2 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref3 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref3 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref3 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref3 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref3 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4 123 4. Snyder H, Engstrom J. The antecedents, forms and consequences of patient involvement: a narrative review of the literature. Int J Nurs Stud 2016;53:351e8. 5. Palumbo R, Cosimato S, Tommasetti A. Dream or reality? A recipe for sustainable and innovative health care ecosystems. TQM J 2017;29(6):847e62. 6. Bodolica V, Spraggon M, Tofan G. A structuration framework for bridging the macro-micro divide in healthcare governance. Health Expect 2016;19(4):790e804. 7. Scholl I, Zill JM, Harter M, Dirmaier J. An integrative model of patient-centeredness e a systematic review and concept analysis. PLoS One 2014;9(9). e107828. 8. Chang AK, Fritschi C, Kim MJ. Nurse-led empowerment strategies for hypertensive patients with metabolic syndrome. Contemp Nurse 2012;42:118e28. 9. Nafradi L, Nakamoto K, Csabai M, Papp-Zipernovszky O, Schulz PJ. An empirical test of the Health Empowerment Model: does patient empowerment moderate the effect of health literacy on health status? Patient Educ Counsel 2018;101:511e7. 10. Cerezo PG, Juve-Udina ME, Delgado-Hito P. Concepts and measures of patient empowerment: a comprehensive review. Rev Esc Enferm USP 2016;50(4):664e71. 11. Fumagalli LP, Radaelli G, Lettieri E, Bertele P, Masella C. Patient empowerment and its neighbors: clarifying the boundaries and their mutual relationships. Health Policy 2015;119:384e94. 12. Schulz PJ, Nakamoto K. Health literacy and patient empowerment in health communication: the importance of separating conjoined twins. Patient Educ Counsel 2013;90:4e11. 13. Barr PJ, Scholl I, Bravo P, Faber MJ, Elwyn G, McAllister M. Assessment of patient empowerment - a systematic review of measures. PLoS One 2015;10(5):1e24. 14. Spraggon M, Bodolica V. Managing organizations in the United Arab Emirates: dynamic characteristics and key economic developments. New York: Palgrave Macmillan; 2014. 15. Koornneef E, Robben P, Blair I. Progress and outcomes of health systems reform in the United Arab Emirates: a systematic review. BMC Health Serv Res 2017;17:672. 16. Bodolica V, Spraggon M, Shahid A. Strategic adaptation to environmental jolts: an analysis of corporate resilience in the property development sector in Dubai. Middle East J Manag 2018;5(1):1e20. 17. UAE Government. Good health and wellbeing. 2018. Retrieved from: https://government.ae/en/about-the-uae/leaving-no- one-behind/3goodhealthandwellbeing. 18. Federal Competitiveness and Statistics Authority. Government and private sector health services statistics. 2018. Retrieved from, http://fcsa.gov.ae/en-us. 19. Legatum Institute. The legatum prosperity Index 2017. Retrieved from www.properity.com. 20. World Economic Forum. The global competitiveness report 2017e2018. Retrieved from https://www.weforum.org. 21. Abuhejleh A, Dulaimi M, Ellahham S. Using Lean management to leverage innovation in healthcare projects: case study of a public hospital in the UAE. BMJ Innov 2016;2:22e32. 22. Gachiri W. A gradient analysis of economic development and chronic diseases. Case of diabetes and life expectancy in United Arab Emirates. J App Bus Econ 2017;19(1):35e43. 23. Weber AS, Turjoman R, Shaheen Y, Al Sayyed F, Hwang MJ, Malick F. Systematic thematic review of e-health research in the Gulf Cooperation Council (Arabian Gulf): Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and United Arab Emirates. J Telemed Telecare 2017;23(4):452e9. 24. Al Slamah T, Nicholl BI, Alslail FY, Melville CA. Self- management of type 2 diabetes in gulf cooperation council countries: a systematic review. PLoS One 2017;12(12). e0189160. 25. Bodolica V, Spraggon M. Mergers and acquisitions and executive compensation. Series: routledge studies in corporate governance. New York: Routledge; 2015. 26. Bodolica V, Spraggon M. Merger and acquisition transactions and executive compensation: a review of the empirical evidence. Acad Manag Ann 2009;3(1):109e81. 27. Malek MM, Abdul Rahman NN, Hasan MS, Abdullah LH. Islamic considerations on the application of patient's autonomy in end-of-life decision. J Relig Health 2018. https:// doi.org/10.1007/s10943-018-0575-5. 28. Bridges JFP, Anderson BO, Buzaid AC, Jazieh AR, Niessen LW, Blauvelt BM, Buchanan DR. Identifying important breast cancer control strategies in Asia, Latin America and the Middle East/North Africa. BMC Health Serv Res 2011;11:227. 29. Bravo P, Edwards A, Barr PJ, Scholl I, Elwyn G, McAllister M. Conceptualizing patient empowerment: a mixed methods study. BMC Health Serv Res 2015;15:252. 30. Castro EA, Regenmortel TV, Vanhaecht K, Sermeus W, Hecke AV. Patient empowerment, patient participation and patient-centeredness in hospital care: a concept analysis based on a literature review. Patient Educ Counsel 2016;99:1923e39. 31. Boudioni M, McLaren S, Lister G. The role of citizenship, culture and voluntary community organizations towards patient empowerment in England and Greece. Int J Caring Sci 2017a;10(1):303e12. 32. Boudioni M, McLaren S, Lister G. A critical analysis of national policies, systems, and structures of patient empowerment in England and Greece. Patient Prefer Adherence 2017b;11:1657e69. 33. Bodolica V, Spraggon M. Clinical governance infrastructures and relational mechanisms of control in healthcare organizations. J Health Manag 2014;16(2):183e98. 34. Tofan G, Bodolica V, Spraggon M. Governance mechanisms in the physician-patient relationship: a literature review and conceptual framework. Health Expect 2013b;16(1):14e31. 35. Risling T, Martinez J, Young J, Thorp-Froslie N. Evaluating patient empowerment in association with eHealth technology: scoping review. J Med Internet Res 2017;19(9):e329. 36. Groen WG, Kuijpers W, Oldenburg HS, Wouters MW, Aaronson NK, van Harten WH. Empowerment of cancer survivors through information technology: an integrative review. J Med Internet Res 2015;17(11). e270. 37. Calvillo J, Roman I, Roa LM. How technology is empowering patients? A literature review. Health Expect 2013;18(5):643e52. 38. te Boveldt N, Vernooij-Dassen M, Leppink I, Samwel H, Vissers K, Engels Y. Patient empowerment in cancer pain management: an integrative literature review. Psycho Oncol 2014;23:1203e11. 39. Nafradi L, Nakamoto K, Schulz PJ. Is patient empowerment the key to promote adherence? A systematic review of the relationship between self-efficacy, health locus of control and medication adherence. PLoS One 2017;12(10). e0186458. 40. Sulaiman N, Hamdan A, Al-Bedri DAM, Young D. Diabetes knowledge and attitudes towards prevention and health promotion: qualitative study in Sharjah, United Arab Emirates. Int J Food Saf Nutr Public Health 2009;2(1):78e88. 41. Laurance J, Henderson S, Howitt PJ, Matar M, Al Kuwari H, Edgman-Levitan S, Darzi A. Patient engagement: four case studies that highlight the potential for improved health outcomes and reduced costs. Health Aff 2014;33(9):1627e34. 42. Hasan S, Farghadani G, Al Haideri SK, Fathy MA. Pharmacist opportunities to improve public self-medicating practices in the UAE. Pharmacol Pharm 2016;7:459e71. 43. Hashim MJ, Prinsloo A, Mirza DM. Quality improvement tools for chronic disease care e more effective processes are less http://refhub.elsevier.com/S0033-3506(19)30023-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref7 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref7 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref7 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref7 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref8 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref8 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref8 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref8 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref9 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref10 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref10 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref10 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref10 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref11 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref11 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref11 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref11 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref11 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref12 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref12 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref12 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref12 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref15 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref15 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref15 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref16 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref16 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref16 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref16 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref16 https://government.ae/en/about-the-uae/leaving-no-one-behind/3goodhealthandwellbeing https://government.ae/en/about-the-uae/leaving-no-one-behind/3goodhealthandwellbeing http://fcsa.gov.ae/en-us http://www.properity.com https://www.weforum.org http://refhub.elsevier.com/S0033-3506(19)30023-X/sref21 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref21 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref21 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref21 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref21 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref22 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref22 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref22 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref22 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref24 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref24 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref24 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref25 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref25 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref25 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref26 https://doi.org/10.1007/s10943-018-0575-5 https://doi.org/10.1007/s10943-018-0575-5 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref28 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref28 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref28 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref28 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref28 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref29 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref29 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref29 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref30 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref31 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref31 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref31 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref31 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref31 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref33 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref33 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref33 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref33 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref34 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref34 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref34 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref34 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref36 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref36 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref36 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref36 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref37 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref37 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref37 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref37 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 1 4 e1 2 4124 likely to be implemented in developing countries. Int J Health Care Qual Assur 2013;26(1):14e9. 44. Al-Yateem N, Rossiter R. Nutritional knowledge and habits of adolescents aged 9 to 13 years in Sharjah, United Arab Emirates: a cross-sectional study. East Mediterr Health J 2017;23(8):551e8. 45. Shehnaz SI, Khan N, Sreedharan J, Issa KJ, Arifulla M. Self- medication and related health complaints among expatriate high school students in the United Arab Emirates. Pharm Pract 2013;11(4):211e8. 46. Shehnaz SI, Khan N, Sreedharan J, Arifulla M. Drug knowledge of expatriate adolescents in the United Arab Emirates and their attitudes towards self-medication. Int J Adolesc Med Health 2014;26(3):423e31. 47. Baynouna LM, Neglekerke NJD, Ali HE, ZeinAlDeen SM, Al Ameri TA. Audit of healthy lifestyle behaviors among patients with diabetes and hypertension attending ambulatory health care services in the United Arab Emirates. Glob Health Promot 2014;21(4):44e51. 48. McLean M, Al Ahbabi S, Al Ameri M, Al Mansoori M, Al Yahyaei F, Bernsen R. Muslim women and medical students in the clinical encounter. Med Educ 2010;44:306e15. 49. Sayiner A, Alzaabi A, Obeidatc NM, Nejjari C, Beji M, Uzaslan E. Attitudes and beliefs about COPD: data from the BREATHE study. Respir Med 2012;106(S2):S60e74. 50. Hashim MJ, Mustafa H, Ali H. Knowledge of diabetes among patients in the United Arab Emirates and trends since 2001: a study using the Michigan Diabetes Knowledge Test. East Mediterr Health J 2016;22(10):742e8. 51. Abdulelkarem AR, Sackville MA. Changes of some health indicators in patients with type 2 diabetes: a prospective study in three community pharmacies in Sharjah, United Arab Emirates. Libyan J Med 2009;4(1):31e6. 52. Al-Kaabi J, Al-Maskari F, Saadi H, Afandi B, Parkar H, Nagelkerke N. Physical activity and reported barriers to activity among type 2 diabetic patients in the United Arab Emirates. Rev Diabet Stud 2009;6(4):271e8. 53. Al-Kaabi JM, Al Maskari F, Cragg P, Afandi B, Souid A-K. Illiteracy and diabetic foot complications. Prim Care Diabetes 2015;9:465e72. 54. Abasaeed A, Vlcek J, Abuelkhair M, Kubena A. Self-medication with antibiotics by the community of Abu Dhabi Emirate, United Arab Emirates. J Infect Dev Ctries 2009;3(7):491e7. 55. Alhyas L, McKay A, Balasanthiran A, Majeed A. Quality of type 2 diabetes management in the states of the Co-operation Council for the Arab states of the Gulf: a systematic review. PLoS One 2011;6(8). e22186. 56. Lorig KR, Holman H. Self-management education: history, definition, outcomes, and mechanisms. Ann Behav Med 2003;26(1):1e7. 57. Zimbudzi E, Lo C, Misso M, Ranasinha S, Zoungas S. Effectiveness of management models for facilitating self- management and patient outcomes in adults with diabetes and chronic kidney disease. Syst Rev 2015;4:81. 58. Al-Kaabi J, Al-Maskari F, Saadi H, Afandi B, Parkar H, Nagelkerke N. Assessment of dietary practice among diabetic patients in the United Arab Emirates. Rev Diabet Stud 2008;5(2):110e5. 59. Sulaiman N, Hamdan A, Tamim H, Mahmood DA, Young D. The prevalence and correlates of depression and anxiety in a sample of diabetic patients in Sharjah, United Arab Emirates. BMC Fam Pract 2010;11:80. 60. UAE Government. Health and fitness. 2018. Retrieved from: https://government.ae/en/information-and-services/health- and-fitness. 61. Loney T, Aw T-C, Handysides DG, Ali R, Blair I, Grivna M, et al. An analysis of the health status of the United Arab Emirates: the ‘Big 4’ public health issues. Glob Health Action 2013;6:20100. 62. Bodolica V, Spraggon M. Life on heels and making deals: a narrative approach to female entrepreneurial experiences in the UAE. Manag Decis 2015;53(5):984e1004. 63. Nair SC, Satish KP, Sreedharan J, Ibrahim H. Assessing health literacy in the Eastern and Middle-Eastern cultures. BMC Public Health 2016;16:831. 64. Rathor MY, Rani MF, Shah AS, Leman WI, Akter SF, Omar AM. The principle of autonomy as related to personal decision-making concerning health and research from an ‘Islamic viewpoint’. J Islamic Med Assoc North America 2011;43(1):27e34. 65. Elbarazi I, Devlin NJ, Katsaiti M-S, Papadimitropoulos EA, Shah KK, Blair I. The effect of religion on the perception of health states among adults in the United Arab Emirates: a qualitative study. BMJ Open 2017;7. e016969. 66. Bodolica V, Spraggon M, Zaidi S. Boundary management strategies for governing family firms: a UAE-based case study. J Bus Res 2015;68(3):684e93. 67. Househ M, Alsughayar A, Al-Mutairi M. Empowering Saudi patients: how do Saudi health websites compare to international health websites? Stud Health Technol Inf 2013;183:296e301. 68. Househ M, Borycki E, Kushniruk A. Empowering patients through social media: the benefits and challenges. Health Inf J 2014;20(1):50e8. 69. Uluc NCI, Ferman M. A comparative analysis of user insights for e-health development challenges in Turkey, Kingdom of Saudi Arabia, Egypt and United Arab Emirates. J Manag Market Logis 2016;3(2):176e89. 70. World Economic Forum. The global information technology report. 2016. https://www.weforum.org. 71. Dahl S, Eagle L. Empowering or misleading? Online health information provision challenges. Market Intell Plann 2016;34(7):1000e20. 72. Spraggon M, Bodolica V. Trust, authentic pride and moral reasoning: a unified framework of relational governance and emotional self-regulation. Bus Ethics Eur Rev 2015;24(3):297e314. 73. Al-Maskari F, El-Sadig M, Al-Kaabi JM, Afandi B, Nagelkerke N, Yeatts KB. Knowledge, attitude and practices of diabetic patients in the United Arab Emirates. PLoS One 2013;8(1). e52857. http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref48 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref48 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref48 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref48 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref49 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref49 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref49 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref49 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref50 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref50 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref50 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref50 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref50 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref51 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref51 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref51 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref51 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref51 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref52 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref52 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref52 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref52 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref52 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref53 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref53 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref53 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref53 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref54 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref54 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref54 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref54 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref55 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref55 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref55 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref55 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref56 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref56 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref56 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref56 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref57 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref57 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref57 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref57 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref58 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref58 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref58 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref58 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref58 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref59 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref59 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref59 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref59 https://government.ae/en/information-and-services/health-and-fitness https://government.ae/en/information-and-services/health-and-fitness http://refhub.elsevier.com/S0033-3506(19)30023-X/sref61 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref61 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref61 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref61 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref62 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref62 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref62 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref62 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref63 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref63 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref63 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref64 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref65 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref65 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref65 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref65 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref66 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref66 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref66 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref66 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref67 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref67 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref67 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref67 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref67 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref68 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref68 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref68 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref68 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref69 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref69 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref69 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref69 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref69 https://www.weforum.org http://refhub.elsevier.com/S0033-3506(19)30023-X/sref71 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref71 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref71 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref71 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref72 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref72 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref72 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref72 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref72 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref73 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref73 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref73 http://refhub.elsevier.com/S0033-3506(19)30023-X/sref73 https://doi.org/10.1016/j.puhe.2019.01.017 https://doi.org/10.1016/j.puhe.2019.01.017 Toward patient-centered care and inclusive health-care governance: a review of patient empowerment in the UAE Introduction Methods Results PE-related review articles PE-related issues in the UAE Discussion Opportunity (1)—consolidate Opportunity (2)—expand Opportunity (3)—initiate Author statements Ethical approval Funding Competing interests References Economic-evaluations-of-public-health-implementation-interventi_2019_Public- ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3 Available online at w Public Health journal homepage: www.elsevier.com/puhe Review Paper Economic evaluations of public health implementation-interventions: a systematic review and guideline for practice P. Reeves a,b,c,*, K. Edmunds b, A. Searles b,a,c, J. Wiggers a,b,c,d a School of Medicine and Public Health, University of Newcastle, Callaghan, New South Wales 2308, Australia b Hunter Medical Research Institute, Newcastle, New South Wales 2305, Australia c Priority Research Centre for Health Behaviour, University of Newcastle, New South Wales 2308, Australia d Hunter New England Population Health, Wallsend, New South Wales 2287, Australia a r t i c l e i n f o Article history: Received 19 October 2018 Received in revised form 2 January 2019 Accepted 15 January 2019 Available online 13 March 2019 * Corresponding author. School of Medicine E-mail addresses: penny.reeves@hmri.or https://doi.org/10.1016/j.puhe.2019.01.012 0033-3506/© 2019 The Authors. Published by under the CC BY-NC-ND license (http://crea a b s t r a c t Objectives: Implementation interventions applied in public health are about using proven strategies to influence the uptake of evidence-based prevention and health promotion initiatives. The decision to invest in implementation has an opportunity cost, which can be overlooked. The purpose of this study was to assess the extent to which economic eval- uations have been applied to implementation interventions in public health. Study design: We conducted a systematic review of empirical studies examining the costs and consequences, cost-effectiveness or cost-benefit of strategies directed towards enhancing the implementation of public health interventions and policies in developed countries. Methods: The following databases were searched for English language publications reporting both effect measures and costs, from 1990 to current: MEDLINE, Embase, PsycINFO, CINAHL, EconLit, EPPI-Centre database of health promotion research, Cost-Effectiveness Analysis Registry, NHS Economic Evaluation Database, Informit and Scopus. Results: The search strategy returned 3229 records after duplicate removal, from which we included 14 economic evaluations. All the included evaluations were conducted and published after 2000. Twelve of the 14 evaluations were based on controlled trials and two reported hy- pothetical modelled scenarios. The methodologic rigour and compliance with reporting guidelines for economic evaluations was highly varied and not related to the publication date. Conclusions: Our findings offer the first insight into the application and methodologic rigour of economic evaluations of implementation strategies supporting public health policies and interventions. To usefully inform public health policy and investment decisions, there needs to be greater application of economic evaluation to understand the cost- effectiveness of alternative implementation efforts. This review highlights the great paucity and mixed quality of the evidence on this topic and offers guidance by way of a checklist to improve the quality and reporting of future evaluations. © 2019 The Authors. Published by Elsevier Ltd on behalf of The Royal Society for Public Health. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). and Public Health, University of Newcastle, Callaghan, New South Wales 2308, Australia. g.au (P. Reeves), john.wiggers@hnehealth.nsw.gov.au (J. Wiggers). Elsevier Ltd on behalf of The Royal Society for Public Health. This is an open access article tivecommons.org/licenses/by-nc-nd/4.0/). http://creativecommons.org/licenses/by-nc-nd/4.0/ mailto:penny.reeves@hmri.org.au mailto:john.wiggers@hnehealth.nsw.gov.au http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.012&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 http://creativecommons.org/licenses/by-nc-nd/4.0/ p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3102 Introduction database limits: MEDLINE, Embase, PsycINFO, CINAHL, Econ- In health care, the use of evidence-based, cost-effective practice does not spontaneously occur. Getting evidence into routine practice is the underlying objective of implementation science so an increasing application in health care is not surprising. Proven strategies and programmes have evolved to influence the uptake of evidence-based health care delivery methods.1 Examples include audit and feedback, education, financial incentives and regulation. The real-world effective- ness of practice change strategies has been examined in tar- geted public health settings such as schools, workplaces and healthcare facilities and has been shown to improve health outcomes.2 However, while effective, these strategies have resource requirements, which can be overlooked. In eco- nomics, the value of these resources is measured in terms of their opportunity cost. That is, the value forgone from an alternative use. Implementation interventions compete for the same resources as other interventions. Identifying, measuring and valuing the resources directed towards implementation, in conjunction with an assessment of effectiveness, informs the value for money derived from implementation investment and is important for overall health service efficiency. The equity impacts of implementa- tion investment should also be assessed and is equally overlooked.3 While prolific in clinical settings, public health initiatives are less likely to be accompanied by economic evaluations4e6 and some specific methodologic challenges have been iden- tified to explain why this could be the case, namely the diffi- culty in attributing effects, measuring and valuing outcomes, identifying intersectoral costs and incorporating equity con- siderations.6 In relation to implementation interventions specifically, it has been shown that the application of eco- nomic evaluation to implementation strategies targeting clinical practice change is not routine.7,8 When they are un- dertaken, the methodological quality is variable.8 The extent to which economic evaluations have been applied to public health implementation interventions is unknown. To that end, the purpose of this study was to conduct a systematic review of empirical studies examining the costs and conse- quences, cost-effectiveness or cost-benefit of strategies directed towards enhancing the implementation of public health interventions and policies. Our specific aims were to: (1) identify empirical economic evaluations of public health implementation interventions and policies published since 1990; (2) assess the quality of these evaluations; (3) synthesise the evidence of the study findings; and (4) develop recom- mendations to facilitate their conduct. Methods Search strategy and study selection The following databases were searched for English language publications from 1990 to November 2017, depending on the Lit, EPPI-Centre database of health promotion research, Cost- Effectiveness Analysis Registry (CEA), NHS Economic Evalua- tion Database (NHS EED), Informit and Scopus. This period was selected to reflect the prominence of implementation science as a discipline from the 1990s. A preliminary exami- nation of literature published before this date did not return any relevant publications. Table 1 details the MEDLINE search strategy and PICOS criteria. The study eligibility criteria were reported to be full eco- nomic evaluations of implementation interventions applied to a public or population health area in a community setting. Studies were considered to be full economic evaluations if they included information on both the costs and effects of implementation strategies.9,10 The definition of public health for this review was consistent with an accepted definition cited in the 2004 Wanless report.5 Included studies were randomised controlled trials, controlled trials and controlled before-and-after studies that involved economic evaluations of implementation strategies, including modelled evaluations. Implementation interventions were defined as any initiative designed to influence the uptake of public or population health interventions in community settings. Included strate- gies followed the taxonomy of professional, organisational, financial and regulatory strategies developed by the Effective Practice and Organisation of Care (EPOC) group11 and were consistent with the types of strategies adopted in the health technology assessment review by Grimshaw et al.12 All titles and abstracts of the search results were inde- pendently screened by two reviewers (P.R. and K.E. or A.S.) to identify potentially relevant studies. Full-text articles were sought for those titles/abstracts that appeared to meet the eligibility criteria. Based on the full-text reading, two re- viewers (P.R. and K.E.) independently assessed the identified studies for eligibility for inclusion. Furthermore, the reference lists of the included studies were also screened for potentially relevant articles. Disagreements between reviewers relating to study inclusion choices were resolved by consensus among all the reviewers (P.R., K.E. and A.S.). Data extraction and analysis Two reviewers (P.R. and K.E.) independently extracted data from the included studies on the following: public health issue addressed; intervention target population and study setting; country; year of study; year of reference case for the analysis and publication date; analysis perspective (societal, health- care sector, public finance or other); category of approach (trial-based, modelled or other); classification of outcomes (intermediate or final); and method of analysis (cost-conse- quence, cost-minimisation, cost-effectiveness, cost-utility, cost-benefit or other). Methodological quality The quality of the economic studies was assessed using the Drummond 10-point checklist.9 Reporting quality was assessed using the Consolidated Health Economic Evaluation https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 Table 1 e MEDLINE search strategy. (O) Outcomes 1. *Cost-Benefit Analysis/ 2. Economic evaluation.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 3. Cost effectiveness.mp. 4. (cost adj2 (effective* or utilit* or benefit* or minimi* or analy* or outcome*)).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 5. 1 or 2 or 3 or 4 (P) Population 6. Public Health/ 7. Public health.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 8. (("population level" or "population based" or population or "community level" or "community based" or communit*) adj2 (intervention* or implement*)).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 9. ("public health" adj2 (intervention* or implement*)).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 10. Health Promotion/ec [Economics] 11. 6 or 7 or 8 or 9 or 10 (I) Intervention 12. Implement*.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 13. Implement* strat*.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 14. Dissemination.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 15. (organi?ational adj change*).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 16. (system adj2 change*).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 17. Quality improvement.mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 18. ((adherent or complian*) adj3 policy).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 19. ((polic* or practic* or progra* or innovat*) adj5 (perform* or feedback or prompt* or reminder* or incentive* or penalt* or communic* or change manag* or train* or audit*)).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 20. (intervention$ or prevent* or polic* or program$).mp. 21. 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 22. 21 and 20 23. 5 and 11 and 22 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e 1 1 3 1 0 3 https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 (P) Population 24. (developing countr* or third world or underdeveloped countr* or under developed countr*).mp. [mp ¼ title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms] 25. Exp africa/or americas/or exp caribbean region/or exp central america/or latin america/or mexico/or exp south america/ 26. Exp europe, eastern/or exp transcaucasia/ 27. New Guinea/or asia/or exp asia, central/or asia, southeastern/or borneo/or cambodia/or east timor/or indonesia/or laos/or malaysia/or mekong valley/or myanmar/or philippines/or thailand/or vietnam/or asia, western/or bangladesh/or bhutan/or india/or middle east/or afghanistan/or iran/or iraq/or jordan/or lebanon/or oman/or saudi arabia/or syria/or turkey/or yemen/or nepal/or pakistan/or sri lanka/or far east/or china/or tibet/or exp korea/or mongolia/ 28. (Afghanistan or Africa or Albania or Algeria or Angola or Antigua or Argentina or Armenia or Azerbaijan or Bangladesh or Barbados or Barbuda or Belarus or Belize or Brazil or Bhutan or Bolivia or Bosnia or Botswana or Bulgaria or Burkina Faso or Burundi or Cambodia or Cameroon or Central African Republic or Chad or Chile or Colombia or Comoros or Congo or Costa Rica or Croatia or Cuba or Czech* or Congo or Djibouti or Dominica or Dominican or East Timor or Ecuador or Egypt or El Salvador or Equatorial Guinea or Eritrea or Estonia or Ethiopia or Fiji or Gabon or Gambia or Ghana or Grenada or Guatemala or Guinea-Bissau or Guyana or Haiti or Honduras or Hungary or India or Indonesia or Iran or Iraq or Ivory Coast or Jamaica or Jordan or Kazakhstan or Kenya or Kiribati or Kyrgyzstan or Laos or Latvia or Lebanon or Lesotho or Liberia or Libya or Lithuania or Madagascar or Malawi or Malaysia or Maldives or Mali or Marshall Islands or Mauritania or Mauritius or Mexico or Micronesia or Moldova or Mongolia or Montenegro or Morocco or Mozambique or Myanmar or Namibia or Nepal or New Guinea or Nicaragua or Niger or Nigeria or Korea or Oman or Pakistan or Palau or Panama or Papua New Guinea or Paraguay or Benin or China or Peru or Philippines or Poland or Cape Verde or Georgia or Kosovo or Macedonia or Yemen or Romania or Russia or Rwanda or Saint Kitts or Saint Vincent or Saint Lucia or Sao Tome Principe or Saudi Arabia or Senegal or Serbia or Seychelles or Sierra Leone or Slovak* or South Africa or Solomon Islands or Somalia or Sri Lanka or Sri-Lanka or Sudan or Suriname or Swaziland or Syria or Tajikistan or Tanzania or Thailand or Togo or Tonga or Trinidad or Tobago or Tunisia or Turkey or Turkmenistan or Uganda or Ukraine or Uruguay or Uzbekistan or Vanuatu or Venezuela or Vietnam or Samoa or Zambia or Zimbabwe).af. 29. 24 or 25 or 26 or 27 or 28 30. (developed countries or european union).af. 31. Europe/or andorra/or austria/or belgium/or exp france/or exp germany/or exp united kingdom/or greece/or ireland/or exp italy/or liechtenstein/or luxembourg/or monaco/or netherlands/or portugal/or exp "scandinavian and nordic countries"/or spain/or switzerland/or exp australia/or new zealand/ 32. North america/or exp canada/or exp united states/ 33. (united kingdom or england or scotland or wales or denmark or finland or iceland or norway or sweden).af. 34. (north america or canada or oecd or united states).af. 35. (europe or andorra or austria or belgium or france or germany or greece or ireland or italy or liechtenstein or luxembourg or monaco or netherlands or portugal or spain or switzerland or australia or new zealand).af. 36. 30 or 31 or 32 or 33 or 34 or 35 37. 23 and 36 38. 23 not 29 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e 1 1 3 1 0 4 https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3 105 Reporting Standards (CHEERS) checklist.10 These checklists are designed to improve the clarity of published economic evaluations and to increase reporting transparency. We assessed whether all relevant costs and effects were identi- fied, measured and valued appropriately. Guidelines for analysis and interpretation of results relate to the treatment of the timing of costs and effects; the reporting and interpre- tation of incremental analysis; treatment of uncertainty; and the context for the results. All the listed items were also aligned with the recommendations for conduct, methodo- logical practices and reporting of cost-effectiveness analyses developed by the JAMA panel on cost-effectiveness in health and medicine.13,14 Costs and effects Compared with economic evaluations of clinical healthcare interventions, the range of possible costs and effects asso- ciated with implementation interventions is argued to be wider.8,12 For this reason, we also assessed whether included studies reported (i) implementation strategy development costs; (ii) implementation strategy execution costs; and (iii) consequent changes in the cost of healthcare provision. We extracted data on the relevant economic outcomes: incremental cost-effectiveness ratios, net mon- etary benefit statistics and costebenefit ratios. Any stated decision rules such as willingness-to-pay thresholds were also recorded. Furthermore, in light of the relative paucity of economic evidence of public health interventions,4e6 we also extracted data on whether cost-effectiveness of the index policy or intervention was reported or discussed. We were also inter- ested to assess what measures of effect were included in the studies and how they were measured. If the evidence base for Fig. 1 e PRISMA the index policy is documented and includes a robust asso- ciation between the intermediate and final outcomes, it may be efficient to limit outcome measurement in implementation studies to intermediate outcome measures only such as change in healthcare practitioner behaviour, uptake or reach.15 Evidence synthesis To address our third aim, a quantitative evidence synthesis of the review results was planned, subject to assessment of heterogeneity. All data were entered in a Microsoft Excel (2013) database. Results Results of searches and screening The searches of the MEDLINE, Embase, PsycINFO, CINAHL, EconLit, EPPI-Centre database of health promotion research, CEA, NHS EED databases produced 5313 hits. The searches of the Informit and Scopus databases produced 244 hits (Fig. 1). From the title and abstract review, a total of 23 records met the eligibility criteria. On the basis of full-text assessments, 12 studies were excluded. The reasons for exclusion were the following: (1) the purpose of the study did not specifically relate to improving population or public health; and (2) studies only reported efficacy or effectiveness metrics with no corresponding data on costs. A further three studies were included from the search of the references of included arti- cles. In total, 14 studies19e32 were included in our review. No systematic reviews of relevant economic evaluations were identified. flowchart. https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3106 Study characteristics Table 2 summarises the characteristics of all included studies. Nine (64%) studies were conducted in the United States,19e22,24,26e28,32 two (14%) in Australia25,29 and the remaining three studies in Japan,23 Spain33 and the Netherlands.30 In the majority of studies (n ¼ 12), the popu- lation targeted by the interventions was the general community19e21,23e28,30,32,33 with only two studies targeting healthcare providers22 and schools,29 respectively. Nine (64%) of the studies focussed on strategies to increase cancer screening participation.19e21,23,26e28,32,33 The public health focus of the remaining studies included physical ac- tivity alone29 and in combination with healthy eating,30 alcohol-related crime25 and immunisation.22,24 The range of implementation strategies used in the studies, while highly varied, included predominantly demand-side interventions. Six studies included multiple intervention arms testing alternate strategies.19,20,27,30e33 Six studies reported results from single interventions designed with multiple strategies.19,26e31 The strategies that were most commonly combined included printed educational materials together with telephone education and counselling. The majority of the studies were trial-based eval- uations,19e23,25e31,33 two studies were modelled evaluations using decision analysis,24,32 one study used modelling to Table 2 e Summary of study characteristics. Characteristic Country USA Japan Australia Spain The Netherlands Population and setting Healthcare providers General community Other (schools) Index intervention or policya Mammography screeni Cervical cancer screeni Colorectal cancer scree Immunisation rates Other (alcohol and crim behaviours) Strategy typea HCP education Tailored risk printed m Public awareness raisin Financial incentives Telephone education a Multistrategy practice c Evaluation method Cost consequence Cost effectiveness Cost benefit Cost utility Other Unclear Perspective Healthcare provider Public finance Societal Other Not stated HCP, health care practitioner. a Studies are included in multiple categories. estimate costs28 and one study included a modelled trans- formation of intermediate to final outcomes.20 Methodological quality Evaluation design The method of economic evaluation was defined in all the studies but clearly justified in only five studies19,24,26,29,33 (Table 2). Most studies reported cost-effectiveness analy- ses,19e24,26e29,32,33 one study presented a costebenefit anal- ysis25 and one study conducted both cost-effectiveness and cost-utility analysis.30,31 An analysis perspective was clearly stated in 11 studies but justification was only provided in seven.19,20,24,26,28e31 The analysis time horizon was explicitly stated in six studies19,26,29e33 and could be inferred in one further study.28 An over view of all the included studies is provided in Table 3. Data included in the evaluations Table 4 summarises the quality of data and reporting per- taining to the identification, measurement and valuation of effectiveness of the intervention. The primary and secondary outcome measures were clearly specified in 11 of the 14 studies.19e25,27,29e31,33 The methods of measurement were reported in detail in only nine studies.19e23,25,27,29e31 Thirteen of the studies reported intermediate outcomes,19e24,26e33 Description n (%) 9 (64) 1 (7) 2 (14) 1 (7) 1 (7) 1 (7) 12 (86) 1 (7) ng 7 (50) ng 2 (14) ning 1 (7) 2 (7) e rates, physical activity and multihealth 3 (21) 3 (21) essaging 5 (35) g (media and community activities) 2 (14) 1 (7) nd counselling 9 (63) hange initiatives 6 (42) 0 (0) 12 (86) 1 (7) 1 (7) 0 (0) 0 (0) 4 (28) 1 (7) 6 (42) 0 (0) 3 (21) https://doi.org/10.1016/j.puhe.2019.01.012 https://doi.org/10.1016/j.puhe.2019.01.012 Table 3 e Overview of included studies. Author (year), Country Targeted public health issue Study setting and participants Type of economic evaluation Implementation strategy category Comparators Measure of effect (intermediate/final outcome) Cost categories included (perspective) Modelled or trial-based analysis Time horizon Health economic results Franzini, Boom et al. (2007),22 USA Immunisation coverage among preschoolers Paediatric and family medicine providers in Houston GP practices CEA HCP education Usual practice Self-reported provider behaviour Immunisation rates Direct expenses Time costs TBA NR Incremental cost ¼ $212 per provider Cost per 1 additional correct answer to immunisation guideline knowledge $437 Cost per 1 additional correct answer to 11 attitude and behaviour questions $474 Cost to increase immunisation rates by 1% $424-$555 Hirai, Ishikawa et al. (2016),23 Japan Colorectal cancer screening participation Japanese community setting, 46e66 years CEA Tailored risk printed messaging Generic reminder that FOBT was due Participation in screening within 5 m of receiving print reminder Individual assessment Overheads Reminder TBA NR Cost per participant 524 JPY Cost to increase screening participation by 1 3740 JPY for tailored matched and 4783 JPY for tailored unmatched and 2747 JPY for control Kim and Yoo,24 (2015), USA Influenza vaccination in the elderly US Medicare elderly CEA Public awareness raising via TV media No TV campaign Increase in vaccination rate # of additionally vaccinated Medicare elderly Cost of TV campaign Modelled NR Deterministic model ICER $17.79 USD per additionally vaccinated Medicare elderly Stochastic model ICER $23.54 USD per additionally vaccinated Medicare elderly Navarro, Shakeshaft et al. (2013),25 Australia Alcohol-related violent crime Communities in NSW population between 5000 and 20,000, >100 km
from a major urban
centre

CBA Public awareness
raising and
multistrategy practice
change initiatives

Control communities,
no intervention

Violent crimes, typically
alcohol related

Intervention costs:
additional policing
Media releases, mail-
outs and feedback
Intervention benefits:
reduced medical costs of
assault
Increased productivity

TBA NR Additional average total
cost of the intervention
$187,905 AUS
Estimated value of the
benefit $4,126,123 AUS
Benefit cost ratio 22:1
Net social benefit
$3,938,218 AUS

Schuster, Frick
et al. (2015),26

USA

Breast and
cervical cancer
screening
participation

Korean American
women attending
one of 23 ethnic
churches in
Baltimore

CEA Telephone education
and counselling

Wait list control, 1-
day training

Screening adherence Intervention
development costs
Training
Implementation

TBA 1 year
and
2 years

ICER ¼ $236 USD per
additional screening
excluding programme
development costs

Slater, Parks et al.
(2017),27 USA

Breast cancer
screening
participation

Female Medicare
beneficiaries aged
65-84 years

CEA Financial
incentives þ tailored
messaging

Control, no
intervention

Completion of a screening
mammogram

Phone centre costs
(logistics þtime) direct
mail materials
mailing list costs,
intervention
development costs

TBA 1 year ICER ¼ $1602.51 USD per
participant who
received a mammogram
for the direct mail
relative to control
ICER ¼ $1135.19 USD for
the direct
mail þ incentive arm
relative to control

(continued on next page)

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
0
1
e
1
1
3

1
0
7

https://doi.org/10.1016/j.puhe.2019.01.012

Table 3 e (continued)

Author (year),
Country

Targeted public
health issue

Study setting and
participants

Type of
economic
evaluation

Implementation
strategy category

Comparators Measure of effect
(intermediate/final

outcome)

Cost categories included
(perspective)

Modelled or
trial-based
analysis

Time
horizon

Health economic results

Stockdale, Keeler
et al. (2000),28

USA

Breast cancer
screening
participation

Women attending
churches in Los
Angeles

CEA Telephone education
and counselling

Usual practice, no
intervention

Number of additional
screenings
Modelled extrapolation
life years saved

Personnel, non-
personnel, indirect costs

TBA and
modelled

1 year Total cost per church
$609.98-$2926.95 USD
(cost models 1e3)
Cost per additional
screen $188.27-$903.42
USD (models 1e3)
Total cost per life year
saved by intervention
and subsequent
mammography $19,678-
$46,308 USD (models 1
e3)

Sutherland,
Reeves et al.
(2016),29

Australia

Physical activity
in adolescents

Secondary schools
and students in
NSW

CEA Multistrategy practice
change initiatives

Usual practice, no
intervention

Minutes of MVPA; percent
reduction in BMI; MET
hours gained per person
per day

Personnel costs
materials and printing

TBA 2 years Cost per additional
minute of MVPA per day
gained:
Based on the finding of a
difference in change of
7.0 (95% CI 2.68e11.36)
minutes per student per
day of MVPA for
students in the
intervention vs control
ICER $56 AUS [95% CI $35
e$147] per additional
minute of MVPA per day

Trapero-Bertran,
Acera Perez
et al. (2017),33

Spain

Cervical cancer
screening
participation

Women aged 30e70
years in selected
community area of
Spain

CEA Tailored risk printed
messaging þ telephone
education and
counselling

Active interventions:
personalised letter;
personalised
letterþ information
leaflet; personalised
letter, leaflet þ phone
call

Screening coverage over
42 months

Direct healthcare costs TBA 3e5 years IG2 is strongly
dominated (more
expensive and less
effective than IG1)
IG1 costs V 2.78 per 1%
increase in coverage
compared with an
opportunistic screening
IG3 costs V 13.73 per 1%
increase in coverage
more than an
opportunistic screening,
making IG1 more cost-
effective.
In the comparisons with
the next best
alternative, IG3 costs V
60.73 per 1% increase in
coverage more than IG1.
For women of all ages,
IG1 is the most cost-
effective alternative.

van Keulen,
Bosmans et al.
(2010),30,31 the
Netherlands

Multihealth
behaviours
(physical activity
and fruit and
vegetable
consumption)

Dutch general
practices

CUA Tailored risk printed
messaging þ telephone
education and
counselling

(1) Printed tailored
letters; (2)
motivational phone
calls; (3) combination
of 1 and 2; (4) no
intervention

Increase in the total
number of public health
guidelines met;
QALYs

Fixed costs
(development, training,
implementation,
overhead)
Variable costs
Total costs

TBA 73 weeks Combined intervention
and TMI both dominated
by control and TPC for
the difference in the
total number of
guidelines met
ICER for the TPC group
relative to control V160

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
0
1
e
1
1
3

1
0
8

https://doi.org/10.1016/j.puhe.2019.01.012

Wu, Fung et al.
(2004),32 USA

Breast cancer
screening
participation

Hypothetical cohort
based on 2000 Texan
female population
aged 50e79 years

CEA Telephone education
and counselling
Brief clinician-led
advice

Active controls:
Tailored telephone
counselling; clinic- or
physician-based
interventions
Usual care

Screening rates cancer
mortality rates

Screening cost
Diagnostic workup
Treatment costs
Terminal care costs

Modelled 5 years ICER more favourable
for tailored telephone
counselling
interventions (tailored
telephone counselling
vs. Clinic-based
intervention 88,520 vs.
172,960)

Andersen et al.
(2002),19 USA

Breast cancer
screening
participation

Woman aged 40e80
years living in a
rural community

CEA Mammography
promotion by
volunteers in rural
communities

Active controls: (1)
Community activities;
(2) Individual
counselling;
combined 1 and 2; (3)
no intervention

Use of mammography in
the general community
(percent of women)
Use of mammography by
underusers

Direct costs
Development and
training of volunteers
Materials
Personnel costs
Indirect costs

TBA
Modelled
extrapolation

31
months

General community
population ICER range
infinited$437 USD per
additional mammogram
ICCAs ICER range
infinited$608 per
additional mammogram
Cost per year of life
saved associated with
mammography
promotion was
approximately $56,000
USD per year of life
saved

Crane et al.
(2000),21 USA

Breast cancer
screening
participation

Woman aged 50
years or older living
in Colorado

Cost
analysis þ
rudimentary
CEA

Multiple outcall
strategy to promote
screening
mammography

(1)Single telephone
call; (2) single call
preceded by mail-out;
(3) no intervention

Adherence to screening
guidelines

Cost for delivery of calls
Printing and postage
Personnel costs
Overheads

TBA NR The cost per participant
who changed their
screening behaviour
was estimated to be
$288, $390 and $154 USD
for each of the single
outcall, advance card 1
single outcall, and
multiple outcall
interventions,
respectively

Costanza et al.
(2000),20 USA

Breast cancer
screening
participation

Female underusers
of mammography
aged 50e80 years
who were members
of an HMO and
living in
Massachusetts

CEA Telephone education
and counselling þ HCP
education

(1) Mailed screening
reminders only; (2)
mailed screening
reminder þtelephone
counselling; (3)
mailed reminders to
women þ physician-
based education

Self-report of
mammography use

Start-up costs
Ongoing costs
Total direct costs
Total indirect costs

TBA 3 years Reminder system cost $5
USD per woman in the
practice for all three
arms
BSTC was 3� more
expensive than routine
care and MD-ED, 5�
more expensive
Cost per additional user
was $726 USD for BSTC
and $2179 for MD-ED
Cost of BSTC $181,418
USD to $362,835 per life
saved for women aged
50e59 years

BMI, body mass index; BSTC, barrier-specific telephone counseling; CI, confidence interval; CEA, cost-effectiveness analysis, CUA, cost-utility analysis; FOBT, faecal occult blood test; GP, general

practitioner; HCP, health care practitioner; MVPA, moderate to vigorous physical activity; QALYs, quality adjusted life years; NR, not reported; NSW, New South Wales; TBA, trial-based analysis; TMI,

tailored messaging intervention; TPC, telephone counselling.

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
0
1
e
1
1
3

1
0
9

https://doi.org/10.1016/j.puhe.2019.01.012

Table 4 e Summary of data collection for evaluating
effectiveness: identification, measurement and valuation
of outcomes.

Reporting quality variable % done (n)

Identification of effects

Primary and secondary outcomes clearly outlined 79% (11/14)

Sufficiently detailed reporting of effects 64% (9/14)

Intermediate effects and final outcomes reported 29% (4/14)

Measurement of effects

Description of methods for measuring effects 79% (11/14)

Sufficiently detailed reporting of methods 50% (7/14)

Measurement in appropriate units 86% (12/14)

Valuation of effects

Description of methods to value effects 100% (2/2)

Credible methods to value effects 50% (1/2)

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3110
three studies reported both intermediate and final out-

comes19,30e32 and one study reported only final outcomes.25

The outcomes were valued in two studies25,30,31 (one cost

benefit analysis (CBA) and one cost-utility analysis), but the

approach to the valuation was clearly specified in only one

study.30,31

A summary of the collection of data for estimating the

costs of implementation is presented in Table 5. Only two

studies reported the costs of developing the implementation

strategies.26,30,31 In these studies, resource use identification

was reported consistently with the recommended guidelines,

but resource use measurement was not reported. Changes in

the cost of any associated healthcare provision was accounted

for in only one study32 and assumed to be of no difference

between study arms in one further study.30,31

All 14 studies reported the costs of executing these stra-

tegies; however, sufficient detail was reported in only five

(36%) studies.19,22,29e32 Only six studies reported the quantities

of resources used for strategy execution purposes separately

from their unit costs,19,21,23,28e31 and only four studies pro-

vided adequate description of methods for estimating

resource quantities.19,22,29e31

The reference year or price base for the analysis was un-

specified in four studies.20,26e28,30,31 Currency data were made

explicit in 11 of the studies and inferred from descriptions in

three studies. In this review, 13 of the studies made reference

to the efficacy of the index policy targeted by the imple-

mentation strategies, and a further six studies referenced the

cost-effectiveness of the policy.

Analysis and interpretation of evaluation results

Three studies appropriately25,26,32 reported discounting of

costs and effects. A further five studies19,20,29e31,33 with anal-

ysis periods greater than 12 months did not report discount-

ing. Incremental analyses was performed in 12 of the 14

studies and omitted in two studies.21,23

None of the studies explicitly associated the efficiency of

implementation strategies and the cost-effectiveness of the

index policy being promoted. While uncertainty analysis was

technically performed in 10 of the 14 studies,19,21,22,24,25,28e33

this involved rudimentary, deterministic univariate analyses

in nine of the 10.
Synthesis of review results
Compared with effectiveness studies, it is argued that the

generalisability of cost-effectiveness results is limited by the

wide range of factors necessarily involved in economic anal-

ysis studies, and the recent commentary on the conduct of

systematic reviews of economic evaluations has highlighted

specific challenges in relation to undertaking meta-analyses

of the results.16e18 In this review, the different populations,

jurisdictions, study settings, time horizons, measures of effect

and economic outcome measures stemming from the delib-

erate choice to adopt a broad definition of public health and no

limitation on the type of implementation intervention or type

of economic evaluation meant that the results from the

included studies could not be meaningfully combined in the

form of a meta-analysis or meta-regression. We, therefore,

summarised the results in the form of a narrative synthesis

only. A descriptive summary of the economic evaluation re-

sults is provided in Table 3.

The majority of the studies (nine) concluded that the re-

sults are evidence that the implementation interventions

are cost-effective or have a positive costebenefit

ratio.19e21,24e27,29,32 Three studies state that the in-

terventions were not deemed cost-effective,22,23,30 and the

remaining two make no claim regarding cost-effective-

ness.28,33 However, of the eight studies using cost-

effectiveness analysis and making a cost-effectiveness

claim, only five provided a context for the results in terms of

either explicit willingness-to-pay thresholds for the selected

outcome measures or benchmark ICERS.24,26,27,29,32
Discussion

To assess the scope and quality of evidence of economic

evaluations applied to implementation investment in a public

health setting, we conducted a systematic review of English

language literature from January 1990 to September 2017. Our

first aim was to identify empirical studies affecting public

health policies and practices over this period. The absence of

any relevant systematic reviews on this topic is in contrast to

the estimated 35e50 reviews of economic evaluations esti-

mated to be published each year.34 On the grounds that sys-

tematic reviews are the vehicle for providing information on

what is known and where knowledge gaps exist, it can be

concluded that there is a great deal yet to be learned on this

topic. From this one review, it is not possible to make any

broad conclusions regarding the value for money offered by

implementation interventions. Furthermore, the dominance

of trial-based analyses with restricted time horizons lends

weight to the argument made earlier by Hoomans and

Severens15 that greater use can be made of economic model-

ling to increase the efficiency of economic evaluations.

As referenced in the introduction, research has been pre-

viously conducted to better understand the paucity of eco-

nomic evaluations conducted in public health.5,6 Yet there are

further reasons that may explain the absence of good quality

economic evaluations in public health implementation in-

terventions. First, it may be challenging to sufficiently differ-

entiate the costs of implementation from the cost of the

https://doi.org/10.1016/j.puhe.2019.01.012

Table 5 e Summary of data collection for evaluating costs: identification, measurement and valuation of resource use.

Reporting quality variable Implementation intervention

development costs,

% done (n)

Implementation intervention

execution costs, % done (n)

Identification of costs

Reporting of costs 14% (2/14) 100% (14/14)

Sufficiently detailed reporting of costs 0% (0/14) 36% (5/14)

Inclusion of fixed costs 7% (1/14) 57% (8/14)

Inclusion of variable costs 14% (2/14) 93% (13/14)

Inclusion of opportunity costs 7% (1/14) 71% (10/14)

Omission of any costs justified 0% (0/14) 29% (4/14)

Measurement of costs

Separate reporting of resource quantities from unit costs 0% (0/14) 43% (6/14)

Description of methods for estimating resource quantities 0% (0/14) 29% (4/14)

Sufficiently detailed description of methods for estimation 0% (0/14) 29% (4/14)

Measurement in appropriate units 0% (0/14) 86% (12/14)

Valuation of costs

Recording of currency data 14% (2/14) 79% (11/14)

Recording of price data 14% (2/14) 50% (7/14)

Description of methods to value unit costs 0% (0/14) 7% (1/14)

Details of any adjustments for inflation or currency conversions 0% (0/14) 21% (3/14)

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3 111
intervention. Second, implementation is typically the purview

of local health services where cost-impact assessments, if

conducted at all, are undertaken internally and not neces-

sarily reported or published in the research community.

Further aims of this research were to assess the method-

ological quality of the evaluations and consistency of report-

ing with recommended guidelines. The clear majority (13 of 14

studies) were cost-effectiveness/utility analyses, where the

outcomes of the implementation strategies were represented

by a single metric. This result is important for the reason that

in the public health context, the outcomes and impacts of

implementation investment can be many and varied,

prompting some commentators to suggest that cost conse-

quence or multicriteria decision analysis may be appropriate

alternate methods, allowing decision makers (or analysts) to

weight the importance of different effects.6 Evaluation quality

and reporting was judged according to consistency with

guidelines.9,10,13,14 No single study met every reporting

criteria, and methodologic rigour and consistency with

reporting guidelines was highly variable. Compliance with the

reporting criteria for measures of effect was higher than for

measurement of resource use. This is likely correlated with

the high proportion of trial-based analyses, where interven-

tion efficacy is the primary trial outcome.

Regarding the data included in the economic evaluations,

we observed that only one study reported the costs of devel-

oping strategies and only one study considered downstream

healthcare costs. It was also observed that in one of the three

studies involving healthcare practitioner education, the eco-

nomic evaluation wrongly omitted to consider the opportunity

cost of healthcare professional time spent on attending the

educational meetings. It is appropriate that heathcare policy-

makers are cognizant of the balance among development costs,

implementation or execution costs and the costs of any asso-

ciated changes in healthcare provision. The focus of the studies

included in our review was clearly dominated by intervention

execution costs. As noted by other commentators, the data

included in these studies may be insufficient for a full economic

evaluation;9,35 however, economic modelling can be used to
transform or supplement relevant costs.9,36 In the studies

included in our review, the evaluations were either wholly trial-

based with no further attempt to transform the outcomes or

costs beyond the life of the study or wholly modelled.

Our review has a number of limitations and strengths

which should be noted. First, there was no prespecified pro-

tocol published for this review; however, the reviewers fol-

lowed the PICOS criteria for inclusion, and the search strategy

has been reported for transparency. Second, it is possible that

some economic evaluations will have been missed. Invest-

ment in implementation of public health initiatives is typi-

cally the responsibility of local health authorities, some of

which do not have the research capacity or capability to

conduct and publish economic evaluations. Furthermore,

implementation may occur in the form of quality improve-

ment initiatives, which often escape the lens of evaluation.

Our findings offer the first insight into the application and

methodologic rigour of economic evaluations of implementa-

tion strategies supporting public health policies and in-

terventions. While restricted in the setting of public health, our

findings are in accordance with similar commentary on the

economics of implementation in clinical health care.8,15,37,38

By highlighting the paucity of evidence and the deficiencies

in the application of economic evaluation methods, the

quality of future evaluations can be improved and decision-

making with respect to investment in implementation can

be better informed. Public health policies and programmes,

especially in the area of health prevention, have great po-

tential to mitigate escalating healthcare costs. Ensuring not

just their effective implementation but cost-effective imple-

mentation is critical if governments are to succeed in realising

improved population health outcomes and contained per

capita healthcare expenditure. To that end, the short checklist

below, developed from the findings of this review, provides

additional pragmatic guidance for conducting and reporting

economic evaluations of implementation interventions in

public health. Our checklist (Table 6) is intended to be used in

conjunction with the Drummond9 and CHEERS checklists10

and adds supplementary questions for consideration.

https://doi.org/10.1016/j.puhe.2019.01.012

Table 6 e Checklist to guide the conduct and reporting of economic evaluations of implementation interventions in public
health.

Item Drummond

checklist itema
CHEERS

checklist itemb
Recommendation relevant to implementation

interventions in public health

Background and objectives N/A 3 Provide an explicit statement of the economic evidence of the

index policy or programme targeted by the implementation

intervention

Consider the policy perspective for the interpretation of the

evaluation result. Will an incremental cost per unit change in

outcome inform decision-making or would an alternate

analysis type (e.g., cost consequence) be more appropriate?

Target population and

subgroups

10.4 4 Include study participant characteristics relevant for informing

equity considerations, e.g., the health status and ‘gap’ by sex,

age and socio-economic status

Study perspective 4.2 6 Where possible adopt a societal perspective to reflect the broad

nature of the expected costs and benefits

Choice of health outcomes

and measurement of

effectiveness

3, 4 10, 11 Depending on the known evidence of cost-effectiveness of the

index policy/programme, consider measuring intermediate

outcomes relevant to implementation and use economic

modelling to extrapolate to the final health outcomes

Estimating resource use and

costs

4, 5 13 Consider identifying, measuring and valuing resource use

associated with both development and execution of

implementation interventions

Characterising uncertainty

and heterogeneity

9 21, 22 Consider variation in parameters relevant to implementation,

e.g., reach and uptake

Distributional impacts 10.4 N/A If applicable (refer item no.4), make an explicit effort to include

equity considerations (e.g., undertake distributional cost-

effectiveness analysis)

a Corresponds directly to Drummond checklist item.9

b Corresponds directly to CHEERS checklist item.10

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3112
Author statements

Acknowledgements

The authors would like to thank Stephen Mears and Debbie

Booth for their assistance with the search strategy.

Ethical approval

None sought.

Funding

None.

Competing interests

The authors declare no conflicts of interest.

Authors’ contributions

PR, AS and JW conceived the review idea. PR, KE and AS per-

formed the title, abstract and full text review. PR and KE per-

formed the data extraction. PR led drafting the manuscript. All

authors read and approved the final manuscript.

r e f e r e n c e s

1. Brownson RC, Colditz GA, Proctor EK. Dissemination and
implementation research in health: translating science to practice.
Oxford Scholarship Online; 2012.
2. Lobb R, Colditz GA. Implementation science and its
application to population health. Annu Rev Public Health
2013;34:235e51.

3. Lal A, Moodie M, Peeters A, Carter R. Inclusion of equity in
economic analyses of public health policies: systematic
review and future directions. Aust N Z J Public Health
2018;42:207e13.

4. Drummond M, Weatherly H, Claxton K, Cookson R,
Ferguson B, Godfrey C, et al. Assessing the challenges of applying
standard methods of economic evaluation to public health
interventions. York: Public Health Research Consortium; 2007.

5. Wanless D. Securing good health for the whole population: final report.
2004. http://webarchive.nationalarchives.gov.uk/þ/http:/www.
hm-treasury.gov.uk/media/D/3/Wanless04_summary .

6. Weatherly H, Drummond M, Claxton K, Cookson R,
Ferguson B, Godfrey C, et al. Methods for assessing the cost-
effectiveness of public health interventions: key challenges
and recommendations. Health Policy 2009;93:85e92.

7. Hoomans T, Ament AJHA, Evers SMAA, Severens JL.
Implementing guidelines into clinical practice: what is the
value? J Eval Clin Pract 2011;17:606e14.

8. Hoomans T, Evers SM, Ament AJ, Hubben MW, van der
Weijden T, Grimshaw JM, et al. The methodological quality of
economic evaluations of guideline implementation into
clinical practice: a systematic review of empiric studies. Value
Health : J Inter Soc Pharmacoecon Outcomes Res 2007;10:305e16.

9. Drummond MF, O’Brien BJ, Stoddart GL, Torrance GW.
Methods for the economic evaluation of health care programmes. 2
ed. Oxford: Oxford Univeristy Press; 1997.

10. Husereau D, Drummond M, Petrou S, Carswell C, Moher D,
Greenberg D, et al. Consolidated health economic evaluation
reporting Standards (CHEERS) statement. Int J Technol Assess
Health Care 2013;29:117e22.

11. Effective Practice and Organization of Care Group (EPOC).
Data collection checklist. 2002. Available from: http://epoc.

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref1

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref2

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref3

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref4

http://webarchive.nationalarchives.gov.uk/+/http:/www.hm-treasury.gov.uk/media/D/3/Wanless04_summary

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref6

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref7

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref8

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref9

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref10

http://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/datacollectionchecklist

https://doi.org/10.1016/j.puhe.2019.01.012

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 1 e1 1 3 113
cochrane.org/sites/epoc.cochrane.org/files/public/uploads/
datacollectionchecklist . [Accessed 23 October 2017].

12. Grimshaw JM, Thomas RE, MacLennan G, Fraser C,
Ramsay CR, Vale L, et al. Effectiveness and efficiency of
guideline dissemination and implementation strategies. In:
Health technology assessment (Winchester, England), vol. 8; 2004.
p. 1e72. iii-iv.

13. Sanders GD, Neumann PJ, Basu A, Brock DW, Feeny D,
Krahn M, et al. Recommendations for conduct,
methodological practices, and reporting of cost-effectiveness
analyses: second panel on cost-effectiveness in health and
medicine. Jama 2016;316:1093e103.

14. Siegel JE, Weinstein MC, Russell LB, Gold MR.
Recommendations for reporting cost-effectiveness analyses.
Panel on cost-effectiveness in health and medicine. Jama
1996;276:1339e41.

15. Hoomans T, Severens JL. Economic evaluation of implementation
strategies in health care. Implement Sci 2014;9:168.

16. Luhnen M, Prediger B, Neugebauer EAM, Mathes T.
Systematic reviews of health economic evaluations: a
protocol for a systematic review of characteristics and
methods applied. Syst Rev 2017;6:238.

17. Anderson R. Systematic reviews of economic evaluations:
utility or futility? Health Econ 2010;19:350e64.

18. Gonzalez-Perez JG. Developing a scoring system to quality
assess economic evaluations. Eur J Health Econ : HEPAC : Health
Econ Prev Care 2002;3:131e6.

19. Andersen MR, Hager M, Su C, Urban N. Analysis of the cost-
effectiveness of mammography promotion by volunteers in
rural communities. Health Educ Behav: Off Pub Soc Pub Health
Edu 2002;29:755e70.

20. Costanza ME, Stoddard AM, Luckmann R, White MJ, Spitz
Avrunin J, Clemow L. Promoting mammography: results of a
randomized trial of telephone counseling and a medical
practice intervention. Am J Prev Med 2000;19:39e46.

21. Crane LA, Leakey TA, Ehrsam G, Rimer BK, Warnecke RB.
Effectiveness and cost-effectiveness of multiple outcalls to
promote mammography among low-income women. Cancer
Epidemiol Biomarkers Prev: Pub Am Assoc Cancer Res Cosponsored
Am Soc Prev Oncol 2000;9:923e31.

22. Franzini L, Boom J, Nelson C. Cost-effectiveness analysis of a
practice-based immunization education intervention. Ambul
Pediatr 2007;7:167e75.

23. Hirai K, Ishikawa Y, Fukuyoshi J, Yonekura A, Harada K,
Shibuya D, et al. Tailored message interventions versus
typical messages for increasing participation in colorectal
cancer screening among a non-adherent population: a
randomized controlled trial. BMC Public Health 2016;16:431.

24. Kim M, Yoo B-K. Cost-effectiveness analysis of a television
campaign to promote seasonal influenza vaccination among
the elderly. Value Health 2015;18:622e30.

25. Navarro HJ, Shakeshaft A, Doran CM, Petrie DJ. Does
increasing community and liquor licensees’ awareness, police
activity, and feedback reduce alcohol-related violent crime? A
benefit-cost analysis. Int J Environ Res Publ Health
2013;10:5490e506.
26. Schuster AL, Frick KD, Huh B-Y, Kim KB, Kim M, Han H-R.
Economic evaluation of a community health worker-led
health literacy intervention to promote cancer screening
among Korean American women. J Health Care Poor
Underserved 2015;26:431e40.

27. Slater JS, Parks MJ, Malone ME, Henly GA, Nelson CL. Coupling
financial incentives with direct mail in population-based
practice: a randomized trial of mammography promotion.
Health Educ Behav 2017;44:165e74.

28. Stockdale SE, Keeler E, Duan N, Derose KP, Fox SA. Costs
and cost-effectiveness of a church-based intervention to
promote mammography screening. Health Serv Res
2000;35:1037e57.

29. Sutherland R, Reeves P, Campbell E, Lubans DR, Morgan PJ,
Nathan N, et al. Cost effectiveness of a multi-component
school-based physical activity intervention targeting
adolescents: the ‘Physical Activity 4 Everyone’ cluster
randomized trial. Int J Behav Nutr Phys Activ 2016;13:94.

30. van Keulen HM, Bosmans JE, van Tulder MW, Severens JL, de
Vries H, Brug J, et al. Cost-effectiveness of tailored print
communication, telephone motivational interviewing, and a
combination of the two: results of an economic evaluation
alongside the Vitalum randomized controlled trial. Int J Behav
Nutr Phys Activ 2010;7. no pagination.

31. van Keulen HM, Bosmans JE, van Tulder MW, Severens JL, de
Vries H, Brug J, et al. Cost-effectiveness of tailored print
communication, telephone motivational interviewing, and a
combination of the two: results of an economic evaluation
alongside the Vitalum randomized controlled trial”:
Correction. Int J Behav Nutr Phys Activ 2011;8:8. ArtID 4. 2011.

32. Wu JH, Fung MC, Chan W, Lairson DR. Cost-effectiveness
analysis of interventions to enhance mammography
compliance using computer modeling (CAN*TROL). Value
Health 2004;7:175e85.

33. Trapero-Bertran M, Acera Perez A, de Sanjose S, Manresa
Dominguez JM, Rodriguez Capriles D, Rodriguez Martinez A,
et al. Cost-effectiveness of strategies to increase screening
coverage for cervical cancer in Spain: the CRIVERVA study.
BMC Public Health 2017;17:194.

34. van Mastrigt GA, Hiligsmann M, Arts JJ, Broos PH, Kleijnen J,
Evers SM, et al. How to prepare a systematic review of
economic evaluations for informing evidence-based
healthcare decisions: a five-step approach (part 1/3). Expert
Rev Pharmacoecon Outcomes Res 2016;16:689e704.

35. Drummond MF, Jefferson TO. Guidelines for authors
and peer reviewers of economic submissions to the BMJ.
The BMJ Economic Evaluation Working Party. BMJ
1996;313:275e83.

36. Buxton MJ, Drummond MF, Van Hout BA, Prince RL,
Sheldon TA, Szucs T, et al. Modelling in economic evaluation:
an unavoidable fact of life. Health Econ 1997;6:217e27.

37. Cleemput I, Kesteloot K. Economic implications of non-
compliance in health care. Lancet 2002;359:2129e30.

38. Cleemput I, Kesteloot K, DeGeest S. A review of the literature
on the economics of noncompliance. Room for
methodological improvement. Health Policy 2002;59:65e94.

http://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/datacollectionchecklist

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref12

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref13

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref14

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref15

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref16

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref17

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref18

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref19

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref20

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref21

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref22

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref23

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref24

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref25

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref26

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref27

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref28

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref29

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref30

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref31

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref32

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref33

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref34

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref35

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref36

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref37

http://refhub.elsevier.com/S0033-3506(19)30012-5/sref38

https://doi.org/10.1016/j.puhe.2019.01.012

Economic evaluations of public health implementation-interventions: a systematic review and guideline for practice
Introduction
Methods
Search strategy and study selection
Data extraction and analysis
Methodological quality
Costs and effects
Evidence synthesis

Results
Results of searches and screening
Study characteristics
Methodological quality
Evaluation design
Data included in the evaluations

Analysis and interpretation of evaluation results
Synthesis of review results

Discussion
Author statements
Acknowledgements

Ethical approval
Funding
Competing interests
Authors’ contributions

References

Towards-a-preventative-approach-to-improving-health-and-reducin_2019_Public-

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Themed Papere Review
Towards a preventative approach to improving
health and reducing health inequalities: a view
from Scotland
N. Craig a,*, M. Robinson b

a NHS Health Scotland, 1 South Gyle Crescent, Edinburgh, EH12 9EB, UK
b NHS Health Scotland, 5 Meridian Court Cadogan Street, Glasgow, G2 6QE, UK
a r t i c l e i n f o

Article history:

Received 20 May 2018

Received in revised form

23 October 2018

Accepted 4 February 2019

Available online 12 March 2019

Keywords:

Prevention

Cost-effectiveness

Health inequalities
* Corresponding author.Tel.: 0131 314 5444.
E-mail addresses: neil.craig@nhs.net (N.

https://doi.org/10.1016/j.puhe.2019.02.013
0033-3506/© 2019 The Authors. Published by
under the CC BY-NC-ND license (http://crea
a b s t r a c t

Pressures on the health system are intense. Prevention is often seen as a sustainable way

to manage these pressures. However, the impact of prevention on the demand for health

and social care is not fully understood. It will reflect the balance of opposing forces:

reduced needs for health and social care because of improving health and increased needs

associated with increasing life expectancy and the diseases of old age, mediated by how

the system manages the resulting pressures. This article illustrates how some of these

factors are playing out in Scotland. The article also highlights the substantial growth in the

evidence base on the economics of prevention and identifies policy developments with the

potential to support a shift to prevention that might help move towards more sustainable

demands on the health and social care system. These include recognition of the impor-

tance of the social determinants of health, the integration of health and social care and

‘realistic medicine’. The article suggests that more use needs to be made of available ev-

idence on the economics of prevention and that all stakeholders need to be engaged in

tackling the technical and political challenges posed by the shift to prevention.

Health. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Introduction

Pressures on the health system are intense.1 Demands

continue to grow more rapidly than the resources available.

Preventing the onset or progression of disease and preventing

inappropriate use of health services are seen as offering the

potential to manage growing demand in more sustainable

ways.2,3 However, the impact of prevention on the demand for
Craig), markrobinson1@n

Elsevier Ltd on behalf of
tivecommons.org/license
health care is not fully understood. In this article, we discuss

some of the challenges in understanding the role prevention

can play in managing demands on the health system.

We do not attempt a systematic or comprehensive review

of factors driving the demand for health care. These have been

considered at length elsewhere, for example, by Wanless as

long ago as the early 2000s4,5 and much more recently by the

Institute for Fiscal Studies.6 Rather, we focus on demographic
hs.net (M. Robinson).

The Royal Society for Public Health. This is an open access article
s/by-nc-nd/4.0/).

http://creativecommons.org/licenses/by-nc-nd/4.0/

mailto:neil.craig@nhs.net

mailto:markrobinson1@nhs.net

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.013&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.02.013

http://creativecommons.org/licenses/by-nc-nd/4.0/

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0196
and health criteria and select population health measures that

illustrate some of the challenges in understanding the eco-

nomic implications of prevention. A discussion section con-

siders the implications for future pressures on health and

other services important in improving the population’s health.
The focus is Scotland, but the themes explored are equally

relevant to the rest of the United Kingdom.
Background

Studies modelling future demands on health systems have

considered a wide variety of factors thought to influence the

balance between the demand for and supply of health care:

� changes in the health needs of the population;
� improvements in service quality and advances in medical
technologies;

� economic factors such as input prices and productivity
improvements; and

� changes in public and patient expectations, driven in part
by rising incomes.

In this article, we focus on changes in health needs and

how they might influence demand. Health needs are driven in

part by changes in:

� the age structure of the population, particularly the extent
to which life expectancy (LE) continues to rise and the

number of older people increases;

� the health status of the population, particularly the extent
to which improvements in LE are accompanied by im-

provements in healthy LE. The levels of ill health (partic-

ularly among elderly people) are key determinants of the

use of health and social care; and

� the likelihood of people seeking health care for a given
level of need.

Wanless highlighted the uncertainties in how these trends

might unfold over time and, crucially for the argument here,

that these trends are themselves dependent on decisions

taken at a policy, practice and individual level with respect to

prevention. Demand will depend in part on the age of the

population, but it will also depend on how healthily we age,

which in turn will depend in part on how much and how

efficiently we invest in prevention. Demand will depend on the

health status of the population, but it will also depend on how

much and how efficiently we invest in the systems in place to

manage that demand. Wanless described three scenarios that

made different assumptions about how these trends would

unfold. All three scenarios envisaged that substantial in-

creases in expenditure would be required to provide a publicly

funded, comprehensive and high quality service. The scenario

that led to the slowest growth in the resources required

assumed effective prevention leading to positive changes in

health and the determinants of health, and effective whole

system working to ensure the effective and efficient manage-

ment of demand. Over a decade later, substantial uncertainties

remain about how these trends will unfold.6 The remainder of

the article reflects on how these trends are developing in
Scotland, how the evidence base on the cost-effectiveness of

prevention has improved and what this might mean for future

investment in prevention and its potential to ease pressures on

the health and social care systems.
Trends in health needs and demands

The Scottish population is ageing. People aged 75 years and

above are projected to be the fastest growing age group in

Scotland. Their numbers are expected to increase by 27% over

the next 10 years and by 79% over the next 25 years.7 Com-

bined with the increased use of health and social care in later

life,8 projections such as these lead to stark warnings about

the future pressures faced by the health and social care sys-

tems. In practice, the health of the population will mediate the

link between ageing and demand, underlining the importance

of prevention. We return to this theme in the discussion.

Age-specific death rates over recent decades have declined

among almost all the major causes of deaths: cancer, coronary

heart disease, stroke, chronic obstructive pulmonary disease,

accidents and suicides. The only exception has been alcohol-

related death rates, which peaked in 2003 but then fell until

2012, when the downward trend in Scotland stalled. For

women, rates have actually increased in each of the last 3

years. Falling death rates are reflected in a general improve-

ment in LE at birth in Scotland since 1980, although the rate of

improvement has slowed in recent years and now appears to

have stalled.9,10

Male LE at birth in Scotland increased by nearly 4 years,

and female LE by just over 2 years since the early 2000s but

both changed little between 2012e14 and 2014e16, and LE at

birth in Scotland actually declined in 2015e2017 for males and

females. It also stopped falling in the United Kingdom as a

whole.11,12 This slowdown in the rate of improvement

occurred primarily amongst elderly men and women, in

whom mortality has actually increased in recent years.13

Similar trends have occurred in England and Wales.14

Healthy life expectancy (HLE) in Scotland changed little

between 2009e10 (when changes were made in the way HLE is

calculated, which mean that figures up to 2008 and from 2009

are not comparable) and 2015e2016.15 As a result, the gap

between LE and HLE for males born between 2009 and 2016

has increased slightly from 15.9 years to 17.7 years, and the

percentage of life expected to be spent in a ‘healthy’ state has

fallen, from 79% in 2009 to 77% in 2016. The gap between LE

and HLE has been fairly constant for females born between

2009 and 2016 (around 18.7 years).16 In short, in Scotland, we

are not seeing the compression of morbidity which we need to

see if the impact of an ageing population on demand for

health care is to be reduced.
Trends in behavioural health risk factors in
Scotland

Trends in behavioural risk factors present a mixed picture.

Scotland has among the highest levels of obesity prevalence

for men and women (aged 16e64) among Organisation for

Economic Co-operation and Development (OECD) countries.

https://doi.org/10.1016/j.puhe.2019.02.013

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0 197
Prevalence increased dramatically from 1995 until 2009/2010,

decreased slightly for males until 2014 and increased again in

2015 and 2016. Prevalence for women remained fairly con-

stant over the period 2008 to 2016.17 In contrast, smoking

prevalence dropped between 2003 and 2013, although it has

changed little since then.18 Likewise, after increasing over the

1990s and early 2000s, alcohol sales, generally regarded as the

best measure of consumption,19 stabilised between 2005 and

2009, declined until 2013 and have since remained broadly

stable.20 Data on levels of physical activity in Scotland show

that progress towards meeting national targets for increasing

walking, cycling and overall physical activity levels is

limited.21 Overall, therefore, there have been some successes

in relation to reducing the prevalence of risk factors likely to

increase the demand for care, but progress has been variable.
Trends in health inequalities in Scotland

Health inequalities in Scotland remain substantial. In 2016,

the death rate for the most deprived tenth of the population

(decile) based on small geographical areas was more than

double that for the least deprived decile.10 Inequalities in HLE

are also large. In 2015-16, male HLE at birth in the 10% most

deprived areas in Scotland was 43.9 years, 26.0 years lower

than in the least deprived areas (69.8 years). For women, the

gap was 22.2 years.15 This has changed little since 2009e2010.

With the exception of the Healthy Birthweight indicator, sig-

nificant health inequalities persist for each indicator covered

in the Scottish Government’s Monitoring Long-Term In-
equalities report, although some have narrowed in absolute

and/or relative terms.15

Absolute inequalities in some measures of health have

fallen as the incidence or prevalence has fallen at a population

level. For some, such as alcohol-related mortality, falling ab-

solute inequalities have been driven by reduced rates in the

most deprived areas. But relative inequalities on many mea-

sures remain wide, some have increased even as absolute

inequalities have diminished, and on some measures, neither

absolute nor relative inequalities are narrowing. For example,

the latest data on trends in health inequalities in Scotland for

premature (<75y) mortality and HLE show that for both, in- equalities are now increasing in relative and absolute terms.15 The ageing of the population may reinforce these pressures given that the social gradient in health across the adult pop- ulation is replicated amongst the elderly.22 Improvement in the evidence base There has been a massive growth in the cost-effectiveness evidence available to inform prevention-focused public health strategies,23e27 including strategies for tackling the wider determinants of health.28e31 The evidence remains patchy in terms of coverage and methodological quality. Evi- dence on the cost-effectiveness of preventative measures to reduce health inequalities is particularly limited. However, the implications are consistent: many preventative in- terventions are cost-effective, many have the potential to reduce demand for health and social care and some are potentially cost-saving. Some authors have challenged the emphasis on savings in a health context,32e34 a theme we return to later. Others have stressed the need for economic evaluation to include consideration of the costs of conditions unrelated to people's initial diagnoses and care, which people may experience as a result of enjoying greater LE because of the care they receive. This would ensure that economic eval- uation would give a more accurate picture of the longer term resource implications and opportunity costs of care, both preventative and curative.35 Regardless, the key point is that the evidence base has improved substantially, although it is not yet applied sys- tematically to inform policy, and there is a tendency to- wards ‘lifestyle drift’. That is, despite the long-standing recognition in official publications of the importance of the social and economic determinants of health,4 prevention efforts in practice tend to focus on individuals' behaviours.36 Evidence suggests that the latter have the potential to widen inequalities.37 Discussion Central to current discussions regarding pressures on health and social care is the question of whether and how prevention might affect demand and spending growth by influencing trends and inequalities in population health. Other drivers of spend include the development of new technologies, input prices and rising expectations as incomes rise. Many analyses suggest that these are more powerful drivers than the ageing or the health of the population.38 However, we focus on health and its link with ageing because they drive underlying needs, albeit in complex ways, and because they are amenable to change through effective prevention. The scope of a healthier population to reduce cost growth is crucial. However, the impacts of improvements in popula- tion health on the demand for health and social care are ambiguous. Overall, the impact of prevention will reflect the balance of opposing forces: reduced needs for health and so- cial care because of improving health and increased needs associated with increasing LE and the diseases of old age mediated by how the system manages the resulting pressures. Projections of the implications of demographic change often apply current rates of service use by age group to future population structures. Because spending rises steeply with age, driven by the increasing prevalence of multimorbidity, such analyses suggest that population ageing will lead to ris- ing health and social care expenditure. However, the extent of this effect depends on whether and how the health of the population improves and on the quality and efficiency of service delivery.39 Current rates of service use are not fixed. They can be changed to manage the future growth in demand arising from demographic change, although further work is required to better understand this relationship.38 There are examples, such as alcohol, where, in recent years, rates of alcohol-related hospital admissions have fallen alongside alcohol-related death rates,20 but more generally and in the longer term, falling death rates are not necessarily associated with increasing HLE. Likewise, falling incidence of a disease does not necessarily equate to fewer people with the https://doi.org/10.1016/j.puhe.2019.02.013 https://doi.org/10.1016/j.puhe.2019.02.013 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0198 disease: over the 10 years to 2016, age-adjusted incidence rates for cancer decreased by 2.6% yet the numbers of people diagnosed with cancer increased, in large part owing to the increase in the number of older people in the population.40 The recent Scottish Burden of Disease Study also underlines why the ageing of the population is central to understanding demand on health and social care. Many of the biggest ‘bur- dens’ are for preventable, non-fatal conditions that increase in prevalence with age.41 A distinction needs to be drawn here between slowing demand growth and making financial savings. Savings are often held up as a potential dividend from an emphasis on prevention, but they depend on resources being released from their current use if and when demand falls because of effec- tive prevention. If not, higher quality services may be pro- vided, or other people may be treated to address current levels of unmet need. Both are clearly good things for patients and staff, but they limit the scope to make savings that can be reinvested elsewhere in the health or social care systems.42 Changing how resources are used also poses technical chal- lenges because of the specialist nature of many capital and human resources in health care and political challenges because of popular commitment to local services. In addition, other areas of health care are not required to demonstrate savings and, as noted already, the long-term cost implications of prevention are unclear. In terms of improving health, despite notable progress in some areas, overall progress has been variable, across risk factors and across socio-economic groups. The persistence of health inequalities contributes to what the Christie Commis- sion called ‘failure demand’, that is, demand for public ser- vices that could have been avoided by earlier preventative measures.2 To the extent that reduced absolute inequalities reflect a falling disease burden at a population level, they may be associated with reducing need for health and care, but the scale of the inequalities that remain suggests that inequalities will continue to be an important driver of demand for health and social care, reinforced by the social patterning of many risk factors. The key policy question is what forms of prevention offer the best investments as a way of improving public health whilst reducing these inequalities. A good economic case can be made for prevention in general, albeit that the evidence needs to be considered on a case by case basis as preventative measures are not necessarily cost-effective.27,42 However, an increasing part of the evidence base relates to interventions tackling the socio-economic determinants of health, with much of the evidence suggesting very high returns for the investments made and a greater potential to reduce health inequalities than interventions reducing behavioural risk factors.26,43 There is a growing consensus that in principle, ‘upstream’ prevention is a good buy, but this is also where the evidence base remains most patchy, both in terms of coverage and in terms of methodological rigour and consistency.43,44 Encouragingly, the importance of health inequalities and of addressing them through the social determinants of health is recognised in current policy discourse45 and in the key role for local government in public health reform in Scotland.46,47 Future resource requirements will also be driven in part by how we choose to meet care needs. Integration of health and social care bodies has the potential to accelerate the process of shifting health care into the community, avoiding the hospital- based treatment of health and social care needs best met in the community. However, integration of services on the ground has started slowly because of the complexity of the change and the challenging workforce and governance issues involved.48 The relationship between need and demand has been addressed by the Chief Medical Officer in Scotland in recent annual reports promoting ‘Realistic Medicine’. The most recent report highlights the need to involve patients and their families in decisions about their care, based on realistic ex- pectations about the scope of health care to add value to people's lives. As more people and less severely ill people are treated, health gains might diminish whilst the risks associ- ated with treatment persist. At some point, patients and their families may decide that these risks outweigh the potential benefits and opt for more conservative management of poor health, preventing demand for care of limited value to pa- tients. ‘Realistic Medicine’, by involving patients and their families more in their care decisions, has the potential to ensure that services are shaped more closely to patients' and their families' preferences. Evidence suggests that, for some, this will mean doing less or no treatment is the best option.45 This article highlights two necessary preconditions for moving towards a more preventative approach to improving health and reducing health inequalities. First, we need to make better use of the growing evidence base on the cost- effectiveness of preventative interventions for improving health and reducing health inequalities. The evidence base remains patchy, but it is large, it is growing and it is consistent in demonstrating the cost-effectiveness of many preventative interventions. Second, we need to harness the political will to move towards a more preventative approach with the deter- mination and skills required to use the available evidence in a complex policy environment shaped by competing interests. There are seeds of hope in the growing recognition of the importance of social determinants of health, the integration of health and social care and ‘Realistic Medicine’. These need to be nurtured using what we know about the cost- effectiveness of prevention and by engaging all stakeholders in the technical and political challenges it poses. Otherwise, there is a risk that ‘yet another opportunity to act will have been missed and healthcare services will continue to run faster and faster to stand still.’4 Author statements Acknowledgements The authors are grateful to Dr Andrew Fraser, Dr Gerry McCartney and two anonymous referees for comments on an earlier version of this article. They are also grateful to Seona Hamilton for help with managing the references. Ethical approval None sought. https://doi.org/10.1016/j.puhe.2019.02.013 https://doi.org/10.1016/j.puhe.2019.02.013 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0 199 Funding None declared. Competing interest None declared. r e f e r e n c e s 1. Scotland Audit. NHS in Scotland 2017. Edinburgh: Audit Scotland; 2017. 2. Commission on the Future Delivery of Public Services. Report on the future delivery of public services by the Commission [Internet]. Edinburgh: Scottish Government; 2011 [cited 17 May 2018]. Available from: www.gov.scot/Publications/2011/ 06/27154527/0. 3. Scottish Parliament Finance Committee. Report on preventative spend, SP Paper 555, 1st report 2011, session 3 [Internet]. Edinburgh: Scottish Parliament [cited 17 May 2018]. Available from: http://archive.scottish.parliament.uk/s3/ committees/finance/reports-11/fir11-01.htm. 4. Wanless D. Securing good health for the whole population. Final report. London: HMSO; 2004. 5. Wanless D. Securing our long term health. Taking a long term view. Final report. London: HMSO; 2002. 6. Johnson P, Kelly E, Lee T, Stoye G, Zaranko B, Charlesworth A, et al. Securing the future: funding health and social care to the 2030s. London: The Institute for Fiscal Studies; 2018. 7. National Records of Scotland. Projected population of Scotland (2016-based). National population projections by sex and age, with UK comparisons [Internet]. Edinburgh: National Statistics; 2017. - [cited 17 May 2018]. Available from: https://www.nrscotland. gov.uk/files//statistics/population-projections/2016-based- scot/pop-proj-2016-scot-nat-pop-pro-pub . 8. Information Services Division. Health and social care integrated resource framework. NHS Scotland and local authority social care expenditure. Years ending March 2013-2016 [Internet]. Edinburgh: NHS National Services Scotland; 2017. Available from: https://www.isdscotland.org/Health-Topics/Health- and-Social-Community-Care/Publications/2017-12-05/2017- 12-05-IRF-Health-and-Social-Care-Resource-Use-2012-16- Summary . 9. National Records of Scotland. Life tables for Scotland 2014-16 [Internet]. Edinburgh: National Statistics; 2017 [cited 17 May 2018]. Available from: https://www.nrscotland.gov.uk/ statistics-and-data/statistics/statistics-by-theme/life- expectancy/life-expectancy-at-scotland-level/scottish- national-life-tables/2014-2016/national-life-tables. 10. Scottish Public Health Observatory. Deaths: key points [Internet]. Edinburgh: ScotPHO; 2017 [cited 17 May 2018]. Available from: http://www.scotpho.org.uk/population- dynamics/deaths/key-points/. 11. National Records of Scotland. Life expectancy for administrative areas within Scotland. Time series data [Internet]. Edinburgh: National Statistics; 2017 [cited 17 May 2018]. Available from: https://www.nrscotland.gov.uk/statistics-and-data/statistics/ statistics-by-theme/life-expectancy/life-expectancy-in- scottish-areas/time-series-data. 12. [Internet] Anonymous ONS National life tables, UK: 2015 to 2017. 2018. Available from: https://www.ons.gov.uk/ peoplepopulationandcommunity/birthsdeathsandmarriages/ lifeexpectancies/bulletins/nationallifetablesunitedkingdom/ 2015to2017. 13. McCartney G, Fischbacher C. Understanding mortality in Scotland from 2015. Briefing note. Edinburgh: NHS Health Scotland; 2017. 14. Hiam L, Harrison D, McKee M, Dorling D. Why is life expectancy in England and Wales 'stalling'? J Epidemiol Commun Health 2018;72(5):404e8. 15. Scottish Government. Long-term monitoring of health inequalities: December 2017 [Internet]. Edinburgh: Scottish Government; 2017 [cited 17 May 2018]. Available from: http:// www.gov.scot/Publications/2017/12/4517. 16. Scottish Public Health Observatory. Healthy life expectancy: Scotland [Internet]. Edinburgh: ScotPHO; 2017 [cited 17 May 2018]. Available from: http://www.scotpho.org.uk/ population-dynamics/healthy-life-expectancy/data/ scotland/. 17. Scottish Public Health Observatory. Obesity: key points [Internet]. Edinburgh: ScotPHO; 2017 [cited 17 May 2018]. Available from: http://www.scotpho.org.uk/clinical-risk- factors/obesity/key-points. 18. Scottish Government. Scottish Health Survey 2016: volume 1: main report [Internet]. Edinburgh: Scottish Government; 2017 [cited 1 April 2018]. Available from: http://www.gov.scot/ Publications/2017/10/2970/0. 19. Beeston C, McAdams R, Craig N, Gordon R, Graham L, MacPherson M, et al. Monitoring and evaluating Scotland's alcohol strategy: final report. Edinburgh: NHS Health Scotland; 2016. 20. Giles L, Robinson M. Monitoring and evaluating Scotland's alcohol strategy: monitoring report 2018. Edinburgh: NHS Health Scotland; 2018. 21. Scottish Public Health Observatory. Physical activity: key points [Internet]. Edinburgh: ScotPHO; 2017 [cited 17 May 2018]. Available from: http://www.scotpho.org.uk/behaviour/ physical-activity/key-points/. 22. NHS Health Scotland. Population groups: older people [Internet]. Edinburgh: NHS Health Scotland; 2018. Available from: http:// www.healthscotland.scot/population-groups/older-people. 23. Owen L, Morgan A, Fischer A, Ellis S, Hoy A, Kelly MP. The cost-effectiveness of public health interventions. J Pub Health 2011;34(1):37e45. 24. Merkur S, Sassi F, McDaid D [Internet]. Promoting Health, Preventing Disease: is There an Economic Case? Policy Summary, 6. Geneva: World Health Organization; 2013 [cited 17 Many 2018]. Available from: www.euro.who.int/__data/assets/pdf_ file/0004/235966/e96956 . 25. WHO Regional Office for Europe. The case for investing in public health. Geneva: World Health Organization; 2014. 26. Masters R, Anwar E, Collins B, Cookson R, Capewell S. Return on investment of public health interventions: a systematic review. J Epidemiol Commun Health 2017;71(8):827e34. 27. Owen L, Pennington B, Fischer A, Jeong K. The cost- effectiveness of public health interventions examined by NICE from 2011 to 2016. J Pub Health 2017:1e10. 28. World Health Organization. The economics of the social determinants of health and health inequalities: a resource book [Internet]. Geneva: WHO; 2013 [cited 17 May 2018]. Available from: http://apps.who.int/iris/bitstream/handle/10665/84213/ 9789241548625_eng . 29. Buck D, Gregory S. Improving the public's health. A resource for local authorities [Internet]. London: The King's Fund; 2013. Available from: www.kingsfund.org.uk/publications/ improving-publics-health. 30. Institute for Health Equity and Public Health England. Local action on health inequalities: understanding the economics of investments in the social determinants of health [Internet]. London: Public Health England; 2014 [cited 17 May 2018]. Available from: https://assets.publishing.service.gov.uk/ government/uploads/system/uploads/attachment_data/file/ http://refhub.elsevier.com/S0033-3506(19)30037-X/sref1 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref1 http://www.gov.scot/Publications/2011/06/27154527/0 http://www.gov.scot/Publications/2011/06/27154527/0 http://archive.scottish.parliament.uk/s3/committees/finance/reports-11/fir11-01.htm http://archive.scottish.parliament.uk/s3/committees/finance/reports-11/fir11-01.htm http://refhub.elsevier.com/S0033-3506(19)30037-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref4 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref5 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref6 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref6 https://www.nrscotland.gov.uk/files//statistics/population-projections/2016-based-scot/pop-proj-2016-scot-nat-pop-pro-pub https://www.nrscotland.gov.uk/files//statistics/population-projections/2016-based-scot/pop-proj-2016-scot-nat-pop-pro-pub https://www.nrscotland.gov.uk/files//statistics/population-projections/2016-based-scot/pop-proj-2016-scot-nat-pop-pro-pub https://www.isdscotland.org/Health-Topics/Health-and-Social-Community-Care/Publications/2017-12-05/2017-12-05-IRF-Health-and-Social-Care-Resource-Use-2012-16-Summary https://www.isdscotland.org/Health-Topics/Health-and-Social-Community-Care/Publications/2017-12-05/2017-12-05-IRF-Health-and-Social-Care-Resource-Use-2012-16-Summary https://www.isdscotland.org/Health-Topics/Health-and-Social-Community-Care/Publications/2017-12-05/2017-12-05-IRF-Health-and-Social-Care-Resource-Use-2012-16-Summary https://www.isdscotland.org/Health-Topics/Health-and-Social-Community-Care/Publications/2017-12-05/2017-12-05-IRF-Health-and-Social-Care-Resource-Use-2012-16-Summary https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-at-scotland-level/scottish-national-life-tables/2014-2016/national-life-tables https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-at-scotland-level/scottish-national-life-tables/2014-2016/national-life-tables https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-at-scotland-level/scottish-national-life-tables/2014-2016/national-life-tables https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-at-scotland-level/scottish-national-life-tables/2014-2016/national-life-tables http://www.scotpho.org.uk/population-dynamics/deaths/key-points/ http://www.scotpho.org.uk/population-dynamics/deaths/key-points/ https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-in-scottish-areas/time-series-data https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-in-scottish-areas/time-series-data https://www.nrscotland.gov.uk/statistics-and-data/statistics/statistics-by-theme/life-expectancy/life-expectancy-in-scottish-areas/time-series-data https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/bulletins/nationallifetablesunitedkingdom/2015to2017 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/bulletins/nationallifetablesunitedkingdom/2015to2017 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/bulletins/nationallifetablesunitedkingdom/2015to2017 https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/bulletins/nationallifetablesunitedkingdom/2015to2017 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref13 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref14 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref14 http://www.gov.scot/Publications/2017/12/4517 http://www.gov.scot/Publications/2017/12/4517 http://www.scotpho.org.uk/population-dynamics/healthy-life-expectancy/data/scotland/ http://www.scotpho.org.uk/population-dynamics/healthy-life-expectancy/data/scotland/ http://www.scotpho.org.uk/population-dynamics/healthy-life-expectancy/data/scotland/ http://www.scotpho.org.uk/clinical-risk-factors/obesity/key-points http://www.scotpho.org.uk/clinical-risk-factors/obesity/key-points http://www.gov.scot/Publications/2017/10/2970/0 http://www.gov.scot/Publications/2017/10/2970/0 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref19 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref19 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref19 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref19 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref20 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref20 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref20 http://www.scotpho.org.uk/behaviour/physical-activity/key-points/ http://www.scotpho.org.uk/behaviour/physical-activity/key-points/ http://www.healthscotland.scot/population-groups/older-people http://www.healthscotland.scot/population-groups/older-people http://refhub.elsevier.com/S0033-3506(19)30037-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref23 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref23 http://www.euro.who.int/__data/assets/pdf_file/0004/235966/e96956 http://www.euro.who.int/__data/assets/pdf_file/0004/235966/e96956 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref25 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref25 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref26 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref27 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref27 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref27 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref27 http://apps.who.int/iris/bitstream/handle/10665/84213/9789241548625_eng http://apps.who.int/iris/bitstream/handle/10665/84213/9789241548625_eng http://www.kingsfund.org.uk/publications/improving-publics-health http://www.kingsfund.org.uk/publications/improving-publics-health https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/356051/Briefing9_Economics_of_investments_health_inequalities https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/356051/Briefing9_Economics_of_investments_health_inequalities https://doi.org/10.1016/j.puhe.2019.02.013 https://doi.org/10.1016/j.puhe.2019.02.013 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 9 5 e2 0 0200 356051/Briefing9_Economics_of_investments_health_ inequalities . 31. Institute for Health Equity and Public Health England. Local action on health inequalities: evidence papers [Internet]. London: Public Health England; 2014 [cited 17 May 2018]. Available from: https://www.gov.uk/government/publications/local- action-on-health-inequalities-evidence-papers. 32. Rappange DR, Brouwer WB, Rutten FF, van Baal PH. Lifestyle intervention: from cost savings to value for money. J Pub Health 2009;32(3):440e7. 33. Ferguson B. Cost savings and the economic case for investing in public health. Public Health Matters [Internet]. London: Public Health England; 2018 [cited 9 April 2018]. Available from: https://publichealthmatters.blog.gov.uk/2018/04/09/cost- savings-and-the-economic-case-for-investing-in-public- health/. 34. Buck D. Talking about the ‘return on investment of public health’: why it's important to get it right [Internet]. London: The King's Fund; 2018 [cited 23 April 2018]. Available from: https://www. kingsfund.org.uk/blog/2018/04/return-investment-public- health. 35. Morton A, Adler A, Briggs A, Bell D, Brouwer W, Claxton K, et al. Unrelated future costs and unrelated future benefits: reflections on NICE guidance to the methods of technology appraisal. Health Econ 2016;25(8):933e88. 36. Kriznik N, Kinmonth A, Ling T, Kelly M. Moving beyond individual choice in policies to reduce health inequalities: the integration of dynamic with individual explanations. J Pub Health 2018. https://doi.org/10.1093/pubmed/fdy045. 37. Beeston C, McCartney G, Ford J, et al. Health inequalities policy review for the Scottish ministerial task force on health inequalities [Internet]. Edinburgh: NHS Health Scotland; 2013 [cited 17 May 2018]. Available from: www.healthscotland.com/ documents/23047.aspx. 38. Appleby J. Spending on health and social care over the next 50 years. Why think long-term? London: The King's Fund; 2013. 39. Barbour J, Morton A, Schang L. The Scottish NHS: meeting the financial challenge ahead. Fraser Allender Econ Commen 2014;38(2):126e46. 40. Information Services Division. Cancer incidence in Scotland 2017. Edinburgh: NHS National Services Scotland; 2018. 41. Scottish Public Health Observatory. The Scottish burden of disease study 2015. Overview report. Edinburgh: NHS Health Scotland; 2017. 42. NHS Health Scotland. Economics of prevention. Edinburgh: NHS Health Scotland; 2016. 43. Webber L, Chalkidou K, Morrow S, Ferguson B, McKee M. What are the best societal investments for improving people's health? BMJ 2018:362. 30 August 2018. 44. Craig N. Best preventative investments for Scotland - what the evidence and experts say. Edinburgh: NHS Health Scotland; 2014. 45. Chief Medical Officer. Chief Medical Officer's annual report 2016- 17. Practising realistic medicine. Edinburgh: Scottish Government; 2018. 46. Scottish Government. Review of public health in Scotland. Strengthening the function and re-focussing action for a healthier Scotland. Edinburgh: Scottish Government; 2015. p. 2016. 47. Scottish Government. Health and social care delivery plan. Edinburgh: Scottish Government; 2016. 48. Audit Scotland. Health and social care integration. Edinburgh: Audit Scotland; 2015. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/356051/Briefing9_Economics_of_investments_health_inequalities https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/356051/Briefing9_Economics_of_investments_health_inequalities https://www.gov.uk/government/publications/local-action-on-health-inequalities-evidence-papers https://www.gov.uk/government/publications/local-action-on-health-inequalities-evidence-papers http://refhub.elsevier.com/S0033-3506(19)30037-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref32 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref32 https://publichealthmatters.blog.gov.uk/2018/04/09/cost-savings-and-the-economic-case-for-investing-in-public-health/ https://publichealthmatters.blog.gov.uk/2018/04/09/cost-savings-and-the-economic-case-for-investing-in-public-health/ https://publichealthmatters.blog.gov.uk/2018/04/09/cost-savings-and-the-economic-case-for-investing-in-public-health/ https://www.kingsfund.org.uk/blog/2018/04/return-investment-public-health https://www.kingsfund.org.uk/blog/2018/04/return-investment-public-health https://www.kingsfund.org.uk/blog/2018/04/return-investment-public-health http://refhub.elsevier.com/S0033-3506(19)30037-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref35 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref35 https://doi.org/10.1093/pubmed/fdy045 http://www.healthscotland.com/documents/23047.aspx http://www.healthscotland.com/documents/23047.aspx http://refhub.elsevier.com/S0033-3506(19)30037-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref38 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref39 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref40 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref41 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref42 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref43 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref44 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref45 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref46 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref47 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref48 http://refhub.elsevier.com/S0033-3506(19)30037-X/sref48 https://doi.org/10.1016/j.puhe.2019.02.013 https://doi.org/10.1016/j.puhe.2019.02.013 Towards a preventative approach to improving health and reducing health inequalities: a view from Scotland Introduction Background Trends in health needs and demands Trends in behavioural health risk factors in Scotland Trends in health inequalities in Scotland Improvement in the evidence base Discussion Author statements Acknowledgements Ethical approval Funding Competing interest References Cognitive-biases-in-public-health-and-how-economics-and-socio_2019_Public-He ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2 Available online at w Public Health journal homepage: www.elsevier.com/puhe Themed Papere Original Research Cognitive biases in public health and how economics and sociology can help overcome them M.P. Kelly Primary Care Unit, Department of Public Health and Primary Care, Cambridge Institute of Public Health, University of Cambridge, Forvie Site, Cambridge, CB2 0SR, UK a r t i c l e i n f o Article history: Received 4 April 2018 Received in revised form 16 October 2018 Accepted 4 February 2019 Available online 15 March 2019 Keywords: Heuristics Cognitive dissonance Societal dynamics Mechanisms of prevention Socio-economics Decision theoretic models https://doi.org/10.1016/j.puhe.2019.02.012 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: The objective of this study was to identify important gaps in the public health evidence base and consider the implications of these for public health and public health economics. Study design: This was a review and critique of public health policy in the UK. Methods: Using two key psychological concepts relating to cognitive biases, viz. cogni- tive dissonance and heuristics, the shortcomings in public health approaches to con- fronting the prevalence of non-communicable diseases are described. The implications are drawn out. Results: Two cognitive biases in public health thinking are identified. (i) A dissonance be- tween what is known and what is done, resulting in the repetition of solutions that have previously been shown to have had little or no effect. (ii) The habitual use of set of heu- ristics which mean that simple solutions to complex problems are preferred to undertaking the detailed assessment of how to bring about change. These biases mean that the evi- dence about the dynamics of populations and the ways that the mechanisms of prevention actually operate seldom feature in the way interventions, policy and practice are under- taken. The evidence base is consequently highly skewed. Conclusions: Health economics combined with sociological reasoning has potentially an important role to play in developing the ideas that will overcome the problems attaching to the cognitive biases. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. about complexity from economics and sociology will help to Introduction Two cognitive biases in public health thinking are barriers to effective policy implementation: a dissonance between what is known and what is done and the habitual use of short cuts in thinking involving finding simple solutions to complex problems. This article argues that the consequence of these biases is that evidence-based interventions are not put to work effectively. It is suggested that a coalescence of ideas ic Health. Published by E begin to break down these barriers. The evidence base in public health has expanded and developed considerably in the last two decades. The knowl- edge base is large and growing, and these advances look set to continue (See, for example, https://www.nihr.ac.uk/about-us/ documents/NIHR-Annual-Report-2015-16 ). The evidence base provides a potentially important platform for policy and practice development. Yet, there is a gap between what is lsevier Ltd. All rights reserved. https://www.nihr.ac.uk/about-us/documents/NIHR-Annual-Report-2015-16 https://www.nihr.ac.uk/about-us/documents/NIHR-Annual-Report-2015-16 http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.012&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2164 known and what is done. In spite of the strength and depth of this evidence and in spite of the existence of Cochrane (formerly the Cochrane Collaboration), the National Institute for Health and Care Excellence (NICE) and the infrastructures created by the National Institute for Health Research (NIHR), the School of Public Health Research, the Medical Research Council (MRC), Public Health England (PHE) and the Depart- ment of Health and Social Care, as well as similar de- velopments in Scotland, Wales and Northern Ireland, there is a paradox. Some of the most important research undertaken about primary, secondary and tertiary prevention in the world is funded and conducted in the United Kingdom. There are excellent research centres and arm's-length bodies producing primary research and reviews of the evidence about risk and causal pathways to disease at individual or population level, as well as about the cost-effectiveness of interventions. There are well-argued and extensive data about the patterning of health at population level and in particular about health in- equalities. The Research Excellence Framework 2014 (REF) results in Unit of Assessment 2 (Public Health) showed that the research outputs in that unit stand comparison with, and in many cases surpass, the best from the top centres in the US and elsewhere.1 The quality of British research in public health is world class. However, much of what is known, discovered and revealed by the research is not, or is only partially, implemented in policy and practice. The reasons for this, it will be argued, are two cognitive biases that are found in policy, practice and interventions. First, a dissonance be- tween what is known and what is done, which results in the repetition of that which has previously been shown to have had little or no effect.2e4 Second, the habitual use of set of heuristics or short cuts in thinking which mean that simple solutions to complex problems are preferred to undertaking the detailed assessment of how to bring about change.5,6 Therefore, what is known, and indeed what is known to be not known, is not subject to detailed analysis and scrutiny. A number of consequences follow from this state of affairs. The fact that there are gaps between research and policy and research and practice is not, of course, a new observa- tion.7,8 It has often been noted and commented on, not just in public health and not just in the UK. There are many models, which purport to describe and to explain the problem or which seek to find solutions. There are linear ones such as that pro- posed by Cooksey,9 cyclical models,10 explanations couched in terms of complexity theory11e13 and political models.14 There are journals devoted to implementation science and a system of university performance management in the UK e the REF e designed to bridge the gap in various ways.15 However, Public Health England (PHE) suggests that there is still a 17-year gap between initiation and uptake of research.16 Eerily, all of this echoes the observation made by Archie Cochrane back in 1972 about the failure of the medical profession to keep up to date with emerging evidence and innovation.17 (For an overview about public health see Orton et al.18) Methods This article seeks to identify the cognitive biases specifically as they apply to public health. The author worked for 14 years before his retirement at senior national policy level in public health in England. The observations reported here are based on that experience. The focus is the relationship between what is known from the evidence and why it is applied sub- optimally. The discussion takes place within the framework outlined by Smith.19 She argued that the relationship between research and action/policy/practice is best understood as a constant interaction between many actors, which include, but is not restricted to, researchers, policy makers and practi- tioners. It also embraces opinion formers, journalists, politi- cians, officials, pressure groups, vested interests, industry, commerce and members of the public. The interplay of ideas between these various actors has emergent properties that include policy and practice outcomes. The argument in this article is that (i) within that commerce of ideas, the cognitive biases play a significant role in shaping the discourse and (ii) we need to develop economic ideas to better meet the con- sequences of these biases. It has been noted for many decades, prefiguring the ar- guments in this article, that the nature of economic and social arrangements are the product of the multitude of different actions and the knowledge of countless individual players acting according to their own volition. Furthermore, the collective consequence of these actions is the structure of economic and social systems. Hayek, for example, argued back in the 1940s that the price mechanism operates because of the actions of many players in the market and that it would be impossible therefore to plan an economic system because no single economic actor or the state could com- mand sufficient knowledge and understanding of the com- plexities involved.20 Hayek saw the price system as both the determinant and product of individual actions. The sociolo- gist Antony Giddens used a similar argument about the so- cial structure which he argued was the product of the interactions between the many millions of individual human actions. These many millions of individual actions produce structured patterns at the level of society. The patterned structure in turn constrains and limits the choices open to individuals.21 The importance of this coalescence in eco- nomic and sociological thinking will be considered in the following sections, but the fact is that the failure to acknowledge the importance of the multiplicities of human actions is fundamental to the cognitive biases with which this article is concerned. Results: systematic biases in the evidence base What is known about populations and the way public health policy and practice generally describe populations The first bias is the dissonance between what social scientists have discovered about how populations work22 and the way that populations are conceptualised in public health.23,24 The understanding of the effectiveness of interventions, and how to put them into practice for example, is seriously hindered by a lack of understanding of the mechanisms by which in- terventions work in different segments of the population. This in turn is compounded by a lack of use of knowledge and ev- idence about population dynamics themselves. https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2 165 The way that populations work as dynamic entities and the implications of this for public health policy and practice is neither well understood nor theorised very adequately.25 At its most basic, the categories used to describe the population are crude and coarse (socio-economic status [SES], ethnicity, gender, education and income). These categories are associ- ated with health and illness, but the categories themselves conceal the enormous diversity in the population e its social variegation. They are coarse because they are unidimensional e people are not just their occupation, or their educational background or their ethnicity. People are members of multiple social groupings with distinctive lifeworlds; they engage in diverse social practices, and their behaviour and their health are not the product of, nor can be reduced to, single categor- ical variables. They are crude because they are reductionist e focussing on bringing things down to the individual level of analysis. SES, for example, is treated as a characteristic of an individual, usually with reference to occupation, education or income. However, the collectivities to which people belong and the relationships between those collectivities are drivers of the dynamics of society. In turn, these relationships, and particularly power relationships in the workplace, commu- nities and markets, are fundamental to the patterning of health and disease. However, this relational dynamic is ignored. Instead, the categories are assigned to individuals and then used to count individual events cross sectionally before being aggregated up to population level and then sometimes plotted longitudi- nally. The intersections between the complex dynamic social variations in the population are seldom pressed into service in public health thinking.26 The interactions and synergies be- tween age, gender, ethnicity, occupation, tribe, caste, geog- raphy and housing tenure and the ways these change across time profoundly shape people's lives and their health. The fluctuations and nuances in the intersecting lifeworlds of the populace and the ways these interact with each other and change over time, all of which would allow us to understand better the differential effectiveness of interventions, are simply not usually part of the discourse of disease prevention or health protection and promotion. This is in spite of the fact that there is an evidence base to draw upon. The work of sociologists such as Savage,27 which pick up the nuances of life in contemporary Britain in a highly sophisticated way, seems to have passed by the public health community at large, in spite of the fact that arguments for embracing this type of thinking have been around in the literature for more than a decade.28 Consequently, the cate- gories do not capture the huge amount of individual and col- lective variation, and they fail to treat the population as a dynamic entity. Populations or societies are dynamic and constantly changing. They are not the mere aggregation of the individuals which make up the population.29 Ontologically, populations are real in themselves and can be analysed in themselves and understood in that way.30 In short, society and the communities and neighbourhoods within them are not treated as complex entities with recursive interactions occurring continuously between the social, sub- jective, biological and physical phenomena.31,32 In general, the public health community does not seek to make sense of the complexity. There is a lot of rhetoric about complexity in the literature, although none of this really has a social or eco- nomic theory of complexity embedded in it. The result is that actions designed to help to protect the population from dis- ease, to protect it from hazards and to promote good health are largely complexity-free zones. The default is to fall back on cross-sectional accounts of the categories, which conceal as much as they reveal. This is true even of the landmark at- tempts by MRC to deal with complexity.33e35 These contribu- tions were very important in moving the argument forward but, to date, have been deficient in providing a social theory or an empirical account of complexity themself. There is a collective cognitive dissonance.36 On one side is the knowledge base about the complexity to be found in so- ciology, psychology, anthropology and economics and in some of the landmark statements by physicians such as Engel37 (and even in the rhetoric of public health38). On the other side are high-level descriptions of health inequalities and the continuation of interventions that fail to deal with the prevalence of non-communicable disease.39,40 The mecha- nisms operating at the population level are not described in ways that are useful to do interventions or to change things. Guideline developers and others constructing in- terventions are in effect hamstrung because they do not have empirical or theoretical accounts of dynamic social mecha- nisms.41 They are armed only with data about associations.42 It is very difficult to get down to a level of granularity to help to develop interventions that would be fit for purpose in different sections and segments of populations and to tailor to the needs of specific groups and communities. Data about asso- ciations or correlations do not explain cause, although cause may sometimes very helpfully be inferred from associational data. From a causal point of view, however, we need to describe and understand the mechanisms to be able to describe causes. By articulating the mechanisms involved (by taking the dynamic approach to the social, economic and biological advocated here), we are able to identify points for intervention more forensically and with greater granularity. My argument is that this obvious point is habitually ignored in policy, which remains fixated on individually behaviourally based solutions when individual human behaviour is at best only one part (and not necessarily the most important part) of the mechanisms at work. The necessary knowledge of mechanisms is simply missing. In clinical medicine, the fact there is biological variation between individuals is a sine qua non of practice e at its most basic, not everyone responds to the same drug or treatment in the same way. The implications of the fact of social variation are just as important, but the efforts to explore and understand that variation are, in public health terms, in their infancy. The argument is that the categories such as SES, ethnicity, gender and so on are not used in ways that capture complex dynamic processes and the mechanisms involved. This is because the way they are habitually conceptualised and used in the public health literature, and more specifically in policy, is reductionist.43 The categories do not in themselves imply reductionism, but the way they are used is reductionist, with the consequent concept and policy default to individually based solutions.44 It is however possible to conceptualise the social level or population level separately from in the indi- vidual one, in a way analogous to the economic level of https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2166 analysis of the market. This helps to get through the cognitive biases. It is as if this idea of the dynamic social level is hidden in plain sight. The individual is so obviously and common- sensically the focus because the world is made up of in- dividuals. But, individuals interacting with each other produce a reality that is at once both economic and social, as Hayek and Giddens in their different ways make very clear. Mechanisms of prevention The second bias refers to the focus on the wrong mechanisms and evidence about prevention.45 In public health policy at large, there is a focus on cause of disease and the origins of health inequalities.46e48 There is much less attention paid to the mechanisms of prevention. In essence, the heuristic or cognitive bias is that the assumption is made that if you know the former (the cause), you will be able to do the latter (pre- vention). The empirical fact that the mechanisms of aetiology may be quite different from the mechanisms of prevention (and that this has been known for decades) is seldom brought into sharp focus. This has become a default bias.49 The same decisions are made instead of taking note of the consequences of those decisions previously. This is habitual and almost automatic.50 This collective cognitive bias or heuristic in thinking is understandable. It is an unintended consequence of two great public health advances. First, with the eventual discovery at the end of the nineteenth century that germs present in dirty water could cause disease e and the coincidental realisation that the provision of clean water and sewage disposal was an effective prevention strategy e cause and prevention were coupled together. Certainly, protecting people from insanitary conditions is a highly efficacious public health strategy, but knowing which germs are in the water, does not tell you how to build a sewer. Very importantly, we must remember that the engineers who built and designed the sanitary systems did not do so originally because of knowledge about water-borne microorganisms e they were miasmatists and were trying to improve air quality.51 More generally, knowing of the exis- tence of germs and other microorganisms does not neces- sarily tell you how to prevent their spread. If it did, we would have stopped hospital acquired infections and influenza epi- demics long ago. Second, the discovery that exposure to cigarette smoke causes lung cancer in some smokers and a range of other serious illnesses and the observation that if people stopped smoking, then the risks dropped dramatically also coupled cause and prevention. Once again preventing exposure to cigarette smoke is a highly effective public health strategy, but critically, knowing the statistical association or indeed un- derstanding the biology of the aetiology of lung cancer and heart disease does not tell you how to help people to stop smoking. The latter requires knowledge of a whole range of evidence from other sciences such as the elasticity of the price of tobacco products, the psychology of addiction, the influ- ence of advertising and the way peer-group pressure operates for example. In other words, successful tobacco control and sanitation require a quite different evidence base to that which is about the aetiology of carcinoma of the lung or cholera. Understanding cause is the necessary but not sufficient con- dition e it tells you what to do but not how to do it. There is a gap between what we know about the mechanisms of the causes of the public health dangers facing us and our ability to turn that knowledge into practices that will facilitate how to halt or reverse or slow down the epidemics relating to obesity, alcohol misuse, lack of exercise and common infections. A lopsided evidence base A consequence of this is that the evidence base is lopsided and we continue to make it more lopsided. The evidence is heavily skewed towards details about proximal risk factors for communicable and non-communicable disease, towards elegant expositions of the way the wider determinants of health are associated with patterning of mortality and morbidity at population level and towards details about the effectiveness and cost-effectiveness of an array of in- terventions designed to prevent disease to improve popula- tion health or protect the public from various hazards. The evidence base for translating this information into accurate descriptions of the mechanisms needed for effective preven- tion strategies is scant, and the economic evaluation of these matters underdeveloped. In each of the areas e proximal risk factors, wider determinants and the interventions which are derived from the evidence, the mechanisms at work as against the associations and correlations between the various factors, are unexplored and underdeveloped. Therefore, with respect to proximal risk factors, for example, of alcohol con- sumption and liver disease or of calorie consumption and obesity, the associations are well established e as are the biological mechanisms involved. But, the social mechanisms are much less well described. The variations in why, where, how, when, with whom and for how long people engage in swallowing food or alcohol is obviously enormously variable. That is because eating and drinking alcohol are not single behaviours but are embedded in webs of social practice which are the products of individual human agency, on the one hand, and in the social structures which are the product of human agency and also constrain and limit individual choices, on the other.52 Clearly, there are mechanisms at work e but, for sure, the mechanism will be very different in the many different circumstances in which people eat and drink. Yet, seeking to make sense of these different mechanisms linking proximal risk and outcomes, for the most part, remain unexplored, whereas the default posi- tion is that changing behaviour is the answer.53 That has been the direction of policy for decades, even though the results have been, at best, disappointing. The picture is just as dismal with respect to the patterning of disease contingent on the wider determinants of health. The associations between the coarse categories of SES, in- come, education and gender and patterns of health and illness are very well established. However, what are the actual social mechanisms that link poverty and disadvantage and wealth and health, and what is the nature of the links between the social phenomena and the biological phenomena, is another great chasm in the public health armoury.54 In recent years, developments in epigenetics and metabolomics have begun to demonstrate some of the plausible biological mechanisms https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2 167 involved and the possible transgenerational effects of expo- sure to poverty and other noxious agents in the environ- ment.55 Similarly, historical scholarship demonstrates that these processes have a very strong historically recursive pattern of reproduction over generations. Notwithstanding epigenetic transmission across generations, independently the recursive nature of the consequences of local social and economic arrangement over time reinforces, generation on generation, the disadvantages and advantages of social ar- rangements. Much is known, but the gap between the bio- logical and the social, and more particularly the nature of the mechanism between the two, remains at the level of statisti- cal association, and not mechanistic understanding. Yet, it is a scientific question which John Snow, William Duncan and Rudolf Virchow would surely have recognised. The precise links between the biological and social and how it works are yet to be elucidated. Discussion: implications and challenges for public health economics Aside from the public health community generally, there are some interesting challenges for public health economics generated by these cognitive biases. With respect to popula- tion dynamics and especially with respect to the reductionism and individualism inherent in current thinking about pop- ulations, utility theory is individualistic par excellence. The reliance in cost-utility analysis and on data derived from randomised controlled trials, for example, mean that in a paradigmatic sense, the application of alternative ways of thinking are seldom entertained. The challenge mounted by behavioural economics56 does not really change the para- digm. It simply suggests that individuals may be motivated by a range of factors in the environment and notes that their actions are not driven solely by utility maximisation. This is of course true but is still about individuals. Notwithstanding the success and power of the individualistically based economic theories, there is another very interesting issue in play. There is in utility theory (as noted previously) a recognition of dynamism which when linked to social theory might be very useful. The notion of the market is at the social or supra- individual level of analysis. The ideas of structure and agency at the heart of social practice theory57 strongly resonate with the idea of the market as a supra-individual structure that exists as a consequence, but independent, of individual deci- sion-making.58 The arguments proposed by Etzioni59 in what he called socio-economics, are helpful in this regard. His idea that variegated communities are the expression of agency- structure interaction and that identity and self are critical to understanding people's practices offers an interesting poten- tial theoretical contribution. It might also help to move the arguments in health economics beyond the distinction be- tween efficiency and equity. In agency e structure sociological theory, societies are not conceptualised as inherently equi- table or efficient and neither are efficiency and equity seen as goals that the system itself should or will maintain. Rather efficiency, equity, inefficiency and inequity are emergent properties of the dynamics of the complex system. To draw together the social and economic theory might be a very productive avenue for health economics and would help to overcome the bias described previously about population dy- namics. The writings of the economists Piketty60 and Varou- fakis61 are important too, not because they major on socio- economics but because they get into the detail of complex- ities and mechanisms, although different in each case. Both develop economic ideas in which the dynamism and detail of social systems are paramount and from which public health implications flow.62 Empirically, there are undoubtedly opportunity costs arising a consequence of the way that society or populations are presently conceptualised and the public health policies that flow from that conceptualisation. Of course, wrestling with the problems of heterogeneity has been intrinsic to much thinking across epidemiology and evidence-based medicine, and economic modellers, in particular, have been cognisant of this.63 At its most basic, the fact that the different incremental cost-effectiveness ratios will apply to the same intervention introduced into different segments of the population has been acknowledged and was, for example, intrinsic to the way NICE public health models were constructed. However, this was a theoretical rather than an empirical exercise because mostly there were no data on differential outcomes in different population groups because such data were simply not collected in the primary studies. From a heath economics point of view, so long as the pri- mary studies in the public health population sciences do not operate with the levels of granularity required to describe population dynamics, economic modellers will have to fall back on a different strategy e but this itself may be an op- portunity. The opportunity is to require that economic models take the fullest account of the best social science models about human behaviour. In this regard, in the psychological sciences, the advances being made in researching the mech- anisms of behaviour change and the ontologies of the com- ponents of behaviour change interventions will provide a firm basis on which to develop future economic models.64,65 Similarly, from sociology, the use of social practice theory in which simple determinism is eschewed in favour of theoret- ical understandings of the interactions between structure and agency, and the emergent properties in complex systems of human conduct provide another important platform on which to build models.66 Greater interdisciplinary working between economists, psychologists and sociologists would undoubt- edly help this process along. As Horton put it in a recent editorial in the Lancet ‘public health science needed to pay more attention to the lived experiences of people in societies. Public health needed to recognise the importance of identity, reasoning and voice. Public health today is crudely reduc- tionist, often ignoring or denying the lives of those it purports to defend.’67 The social sciences have much to offer here to improve this state of affairs, and public health economics has a central place potentially. Decision theoretic approaches also offer a potential solu- tion to the biases. Threlfall et al. have argued that the com- bination of robust theory, causal understanding and observation are able to provide sufficient evidence of the di- rection of effect in public health interventions.68 They propose moving from what they call the dominant hypothesis-testing approach that is based on the individualism and reductionism https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2168 in the current evidence base. The same team has argued that in many cases, it is simply not possible to conduct trials that are large enough to capture meaningful effect sizes because the nature of populations is such that the randomised controlled trial (RCT) cannot cope with the heterogeneity in any meaningful way. RCTs pick up individual effects and only get to population effects by aggregating the results e not by conceptualising the population as an entity in its own right.69 A decision theory approach utilises all relevant knowledge, theory and data. Building the links between the psychological and sociological ideas identified previously is a way to help develop the decision theoretic approach. The beauty of such a strategy aside from its comprehensiveness with respect to evidence and theory and its intrinsic interdisciplinary nature is that it releases the thinking from the constraints of the hypothesis-testing approach at the heart of the reductionism attached to RCTs and the individualism of epidemiology and utility theory. In respect of mechanisms, there is much work to do. There are two dimensions to this: first, the mechanisms operating between the social and the biological. As noted previously, while plausible biological mechanisms have been described about the links between social exposures and biological con- sequences and this is a fast-moving scientific arena, the gap is between the social and the biological. Sociologists for the most part have been taciturn on this matter (but see Meloni et al.70), and biologists have been seemingly content to see the social as a factor that kick starts the biological processes without saying with any precision as to when and how it happens.71 This is true even when there are known toxins in the envi- ronment. It is assumed that exposure occurs but how social factors interact to produce biological outcomes is not explored. Perhaps the most productive way to conceptualise the process is one in which physical, biological, social and subjective processes (which are normally treated as analyti- cally distinct) are synthesised into a single complex system and in a dynamic and changing process. Within that process, the connections between social and subjective phenomena and the biological and social still need to be elaborated. And here, utility theory, or that part of it that describes market mechanisms as the outcome of individual decision-making but having an economic reality sui generis, can be conjoined with structure and agency in the social theories as suggested previously. In other words, it is out of the very repetitive na- ture of social and economic life, the social and economic practices in which people engage in their everyday lifeworlds that offer a clue. This is an exciting possibility for future development as mentioned previously. Second, the links along the causal chains might be given some further attention. In general, even in the most sophis- ticated economic models, the mechanisms in complex causal chains are treated in a predictive way and as existing in time and space. This is the sense that event A at time T1 has con- sequences B at time T2; T2 occurs chronologically after T1. This allows the prediction that if A happens, then B will also likely occur. Most public health interventions are premised on this assumption, and in a world governed by Newtonian physics, that is a reasonable assumption. However, it is also perfectly possible to flip the coin metaphorically and to think about the question the other way round. In everyday life, most of us think the other way round a lot of the time. Therefore, when we ask the question why did something happen, at time T, we generally commonsensically answer that question by attributing it to a preceding causal event that happened in the past. We then may ask another question which is why did that preceding event occur. And, we could continue to ask why the event that preceded it occurred and so on in a potentially never-ending process of regress. However, this way of thinking can actually be very helpful in model building because it allows us to disassemble phe- nomenon and see them for the multiple complexities that they may involve. This form of regressive forensic inference can be very helpful as we build and construct models and may be very useful in terms of economic thinking. Prediction can indeed be very seductive, although in the social sciences, it is often highly inaccurate. However, if we think about our models or build our models in this way and ask the question about the dynamic economic process at each of the stages, this may prove to be very fruitful. It would certainly assist in articulating the mechanisms of prevention and examining the economic processes at each of the stages. The theoretical and method- ological means of articulating those mechanisms have been described and provide a useful platform on which to build.72 All of this would of course help to get beyond the problem of the skewed evidence base as we build the knowledge base in the arenas where the evidence cupboard is presently pretty bare. Novel economic thinking might also help to get past the cognitive biases and the cognitive dissonance. But, this comes with a health warning. The policy arena is much more than a set of cognitive biases and government failure and ineptitude, unanticipated consequences, internal and external power relations, vested interests, incentives and disincentives in the system all play their parts and in ways that no single agent has the ability, knowledge and wisdom to know.73 Therefore, even if we can overcome the cognitive biases, these other realities of political and policy making life will not go away, although along with ways of implementing policy, they should form part of the evidence as we seek to gain leverage. Conclusion A number of writers have developed ideas that resonate with the arguments here. Rychetnik et al.74 called for the recogni- tion of the importance of evidence about how something should be done and argued that would include information about design, implementation and context and how the intervention was received. Brownson et al.75,76 have noted that the evidence about what should be done based on risk is extensive but also stresses the importance of evidence about the relative effectiveness of specific interventions. Taking our cue from this, we must obviously continue to build the evi- dence for the necessary conditions for effective prevention, i.e., the evidence about aetiology, proximal risk factors and cost- effectiveness. However, we must also build the evidence base about the sufficient conditions for prevention. This is the evidence about how to create population level changes, how to implement policies that benefit all segments of the popu- lation and how to implement policies in ways that recognise how interventions are experienced (and resisted) by different communities. With the exception of tobacco control policy e https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2 169 which incorporates evidence about sufficient conditions such as the price mechanism, the psychology of quitting and addiction and how to manage misinformation from big to- bacco etc. e policy in many areas, including dementia, obesity, alcohol and physical activity, has been implemented sometimes without robust evidence but more usually with little consideration of the evidence about the mechanisms likely to produce optimal effectiveness. Economics needs to be front and centre in working with psychology and sociology in particular to develop new ways of thinking and new models to make this evidence clear. In practical terms, this means that research funders should devote more resources to develop the evidence about the mechanisms of prevention,77 and they should rebalance the portfolio away from aetiology and risk.78 This would involve a focus on the politics of implementation for example and in particular on the power struggles that go on once the evidence of causation and risk have been established. It would involve building the evidence about how complexities of real-world systems (not imagined simple linear ones) operate. Practi- cally, the argument also means that policy makers should become much more cognisant than they usually are, of the cognitive biases which with which they themselves operate. It is interesting to note that most government officials are acutely aware of, and sensitive to, the political nuances of every action they take and propose. But, in my experience, they are for the most part blindsided to the cognitive pro- cesses involved in their own thinking. And, my own tribe, academic researchers, should similarly not imagine the job is done once research findings are published. Frequently what happens now is that authors complete their work and leave it at that, hoping others will take up the baton, assuming that it is self-evident from their published work what ought to happen. More actively minded colleagues, however, some- times get involved in advocacy. This is seldom an effective strategy as governments, decision-makers and industry are mostly impervious to academic advocacy, and in any event, telling people that they are wrong is seldom an effective way to change their behaviour. Instead, the political, social and economic processes that affect the ways that findings are received, interpreted, acted upon and implemented, in the way that Smith79 has described, should be the subject of much greater academic and scientific scrutiny, than is currently the case. These findings should then be used to gain the leverage that is needed. And, it is not as if we do not know this as an academic community, it is just that we choose not to do it, preferring, very often, political gesturing, rather than applying our scientific skills to the processes that we seek to influence. Finally, as noted previously, economists and sociologists have long understood the dynamics of complexity, although most recent accounts of complexity pay almost no attention to this classic literature. But, there is clearly an opportunity for so- ciologist and economists to work together in that conceptual and theoretical space of markets and structure and agency. We need a strategic research programme to develop this, and a consilience between the disciplines should be encouraged. Furthermore, together, the theoretical insights of the two disciplines will, if applied properly, help shift the rather vacuous rhetoric that currently surrounds notions of complexity to replace it with some properly social scientific empirical and theoretical work. The biases described here are no doubt to be found in many other fields of policy and indeed in other jurisdictions. Public health is not unique in this regard. However, it does seem likely to me that those government departments have a longer his- tory of grappling with evidence, and here I single out transport and health in England as exemplars of good practice, will be more likely to able to get into an understanding of the bias problems than those departments where down the years, evi- dence has been less prominent in the decision-making process. Author statements Ethical approval None sought. Funding This work was supported by the annual fund of St John's College, Cambridge and by the Arts and Humanities Research Council (AHRC) (UK) (Grant number AH/M005917/1) (‘Evalu- ating Evidence in Medicine’). Competing interests The author is in receipt of grant funding for public healtherelated research from MRC, ESRC, the Wellcome Trust and NIHR. He also has one consultancy for providing general evidence-based advice on obesity prevention to Slimming World. From 2005 to 2014, he was the Director of the Centre for Public Health at NICE. r e f e r e n c e s 1. Higher Education Funding Council. Research excellence framework 2014: the results. Bristol: Higher Education Funding Council; 2014. 2. Tavris C, Aronson E. Mistakes were made (But Not By Me): why we justify foolish beliefs, bad decisions, and hurtful acts. Boston MA: Mariner Books; 2015. 3. Marteau TM, Hollands GJ, Kelly MP. Changing population behavior and reducing health disparities: Exploring the potential of “choice architecture” interventions. In: Kaplan RM, Spittel M, David DH, editors. Population health: behavioral and social science insights, AHRQ publication No. 15- 0002. Rockville, MD: Agency for Healthcare Research and Quality and Office of Behavioral and Social Sciences Research, National Institutes of Health; 2015. p. 105e26. 4. Kelly MP, Barker M. Why is changing health related behaviour so difficult? Publ Health 2016;136:109e16. https://doi.org/10. 1016/j.puhe.2016.03.030. 5. Kahneman D. Thinking, fast and slow. New York: Farrar, Strauss & Giroux; 2011. 6. Kelly MP, Russo F. Causal narratives in public health: the difference between mechanisms of aetiology and mechanisms of prevention in non-communicable diseases. Sociol Health Illness 2018;40(1):82e99. http://onlinelibrary. wiley.com/doi/10.1111/1467-9566.12621/pdf. http://refhub.elsevier.com/S0033-3506(19)30036-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref2 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref2 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref2 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref3 https://doi.org/10.1016/j.puhe.2016.03.030 https://doi.org/10.1016/j.puhe.2016.03.030 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref5 http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2170 7. National Audit Office. Getting the evidence: using research in policy making. London: HMSO; 2003. 8. Davies HTO, Nutley SM, Smith PC. What works? evidence based policy and practices in public services. Bristol: Policy Press; 2000. 9. Cooksey D. A review of UK health research funding. London: HM Treasury; 2006. 10. Public Health England. Knowledge strategy: harnessing the power of information to improve the public's health. London: Public Health England; 2014. 11. Holmes B, Best A, Davies H, Hunter D, Kelly MP, Marshall M, Rycroft- Malone J. Mobilizing knowledge in complex health systems: A call to action. Evidence and Policy 2017;13(3):539e60. https://doi.org/10.1332/174426416X14712553750311. 12. Ogilvie D, Craig P, Griffin S, Macintyre S, Wareham NJ. A translational framework for public health research. BMC Public Health 2009;9:116. https://doi.org/10.1186/1471-2458-9-116. 13. Orton L, Lloyd-Williams F, Taylor-Robinson D, O'Flaherty M, Capewell S. The use of research evidence in public health decision making processes: systematic review. PLoS One 2011;6(7):e21704. 14. Kingdon J W Agendas. Alternatives and public policies. Boston, MA: Little Brown; 1984. 15. Higher Education Funding Council. Research excellence framework 2014: The Results. Bristol: Higher Education Funding Council; 2014. 16. Public Health England. Doing, supporting and using public health research. London: Public Health England; 2015. 17. Cochrane AL. Effectiveness and efficiency: random reflections on health services. London: British Medical Journal/Nuffield Provincial Hospitals Trust; 1972. 18. Orton L, Lloyd-Williams F, Taylor-Robinson D, O'Flaherty M, Capewell S. The use of research evidence in public health decision making processes: systematic review. PLoS One 2011;6(7):e21704. 19. Smith K. Beyond evidence based policy in public health: the interplay of ideas. Basingstoke, UK: Palgrave Macmillan; 2013. 20. Hayek FA. The use of knowledge in society. Am Econ Rev 1945;35(4):519e30. 21. Giddens A. The constitution of society: outline of the theory of structuration. Cambridge: Polity; 1984. 22. Giddens A. The constitution of society: outline of the theory of structuration. Cambridge: Polity; 1984. 23. McMichael AJ. Prisoners of the proximate: loosening the constraints on epidemiology in an age of change. Am J Epidemiol 1999;149:887e97. 24. Horton R. Apostasy against the public health elites. Lancet 2018;391:643. February 17, 2018, http://www.thelancet.com/ pdfs/journals/lancet/PIIS0140-6736(18)30304-0 . 25. Kriznik NM, Kinmonth AL, Ling T, Kelly MP. Moving beyond individual choice in policies to reduce health inequalities: the integration of dynamic with individual explanations. J Public Health 2018;40(4):764e75. https://doi.org/10.1093/pubmed/ fdy045. https://academic.oup.com/jpubhealth/advance-article/ doi/10.1093/pubmed/fdy045/4931230?guestAccessKey¼ af9f5249-b3b7-4270-92db-421e9c8fb5ac. 26. Kelly MP. The axes of social differentiation and the evidence base on health equity. J R Soc Med 2010;103:266e72. https:// doi.org/10.1258/jrsm.2010.100005. 27. Savage M, Cunningham N, Devine F, Friedman S, Laurison D, McKenzie L, Miles A, Snee H, Wakeling P. Social class in the twenty first century. London: Pelican; 2015. 28. Tugwell P, Petticrew M, Kristjansson EA, Welch V, Ueffing E, Waters E, Bonnefoy J, Morgan A, Doohan E, Kelly MP. Assessing equity in systematic reviews: realising the recommendations of the Commission on Social Determinants of Health. Br Med J 2010;341:873e7. https:// doi.org/10.1136/bmj.c4739. 341:c4739, http://bit.ly/9lUV7k. 29. Kriznik NM, Kinmonth AL, Ling T, Kelly MP. Moving beyond individual choice in policies to reduce health inequalities: the integration of dynamic with individual explanations. J Public Health 2018;40(4):764e75. https://doi.org/10.1093/pubmed/ fdy045. https://academic.oup.com/jpubhealth/advance-article/ doi/10.1093/pubmed/fdy045/4931230?guestAccessKey¼ af9f5249-b3b7-4270-92db-421e9c8fb5ac. 30. Kelly MP. The individual and the social level in public health. In: Killoran A, Kelly MP, editors. Evidence based public health: effectiveness and efficiency. Oxford: Oxford University Press; 2010. p. 425e35. 31. Rutter H, Savona N, Glonti K, Bibby J, Cummins S, Finegood D, Greaves F, Harper L, Hawe P, Moore L, Petticrew M, Rehfuess E, Shiell A, Thomas J, White M. The need for a complex systems model of evidence for public health. Lancet 2017;390(10112):2602e4. https://doi.org/10.1016/S0140- 6736(17)31267-9. 32. Sniehotta FF, Araujo-Soares V, Brown J, Kelly MP, Michie S, West R. Complex systems and individual-level approaches to population health: a false dichotomy? Lancet Public Health September 2017;2. www.thelancet.com/public-health. http:// www.thelancet.com/pdfs/journals/lanpub/PIIS2468-2667(17) 30167-6 . 33. Craig P, Dieppe P, Macintyre S, Michie S, Nazareth I, Petticrew M. Developing and evaluating complex interventions: new guidance. London: Medical Research Council; 2008. http:// www.mrc.ac.uk/complexinterventionsguidance. 34. Campbell M, Fitzpatrick R, Haines A, Kinmonth AL, Sandercock P, Spiegelhalter D. Tyrer. Framework for design and evaluation of complex interventions to improve health. Br Med J 2000;321:694e6. 35. Moore, G, G., Audrey, S., Barker, M., Bond L., Bonell, C., Hardeman W., Moore, L., O'Cathain, A., Tinati, T. Wight, D., Baird, J. Process evaluation of complex interventions, UK Medical Research Council (MRC) guidance. Prepared on behalf of the MRC Population Health Science Research Network (nd). http://www.epc-src.org/src/srcDocuments/Workgroups/ 22129/Session1_3_MRC%20guidance . 36. Festinger L, Riecken HW, Schacter S. When prophecy fails: a social and psycholgical study ofa modern group that predicted the destruction of the world. Minnesota: University of Minnesota Press; 1956. 37. Engel GL. A unified concept of health and disease. Perspect Biol Med 1960;3:459e85. 38. Rutter H, Savona N, Glonti K, Bibby J, Cummins S, Finegood D, Greaves F, Harper L, Hawe P, Moore L, Petticrew M, Rehfuess E, Shiell A, Thomas J, White M. The need for a complex systems model of evidence for public health. Lancet 2017;390(10112):2602e4. https://doi.org/10.1016/S0140- 6736(17)31267-9. 39. Kelly MP, Barker M. Why is changing health related behaviour so difficult? Publ Health 2016;136:109e16. https://doi.org/10. 1016/j.puhe.2016.03.030. 40. Marteau TM, Hollands GJ, Kelly MP. Changing population behavior and reducing health disparities: Exploring the potential of “choice architecture” interventions. In: Kaplan RM, Spittel M, David DH, editors. Population health: behavioral and social science insights, AHRQ publication No. 15- 0002. Rockville, MD: Agency for Healthcare Research and Quality and Office of Behavioral and Social Sciences Research, National Institutes of Health; 2015. p. 105e26. 41. Parkkinen V-P, Wallmann C, Wilde M, Clarke B, Illari P, Kelly, Michael P, Norell Charlie, Russo Federica, Shaw Beth, Williamson Jon. Evaluating evidence of mechanisms in medicine: principles and procedures. London: Springer; 2018xvii 125. https://www.springer.com/gb/book/9783319946092 https:// link.springer.com/book/10.1007/978-3-319-94610-8. 42. Russo F, Williamson J. Interpreting causality in the health sciences. Int Stud Philos Sci 2017;21(2):157e70. http://refhub.elsevier.com/S0033-3506(19)30036-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref10 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref10 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref10 https://doi.org/10.1332/174426416X14712553750311 https://doi.org/10.1186/1471-2458-9-116 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref13 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref13 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref13 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref13 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref14 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref14 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref19 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref19 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref21 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref21 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref23 http://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(18)30304-0 http://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(18)30304-0 https://doi.org/10.1093/pubmed/fdy045 https://doi.org/10.1093/pubmed/fdy045 https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://doi.org/10.1258/jrsm.2010.100005 https://doi.org/10.1258/jrsm.2010.100005 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref27 https://doi.org/10.1136/bmj.c4739 https://doi.org/10.1136/bmj.c4739 http://bit.ly/9lUV7k https://doi.org/10.1093/pubmed/fdy045 https://doi.org/10.1093/pubmed/fdy045 https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac http://refhub.elsevier.com/S0033-3506(19)30036-8/sref30 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref30 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref30 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref30 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref30 https://doi.org/10.1016/S0140-6736(17)31267-9 https://doi.org/10.1016/S0140-6736(17)31267-9 http://www.thelancet.com/public-health http://www.thelancet.com/pdfs/journals/lanpub/PIIS2468-2667(17)30167-6 http://www.thelancet.com/pdfs/journals/lanpub/PIIS2468-2667(17)30167-6 http://www.thelancet.com/pdfs/journals/lanpub/PIIS2468-2667(17)30167-6 http://www.mrc.ac.uk/complexinterventionsguidance http://www.mrc.ac.uk/complexinterventionsguidance http://refhub.elsevier.com/S0033-3506(19)30036-8/sref34 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref34 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref34 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref34 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref34 http://www.epc-src.org/src/srcDocuments/Workgroups/22129/Session1_3_MRC%20guidance http://www.epc-src.org/src/srcDocuments/Workgroups/22129/Session1_3_MRC%20guidance http://refhub.elsevier.com/S0033-3506(19)30036-8/sref36 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref36 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref36 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref37 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref37 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref37 https://doi.org/10.1016/S0140-6736(17)31267-9 https://doi.org/10.1016/S0140-6736(17)31267-9 https://doi.org/10.1016/j.puhe.2016.03.030 https://doi.org/10.1016/j.puhe.2016.03.030 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref40 https://www.springer.com/gb/book/9783319946092%20https://link.springer.com/book/10.1007/978-3-319-94610-8 https://www.springer.com/gb/book/9783319946092%20https://link.springer.com/book/10.1007/978-3-319-94610-8 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref42 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref42 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref42 https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2 171 43. Kriznik NM, Kinmonth AL, Ling T, Kelly MP. Moving beyond individual choice in policies to reduce health inequalities: the integration of dynamic with individual explanations. J Public Health 2018;40(4):764e75. https://doi.org/10.1093/pubmed/ fdy045. https://academic.oup.com/jpubhealth/advance-article/ doi/10.1093/pubmed/fdy045/4931230?guestAccessKey¼ af9f5249-b3b7-4270-92db-421e9c8fb5ac. 44. Marteau TM, Hollands GJ, Kelly MP. Changing population behavior and reducing health disparities: Exploring the potential of “choice architecture” interventions. In: Kaplan RM, Spittel M, David DH, editors. Population health: behavioral and social science insights, AHRQ publication No. 15- 0002. Rockville, MD: Agency for Healthcare Research and Quality and Office of Behavioral and Social Sciences Research, National Institutes of Health; 2015. p. 105e26. 45. Kelly MP, Russo F. Causal narratives in public health: the difference between mechanisms of aetiology and mechanisms of prevention in non-communicable diseases. Sociol Health Illness 2018;40(1):82e99. http://onlinelibrary. wiley.com/doi/10.1111/1467-9566.12621/pdf. 46. McMichael AJ. Prisoners of the proximate: loosening the constraints on epidemiology in an age of change. Am J Epidemiol 1999;149:887e97. 47. Marmot M, Wilkinson R, editors. Social determinants of health. 2nd ed. Oxford: Oxford University Press; 2006. 48. Wilkinson R, Pickett K. The spirit level: why equality is better for everyone. London: Allen Lane; 2009. 49. Samuelson W, Zeckhauser R. Status quo bias in decision making. J Risk Uncertain 1988;1:7e59. 50. Kahneman D. Thinking, fast and slow. New York: Farrar, Strauss & Giroux; 2011. 51. Kelly MP, Russo F. Causal narratives in public health: the difference between mechanisms of aetiology and mechanisms of prevention in non-communicable diseases. Sociol Health Illness 2018;40(1):82e99. http://onlinelibrary. wiley.com/doi/10.1111/1467-9566.12621/pdf. 52. Blue S, Shove E, Carmona C, Kelly MP. Theories of practice and public health: understanding (un) healthy practices. Crit Public Health 2016;26:36e50. https://doi.org/10.1080/09581596. 2014.980396. https://doi.org/10.1080/09581596.2014.980396. 53. Marteau TM, Hollands GJ, Kelly MP. Changing population behavior and reducing health disparities: Exploring the potential of “choice architecture” interventions. In: Kaplan RM, Spittel M, David DH, editors. Population health: behavioral and social science insights, AHRQ publication No. 15- 0002. Rockville, MD: Agency for Healthcare Research and Quality and Office of Behavioral and Social Sciences Research, National Institutes of Health; 2015. p. 105e26. 54. Kelly MP, Kelly R, Russo F. The integration of social, behavioural and biological mechanisms in models of pathogenesis. Perspect Biol Med 2014;57:308e28. 55. Kelly MP, Kelly RS. Quantifying social influences throughout the life course: action, structure and ‘omics’. In: Meloni M, Cromby J, Fitzgerald P, Lloyd S, editors. The palgrave handbook of biology and society. Basingstoke: Palgrave Macmillan; 2018. p. 587e609. 56. Thaler RH, Sunstein CR. Nudge: improving decisions about health. Wealth and Happiness New Haven, CT: Yale University Press; 2008. 57. Shove E, Pantzar M, Watson M. The dynamics of social practice: everydaylife and how it changes. London: Sage; 2012. 58. Hayek FA. The use of knowledge in society. Am Econ Rev 1945;35(4):519e30. 59. Etzioni A. The moral dimension: towards a new economics. New York: Free Press; 1988. 60. Piketty T. Capital in the twenty-first century. Cambridge, Mass: Belknap Press; 2004. 61. Varoufakis Y. The global minotaur: America, Europe and the future of the global economy. New edition. London: Zed Books; 2015. first published 2011.; a)Varoufakis Y. And the weak suffer what they must? Europe, austerity and the threat to Global stability. London: Bodley Head; 2016. 62. Kelly MP. Understanding the mechanisms underpinning health inequalities: lessons from economics. Global Health Promotion 2016;23:3e5. 63. Squires H, Chilcott J, Akehurst R, Burr J, Kelly MP. A systematic literature review of the key challenges for developing the structure of Public Health economic models. Int J Public Health 2016;61:289e98. 64. Michie S, Carey R, Johnston M, Rothman A, de Bruin M, Kelly MP, Connell L. From theory-inspired to theory-based interventions: A protocol for developing and testing a methodology for linking behaviour change techniques to theoretical mechanisms of action. Ann Behav Med 2016. https://doi.org/10.1007/s12160-016-9816-6. http://rdcu.be/nitY https://www.researchgate.net/publication/305218366_From_ Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_ for_Developing_and_Testing_a_Methodology_for_Linking_ Behaviour_Change_Techniques_to_Theoretical_ Mechanisms_of_Action. 65. Michie, S., Thomas, J., Johnston, M., Mac Aonghusa, P., Shawe-Taylor, J., Kelly, M.P., Deleris, L.A., Finnerty, A.N., Marques, M.M., Norris, E., O'Mara-Eves, A., West, R. The Human Behaviour-Change Project: Harnessing the power of Artificial Intelligence and Machine Learning for evidence synthesis and interpretation, Implement Sci. 12:121 DOI 10.1186/s13012-017-0641-5 http://rdcu.be/wRUc. 66. Shove E, Pantzar M, Watson M. The dynamics of social practice: everydaylife and how it changes. London: Sage; 2012. 67. Horton R. Apostasy against the public health elites. Lancet 2018;391(February 17):643. 2018, http://www.thelancet.com/ pdfs/journals/lancet/PIIS0140-6736(18)30304-0 . 68. Threlfall A, Meah S, Fischer AJ, Cookson R, Rutter H, Kelly MP. The appraisal of public health interventions: the use of theory. J Public Health 2015;37:166e71. http://jpubhealth. oxfordjournals.org/content/37/1/166.full þhtml. 69. Fischer AJ, Threlfall A, Meah S, Cookson R, Rutter H, Kelly MP. The appraisal of public health interventions: an overview. J Public Health 2013;35:488e94. http://jpubhealth. oxfordjournals.org/cgi/content/full/fdt076? ijkey¼W0OiEW3vvUj0jgR&keytype¼ref. 70. Meloni M, Cromby J, Fitzgerald P, Lloyd S. The palgrave handbook of biology and society. Basingstoke: Palgrave Macmillan; 2018. 71. Landecker H, Panofsky A. From social structure to gene regulation, and back: a critical introduction to environmental epigenetics for sociology. Annu Rev Sociol 2013;39:333e57. 72. Parkkinen V-P, Wallmann C, Wilde M, Clarke B, Illari P, Kelly MP, Norrell C, Russo F, Shaw B, Williamson J. Evaluating evidence of mechanisms in medicine: principles and procedures. London: Springer; 2018. 73. Hayek FA. The use of knowledge in society. Am Econ Rev 1945;35(4):519e30. 74. Rychetnik L, Hawe P, Waters E, Barratt A, Frommer M. A glossary for evidence based public health. J Epidemiol Community Health 2004;58:538e45. https://doi.org/10.1136/ jech.2003.011585. http://jech.bmj.com/content/jech/58/7/538. full . 75. Brownson RC, Gurney JG, Land GH. Evidence-based decision making in public health. J Publ Health Manag Pract 1999;5:86e97. 76. Brownson RC, Baker EA, Leet TL, et al. Evidence-based public health. Oxford: Oxford University Press; 2003. p. 7. https://doi.org/10.1093/pubmed/fdy045 https://doi.org/10.1093/pubmed/fdy045 https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac https://academic.oup.com/jpubhealth/advance-article/doi/10.1093/pubmed/fdy045/4931230?guestAccessKey=af9f5249-b3b7-4270-92db-421e9c8fb5ac http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref44 http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://refhub.elsevier.com/S0033-3506(19)30036-8/sref46 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref46 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref46 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref46 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref47 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref47 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref48 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref48 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref49 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref49 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref49 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref50 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref50 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref50 http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf https://doi.org/10.1080/09581596.2014.980396 https://doi.org/10.1080/09581596.2014.980396 https://doi.org/10.1080/09581596.2014.980396 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref53 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref54 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref54 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref54 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref54 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref55 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref56 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref56 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref56 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref57 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref57 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref58 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref58 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref58 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref59 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref59 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref60 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref60 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref61 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref61 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref61 http://refhub.elsevier.com/S0033-3506(19)30036-8/bib61a http://refhub.elsevier.com/S0033-3506(19)30036-8/bib61a http://refhub.elsevier.com/S0033-3506(19)30036-8/bib61a http://refhub.elsevier.com/S0033-3506(19)30036-8/sref62 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref62 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref62 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref62 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref63 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref63 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref63 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref63 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref63 https://doi.org/10.1007/s12160-016-9816-6 http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/nitY%20https://www.researchgate.net/publication/305218366_From_Theory-Inspired_to_Theory-Based_Interventions_A_Protocol_for_Developing_and_Testing_a_Methodology_for_Linking_Behaviour_Change_Techniques_to_Theoretical_Mechanisms_of_Action http://rdcu.be/wRUc http://refhub.elsevier.com/S0033-3506(19)30036-8/sref66 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref66 http://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(18)30304-0 http://www.thelancet.com/pdfs/journals/lancet/PIIS0140-6736(18)30304-0 http://jpubhealth.oxfordjournals.org/content/37/1/166.full +html http://jpubhealth.oxfordjournals.org/content/37/1/166.full +html http://jpubhealth.oxfordjournals.org/content/37/1/166.full +html http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://jpubhealth.oxfordjournals.org/cgi/content/full/fdt076?ijkey=W0OiEW3vvUj0jgR&keytype=ref http://refhub.elsevier.com/S0033-3506(19)30036-8/sref70 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref70 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref70 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref71 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref71 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref71 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref71 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref72 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref72 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref72 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref72 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref73 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref73 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref73 https://doi.org/10.1136/jech.2003.011585 https://doi.org/10.1136/jech.2003.011585 http://jech.bmj.com/content/jech/58/7/538.full http://jech.bmj.com/content/jech/58/7/538.full http://refhub.elsevier.com/S0033-3506(19)30036-8/sref75 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref75 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref75 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref75 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref76 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref76 https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 6 3 e1 7 2172 77. Kelly MP, Russo F. Causal narratives in public health: the difference between mechanisms of aetiology and mechanisms of prevention in non-communicable diseases. Sociol Health Illness 2018;40(1):82e99. http://onlinelibrary. wiley.com/doi/10.1111/1467-9566.12621/pdf. 78. Kelly M. Foreword UKCRC public health research centres of excellence final report 2018. London, Belfast, Edinburgh: UK Clinical Research Collaboration; 2018. 79. Smith K. Beyond evidence based policy in public health: the interplay of ideas. Basingstoke, UK: Palgrave Macmillan; 2013. http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://onlinelibrary.wiley.com/doi/10.1111/1467-9566.12621/pdf http://refhub.elsevier.com/S0033-3506(19)30036-8/sref78 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref78 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref78 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref79 http://refhub.elsevier.com/S0033-3506(19)30036-8/sref79 https://doi.org/10.1016/j.puhe.2019.02.012 https://doi.org/10.1016/j.puhe.2019.02.012 Cognitive biases in public health and how economics and sociology can help overcome them Introduction Methods Results: systematic biases in the evidence base What is known about populations and the way public health policy and practice generally describe populations Mechanisms of prevention A lopsided evidence base Discussion: implications and challenges for public health economics Conclusion Author statements Ethical approval Funding Competing interests References Co-payments-for-emergency-department-visits--a-quasi-experim_2019_Public-Hea ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 8 Available online at w Public Health journal homepage: www.elsevier.com/puhe Original Research Co-payments for emergency department visits: a quasi-experimental study P. Petrou a,*, D. Ingleby b a Health Economics, Public Health Program, Department of Health Sciences, European University, Nicosia, Cyprus b Centre for Social Science and Global Health, University of Amsterdam, the Netherlands a r t i c l e i n f o Article history: Received 3 June 2018 Received in revised form 11 December 2018 Accepted 19 December 2018 Available online 26 February 2019 Keywords: Emergency department Co-payment Non-avoidable visits Potentially avoidable visits Cyprus * Corresponding author. Tel.: þ35799587597; E-mail addresses: panayiotis.petrou@st.o https://doi.org/10.1016/j.puhe.2018.12.014 0033-3506/© 2018 The Royal Society for Publ a b s t r a c t Objectives: Financial recession in Cyprus has led to health reforms to promote efficiency and reduce public expenditure. In this context, a co-payment fee was introduced in 2013 for all emergency department (ED) visits, with the aim of reducing potentially avoidable visits. The objective of this study was to assess the short-term intended and unintended impacts of introducing these co-payments. Study design: The study design is an interrupted time series analysis. Methods: We used an autoregressive integrated moving average model for interrupted time series analysis of data on ED visits over 42 consecutive months, from 2013 to 2015 in a regional hospital in Cyprus. The ED visits have been classified to non-avoidable and potentially avoidable visits. Results: The introduction of co-payment had no effect on non-avoidable visits (4% [95% confidence interval {CI}: 4.3e11.08] P ¼ 0.694). However, it had the immediate and sustained effect of reducing potentially avoidable visits, an effect that was statistically significant from the first month onwards (29.8% [95% CI: 22.6e34.1] P < 0.00001). Conclusions: Co-payments can be a valuable tool for reducing potentially avoidable emer- gency department visits, without adversely impacting non-avoidable visits. This is a particularly significant finding for countries experiencing financial pressures and struggling to reduce waste in health expenditure. However, the long-term impact of this policy must be assessed, including potential negative effects on public health, to make sure it does not create barriers in obtaining necessary health care that might actually increase expenses in the long run. In particular, timely access to primary care services must be safeguarded. © 2018 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. Introduction In the operational framework of a health system, emergency departments (EDs) are designed to deal with non-avoidable cases, such as major injuries and other situations where fax: þ35722875021. uc.ac.cy, p.petrou@extern ic Health. Published by E ambulatory care is not appropriate, such as those related to alcohol and drugs.1 The inherently unpredictable nature of these cases implies that an ED must be able to cope with a multitude of diverse and possibly life-threatening conditions. This presupposes that EDs are staffed with multidisciplinary al.euc.ac.cy (P. Petrou). lsevier Ltd. All rights reserved. mailto:panayiotis.petrou@st.ouc.ac.cy mailto:p.petrou@external.euc.ac.cy http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2018.12.014&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 8 51 teams including specialists and are equipped accordingly. These attributes render EDs prone to overuse by patients.2,3 Disincentives to using EDs in the form of co-payments at the point of treatment have been applied in several coun- tries.4,5 The co-payment, as a demand-side cost containment measure, has proved its efficacy in reducing excessive utili- zation of health care by inducing a trade-off between the welfare gains from health-care provision and the welfare losses due to waste.6e11 The ability of co-payment to reach an equilibrium between deterring overuse and not discouraging necessary care- by comprising a barrier- is key to its operational scope.11e13 More specifically, in the context of EDs, co-payment encour- ages patients to undertake a kind of self-triage to determine whether visiting an ED is really necessary and appropriate. Cyprus Cyprus is a unique case among European Union member states, because it is the only one without a national system providing universal health coverage.14,15 The current health- care system consists of two fragmented and uncoordinated components, the private and public sector. The private health sector is completely financed by out-of-pocket (OOP) pay- ments, which can be subsidised by voluntary private insur- ance, while the public health care sector provides free health care (to almost 85% of the total population).15 Beneficiaries of the public sector include people who satisfy one of several socioeconomic criteria and people with certain chronic conditions.15,16 Until 2013, Cyprus was one of the few countries in which the public health-care sector did not apply co-payments. Up to that date, EDs operated in all six public hospitals in Cyprus and anybody could use them free of charge, regardless of their eligibility for public-sector care. Some private hospitals have only basic ED units, capable of coping with only minor cases and not offering the breadth and scope of EDs in public hos- pitals, while charges apply. As a result, the ED services mostly used are the public ones. Following an assessment of the public health-care sector in 2013, in the context of a memorandum of understanding concerning the financial crisis in Cyprus, a technical commit- tee concluded that EDs in Cyprus are overused.16,17 This is considered to be a repercussion of long-waiting lists in the public healthcare sector, in tandem with financial crisis and the austerity measures, that further impeded access. People also assign blame to public sector's time-table, which excludes afternoon, public holidays and weekend shifts. This is further exacerbated by the lack of a national coverage health system, which cascades to the existence of many uninsured patients, for which the EDs emerge as the only pathway of access to health care.16e18 Finally, the absence of auxiliary measures such as call centres, as well as lack of collaboration with pri- mary health-care centres, hinders referral of minor cases to day care settings, which could extenuate the burden on the ED. It is estimated that ED visits cost four times more than pri- mary health service visits for the same treatment, leading to productive inefficiency.19e22 Crowding of EDs can more than double the waiting times, creating risks for people genuinely in need of urgent care.23,24 A Canadian study reported that reduction of ED waiting times by 1 h would reduce mortality in a high-risk patient cohort by 6.5%, while the corresponding mortality reduction in a low-risk patient cohort would be 13%.24 In August 2013, the Cyprus Ministry of Health introduced a co-payment of 10 euros, as a prerequisite of the bail-out agreement for all ED visits, applying to all users.25 The aim of this article is to explore the effect of co-payments on ED use in Cyprus. Specifically, we want to establish the separate effects of co-payment on non-avoidable versus potentially avoidable ED visits. Our approach involves testing the hypothesis that co- payment reduced potentially avoidable visits to the ED, without deterring non-avoidable ones. We used daily data covering the period January 2011 to June 2014 from one regional hospital. Because the economic situation of Cyrus was not stable during this period, we also controlled for underlying trends that might have been caused by economic fluctuations. This study further advances relevant literature, and it builds on a recent publication by Petrou,18 who reported on the co- payment on total number of ED visits, without commenting ontheirunderlying emergent (or not) classification (seeTable 1). Methods Data Statistics were collected on visits to Paphos public ED between January 2011 and May 2014. Paphos Hospital is located in the western part of Cyprus and offers health care to approximately 150,000 inhabitants. The hospital has 150 beds, and every year around 8000 patients are admitted. Its outpatient clinics serve approximately 160,000 cases every year, while an estimated 65,000 patients visited the ED yearly.26 No other public EDs operate in this region, so we may assume that these data captured the full impact of the policy change under assess- ment. The private sector was excluded because co-payments were only introduced in public EDs. We also assume that people did not opt for private health services, because this would perpetuate to a full OOP expenditure, exceeding by far the corresponding co-payment fee, and further burden the already crisis-stricken family disposable income. In the public ED, the co-payment fee applies to everybody (both beneficiaries and non-beneficiaries of public-sector care and with or without a referral). The fee is payable at point of care. Visits were classified by two health-care professionals into potentially avoidable and non-avoidable, on the basis of available algorithms.27e29 The rationale was to distinguish visits that should be treated within the ED and visits that potentially could be referred to a Primary Healthcare center (PHC) without posing any threat to patient's health status. Consequently, the definition of the potentially avoidable visits entails the following: � cases that could be dealt with within 12 h such as derma- tological infections, abrasions and sore throat; � conditions that could be addressed properly at primary health-care centres without compromising the safety and comfort of the patient and conditions that do not require special equipment; https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 Table 1 e STROBE statementdchecklist of items that should be included in reports of observational studies. Item no Recommendation Title and abstract 1 (a) Interrupted time series analysis (b) The abstracts provides an informative and balanced summary of what was done and what was found Introduction Background/rationale 2 The fiscal crisis and the bailout agreement mandated reforms in Cyprus public health care to reduce waste and enhance efficiency Objectives 3 To assess the impact of this measure in non-avoidable and potentially avoidable visits Methods Study design 4 An interrupted time-series analysis Setting 5 Paphos Hospital, emergency department 2013e2015 Participants 6 (a) Cohort studydAll patients that visited the Paphos Hospital ED . No controls apply Variables 7 Emergent/non-avoidable and avoidable emergency department visits. Data sources/measurement 8a Paphos Hospital Emergency Department. Bias 9 Visits were assessed by a team of health professionals regarding their classification. Study size 10 We recorded all visits Quantitative variables 11 All visits, non-emergent/avoidable and emergent/non-avoidable Statistical methods 12 An interrupted time series analysis, which is extensively described in the methodology section Results Participants 13a (a) 213,102 visits to the emergency department, 20,247 defined as emergent/non-avoidable and 192,855, as avoidable Descriptive data 14a 213,102 visits to the emergency department, 20,247 defined as emergent/ non-avoidable and 192,855 as avoidable No control group Outcome data 15a Impact on co-payment emergent/non-avoidable and avoidable Emergency Department visits Main results 16 The introduction of co-payment did not exert any effect on emergent/non- avoidable visits (4% [95% CI: 4.3e11.08] P ¼ 0.694). Nevertheless, co- payments showed an immediate and sustained effect by reducing potentially avoidable emergency department visits, an effect that was statistically significant from the first month onwards (29.8% [95% CI: 22.6 e34.1] P < 0.00001). Discussion Key results 18 Co-payment reduced non-emergent/potentially avoidable visits, whereas it did not affect the rate of emergent/non-avoidable visits Limitations 19 A long-term monitoring for potential adverse events of this measure is imperative. Moreover, the classification between emergent/non-avoidable and non-emergent/potentially avoidable is prone to bias Interpretation 20 Co-payment, in its current form as applied in Cyprus, can contribute to the efficiency of the system, while it does not seem to impede access of people presenting with urgent conditions. Generalisability 21 Cyprus is a homogenous health market, and results can be extrapolated to the rest of the country. Other information Funding 22 No funding was received An explanation and elaboration article discusses each checklist item and gives methodological background and published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely available on the websites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at http://www.annals.org/ and Epidemiology at http://www.epidem.com/). Infor- mation on the STROBE Initiative is available at www.strobe-statement.org. a Give information separately for cases and controls in caseecontrol studies and, if applicable, for exposed and unexposed groups in cohort and cross-sectional studies. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 852 � conditions amenable to ambulatory care that sufficient patient monitoring and follow-up should be able to reduce or eliminate;27 � preventable health conditions for which manifestations and exacerbations are usually caused by lack of timely and proper primary care (such as asthma attacks), which are treatable in primary health-care services.28e32 On the contrary, non-avoidable visits included conditions for which an ED deems necessary as in the case of the following: � life-threatening conditions; � potential impairment of vital functions; � patients admitted in critical condition; � major injuries; http://www.plosmedicine.org/ http://www.annals.org/ http://www.epidem.com/ http://www.strobe-statement.org https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 8 53 � alcohol and substance intoxication and � all other conditions requiring immediate hospitalization. We performed three analyses: one for total visits, one for non-avoidable visits and one for potentially avoidable visits, to show the effects of introducing co-payments on these three categories. Statistical analysis To distinguish the effects of introducing co-payments from those of the financial crisis which was ongoing at the time, we developed an interrupted time series (ITS) analysis, a method which is regarded as one of the strongest quasi-experimental approaches.33e35 ITS analysis is a valuable study design for examining the effectiveness of population-level health in- terventions that have been implemented at a clearly defined point in time.36e39 An ITS is one that is segmented by specific change points where the values of the time series change, in our case, because of the introduction of a new policy. In this context, we performed segmented and dynamic regression analysis.35 These models enabled us to explore the effect of introducing co-payments on ED visits while taking into consideration data autocorrelation. They also enable forecasts to be made on the basis of past values. Using this model avoids the need for a control group by making multiple assessments of the outcome variable. Another significant policy-relevant attribute of ITS analysis is the dynamic presentation of response, which can be classified as instant, delayed, pro- longed or transient; the method can also assess impact on a specific time point, e.g. 5 or 10 months after the introduction of co-payment. Another attribute is the ability to assess the course of a variable (in our case, the number of ED visits) over a prolonged period of time preceding the introduction of co- payment. We specified three autoregressive integrated moving average (ARIMA) models, for total , potentially avoidable and non-avoidable visits. This enabled us to examine the effects of two events, the financial crisis and the introduction of co- payment. The model predicts a value in a time series as a linear combination of the past and current values and errors of the model. Bayesian information criteria were used to define parameters of the ARIMA model. We assume that regression follows the pattern:33e36 Outcome ¼ constant þ a1 timeþ a2 phase þ a3 interact þ er. Before the copayment introduction, this regression is Outcome ¼ constant þ a1time þ er. This model will enable any changes in ED visits due to recession. After the intro- duction of the co-payment, this is transformed to Outcome ¼ constant þ a1* time þ a2*phase þ a3* interact þ er. Variables used in the model are as the following: Outcome ¼ number of visits to ED a 1 ¼ coefficient for time: time is a continuous variable indicating number of months from the start of the obser- vation period. a 2 ¼ coefficient for phase, where phase is a binary dummy variable denoted with 0 before introduction of co-payment and 1 after introduction of co-payment. a 3 ¼ coefficient for interact, where interact is a variable which is coded as zero before co-payment. After the introduction of co-payment, data points remain the same as the time coefficient. er ¼ error of the model (random variability that is not explained by the model). This variable will also elucidate whether the financial crisis exerted any impact on the number of ED visits. We assume a starting date for the effects of the financial crisis at January 2011, an approach adopted by several authors. At this point, Cyprus plunged into recession, as attested by its exclusion from international financial markets.16e18 Therefore, the a3 variable will enable us to evaluate any changes that occurred between the start of the financial crisis and the introduction of co-payment in August 2013. We regard the main determinant of changes in this period as the financial crisis.16,17 In our model, constant estimates the baseline outcome level, that is, the number of ED visits at the beginning of the observation period (at time zero). a1 estimates the change in the average ED visits for each month before introduction of the co-payment and it represent the baseline trend (Fig. 1). a2 estimates the change in average ED visits that occurs after introduction of co-payment, and a3 estimates the change in trend in the mean ED visits after introduction of co-payment, which is compared with the trend before the introduction of co-payment. Post-intervention slope is given by the sum of a1 and a3. Therefore, this model offers the capability of con- trolling for the baseline level and trend, which constitutes a major strength of segmented regression analysis.33e39 In addition to the aforementioned information, we calculated the dynamic forecast for the posteco-payment period starting August 2013. This enabled the estimation of the number of ED visits which would have occurred if co-payment had not been introduced. We then estimated the impact of co-payment at monthly intervals, by comparing observed and predicted ED visits. A 95% confidence interval was applied. We used the BoxeJenkins method to identify the model that fits better and estimated averageerank regression with daily ED visits for 42 consecutive months. Seasonality was assessed through the standard integration tests, and the LjungeBox test was be used for the white noise condition . This enabled us to assess how data fit the model and whether autocorrelation of residuals was random or not. In addition, we calculated the trend of ED visits without the introduction of co-payment. We assessed impact at intervals of one month, up to eleven months after the introduction of co-payment. Statistical analysis was performed using SPSS, v 21.0.40 Results Introduction of co-payment was associated with a statistically significant reduction in total ED visits which was evident from the first month after the introduction of this measure (Table 2, Fig. 2) and persisted until the end of the follow-up period. This https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 is it s R e d u c ti o n o f n o n -a v o id a b le E D v is it s c o m p a re d w it h fo re c a s te d v is it s 1 1 1 .3 8 % (5 .2 e 1 6 .7 7 ) p ¼ 0 .2 9 1 6 .3 % (� 1 .4 e 1 2 ) p ¼ 0 .7 8 1 1 5 .3 % (9 .8 e 1 5 .3 ) p ¼ 0 .9 6 1 5 .9 % (� 1 .5 e 1 2 .3 ) p ¼ 0 .6 9 1 �2 4 .4 9 % (� 1 5 .9 to �1 .5 7 ) p ¼ 0 .4 6 �0 .1 % (� 8 .2 e 6 .8 ) p ¼ 0 .5 6 1 4 .9 % (� 2 .9 e 1 1 .7 ) p ¼ 0 .2 9 1 1 3 .5 % (6 .8 5 e 1 9 .2 9 ) p ¼ 0 .1 8 1 �0 .6 8 % (� 9 .4 e 0 .6 2 ) p ¼ 0 .3 5 1 4 % (� 4 .3 e 1 1 .0 8 ) ¼ 0 .2 8 ) R 2 0 .8 2 3 L ju n g B o x 1 5 .7 (p ¼ 0 .5 4 3 ) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 854 effect was primarily due to the significant reduction of the potentially avoidable visits group (Table 2, Fig. 1). Introduction of co-payment did not have any effect on non-avoidable cases, a consistent finding from 11 months after the intervention (Table 2, Fig. 1). The time-lapse between the advent of the financial crisis and the introduction of co-payment showed no impact on ED visits, either non-avoidable or avoidable visits, as reflected by the slope of the ITS and the interact coefficient (est ¼ 0.7, P ¼ 0.664). We used Ljung Box to assess model fit, which was good, and the white noise residual did not have significant P values (P ¼ 0.432, P ¼ 0.51, P ¼ 0.543) (Table 2). Visual inspection of the autocorrelation function and partial autocorrelation function confirmed this. The model was tested for underfitting (drop- ping of questionable parameters) and overfitting (including extra parameters) through the use of R2.44 R2 (which denotes the square of the correlation between the response values and the predicted response values) was 0.866, 0.892 and 0.823 for total visits, potentially avoidable and non-avoidable visits, respectively. Therefore, we can conclude that model fits well and explains to data variance to a satisfactory degree (Table 2). T a b le 2 e R e d u ct io n o f E D v is it s fo ll o w in g in tr o d u ct io n o f 1 0 e u ro co -p a y m e n t in A u g u s t 2 0 1 3 . T im e p e ri o d a ft e r in tr o d u c ti o n o f c o -p a y m e n t R e d u c ti o n o f to ta l n u m b e r o f E D v is it s c o m p a re d w it h fo re c a s te d v is it s (9 5 % C I) R e d u c ti o n o f p o te n ti a ll y a v o id a b le E D v is it s c o m p a re d w it h fo re c a s te d v 1 m o n th 3 0 .3 % (2 4 .2 e 3 5 .5 ) p < 0 .0 0 0 1 3 2 .2 % (2 5 .1 e 3 6 .2 ) p < 0 .0 0 0 2 m o n th 2 5 .3 % (1 7 .2 e 3 1 .9 ) p < 0 .0 0 0 1 2 7 .2 % (1 8 .2 e 3 0 .7 ) p < 0 .0 0 0 3 m o n th 2 7 .8 % (2 0 .3 e 3 3 .9 ) p < 0 .0 0 0 1 2 9 .2 % (2 1 .5 e 3 2 .7 ) p < 0 .0 0 0 4 m o n th 3 3 .7 % (2 6 .8 e 3 9 .5 ) p < 0 .0 0 0 1 3 0 .8 % (2 5 .6 e 3 8 .1 ) p < 0 .0 0 0 5 m o n th 2 5 % (1 7 e 3 1 .7 ) p < 0 .0 0 0 1 2 7 .6 % (1 7 .1 e 3 0 .7 ) p < 0 .0 0 0 6 m o n th 3 0 .4 % (2 2 .4 e 3 7 ) p < 0 .0 0 0 1 3 3 .1 % (2 3 .4 e 3 6 ) p < 0 .0 0 0 1 7 m o n th 2 8 % (2 0 .6 e 3 4 .1 ) p < 0 .0 0 0 1 2 9 .4 % (2 1 .6 e 3 5 .3 ) p < 0 .0 0 0 8 m o n th 2 5 .5 % (1 7 .2 e 3 2 .1 ) p < 0 .0 0 0 1 2 7 .9 .1 % (1 7 .8 e 3 4 .1 ) p < 0 .0 0 0 9 m o n th 2 8 .3 % (2 0 .3 e 3 4 .9 ) p < 0 .0 0 0 1 3 0 .6 % (2 0 .3 e 3 4 .9 ) p < 0 .0 0 0 1 0 m o n th 2 8 .8 % (2 1 .0 9 e 3 5 .1 ) p < 0 .0 0 0 1 2 9 .8 % (2 2 .6 e 3 4 .1 ) p < 0 .0 0 0 R 2 0 .8 6 6 L ju n g B o x 1 7 (p ¼ 0 .4 3 2 ) R 2 0 .8 9 2 L ju n g B o x 1 7 .7 (p ¼ 0 .5 1 E D , e m e rg e n c y d e p a rt m e n t; C I, c o n fi d e n c e in te rv a l. S ta ti s ti c a ll y s ig n ifi c a n t le v e l is s e t a t P ¼ 0 .0 5 . Discussion The introduction of co-payment in Cyprus led to a significant and immediate reduction of ED visits, which persisted during the 11 months of follow-up with no signs of wearing off. The reduction was mainly due to reduction of use by people pre- senting with potentially avoidable conditions which could be treated in primary health-care centres. This is one of the reasons why the World Health Organization regards easily accessible primary health care as the most important component of an effective health system.41 In line with this, the Government of Cyprus must ensure that people deterred from visiting EDs by the co-payment received for the health care they need.42 For achieving this, an array of measures have been put forward by many authors. Reforms of the pri- mary care sector are necessary, including extension of open- ing hours and implementation of more convenient weekend and night shifts to provide people with potentially avoidable conditions with an accessible alternative to ED use.43e46 It is also imperative to address the issue of those without health coverage, for whom EDs constitute the only affordable pathway to timely health care.17e19 An interrelated issue concerns people who repeatedly visit EDs because of sub- stance use or persistent medical, psychological or social problems. Referring them to specialised services for dealing with the underlying condition is also desirable, although time constraints and reimbursement status may hinder their ac- cess.43 Community nursing and the introduction of minor trauma centres can also play a role in reducing the burden on EDs. Campaigns to raise public awareness concerning the appropriate use of ED centres are essential, because EDs are widely assumed in Cyprus to provide better quality health care than primary health-care centres.4,18 Such campaigns must highlight the fact that EDs are not designed to provide non-urgent care. On the contrary, doctors in EDs cannot easily access the medical records of patients, so the benefit to the non-urgent patient is questionable. In this respect, https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 Fig. 1 e Monthly potentially avoidable visits to emergency department. Dotted line indicates introduction of co-payment. Fig. 2 e Total number of visits to emergency department. Dotted line indicates introduction of co-payment. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 8 55 appropriate use of EDs can both reduce costs and improve health care.18,46 This study also showed that non-avoidable visits were not affected by the introduction of co-payment. This is a signifi- cant finding because it sheds light on one of the main objec- tions to user charges: the danger of impeding access to urgently necessary care.47 Nevertheless, for patients with very limited means and those who use the ED frequently to obtain treatment for persistent or multiple conditions, even a 10 euro charge could be a deterrent. The fluctuations in the economy of Cyprus did not appear to have any impact on ED visits, as seen in the ITS analysis, specifically, in the interaction coefficient (Figs. 1e3). Although some authors have reported a marked shift towards free public health care during crisis periods,47e49 our data do not support this. Also, noteworthy is the high percentage of potentially avoidable visits, which surpassed estimates from other countries50e52 and implies significant waste in the form of productive and allocative inefficiency. In the face of financial crisis, it is tempting for cash- deprived countries to raise levels of co-payment, to compen- sate for reduced tax income due to unemployment and reduced incomes.53 This underpins the importance of moni- toring health indicators in the long term to track potential https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 Fig. 3 e Monthly non-avoidable visits to emergency department. Dotted line indicates introduction of co-payment. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 856 delayed effects of the measure. Excessive use of co-payments as a source of revenue may create inequities in the short term and increased health costs in the long run.2e5,8e10,53 There- fore, they should be (and often are) combined with exemp- tions, ceilings and reimbursements that counteract the tendency to penalise the sick and the poor. Limitations of this study This study is subject to some limitations. It was not possible to check the validity of the classification of visits as ‘non- avoidable’ or ‘potentially avoidable’ against other indicators. The description of symptoms in medical records may not correspond to the experience of the patient; it may also influenced by the decision to treat or not to treat. There was no examination of inequities resulting from the introduction of co-paymentsdthe possibility that people with more health problems and those with limited financial means would be more strongly penalised by co-payments than others. Data were also obtained from only one hospital. Although no sig- nificant regional differences apply, as always, caution is required when extrapolating from these data. Conclusions Introduction of copayment reduced ED use, primarily due to a reduction of potentially avoidable visits, whereas, in the short term (up to 11 months), it did not reduce non-avoidable visits. Therefore, introduction of co-payment in the midst of financial crisis exerts a beneficial and discrete effect on public health- care resources, at least within the time span studied. This can be partly attributed to the low amount of the co-payment, which encourages personal responsibility without providing a serious barrier to access (which could be interpreted as pun- ishment for improper use). Nevertheless, the long-term im- pacts still have to be assessed. To this end and to safeguard equity in health provision, as well as avoiding possible future costs arising from impediments to necessary treatment, it is crucial to establish accessible and affordable primary care and to provide health education to the public. This presupposes a massive overhaul of the system to adapt it to users' needs. Author statements Ethical approval Dr Panagiotis Petrou is grateful to Maria Michael, adminis- tration officer at Health Insurance Organization, for her affirmed support, which embodies the quote of M. Twain “it's not the size of the dog in the fight, it's the size of the fight in the dog” and in this sense, we cannot “bid farewell to Alexandria”. This study was approved by Cyprus Health Research Board (0059/2012). Funding No funding was received. Competing interests None declared. r e f e r e n c e s 1. Weinick RM, Billings J, Thorpe JM. Ambulatory care sensitive emergency department visits: a national perspective. Acad Emerg Med 2003;10(5):525e6. 2. Selby JV, Fireman BH, Swain BE. Effect of a copayment on use of the emergency department in a health maintenance organization. N Engl J Med 1996;334(10):635e41. http://refhub.elsevier.com/S0033-3506(18)30406-2/sref1 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref1 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref1 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref1 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref2 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref2 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref2 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref2 https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 8 57 3. O'Grady KF, Manning WG, Newhouse JP. The impact of cost sharing on emergency department use. N Engl J Med 1985;313(8):484e90. 4. Selby JV. Cost sharing in the emergency department: is it safe? Is it needed? N Engl J Med 1997;336(24):1750e1. 5. Mortensen K. Departments copayments did not reduce medicaid enrollees' nonemergency use of emergency. Health Aff 2010;29(9):1643e50. 6. Falcone M, Fabozzi F, Bachini L. Tuscany case study health impact assessment tools: copayment policies evaluation Agenzia Regionale di Sanit�a Toscana, Italy. 2014. Available at: https:// www.ars.toscana.it/files/aree_intervento/indicatori_equita_ qualita/equity_action/WP4_HIA_CASE_%20STUDY_Tuscany_ 2014.pd. 7. Wong MD, Andersen R, Sherbourne CD, Hays RD, Shapiro MF. Effects of cost sharing on care seeking and health status: results from the Medical Outcomes Study. Am J Public Health 2001;91:1889e94. 8. Remler DK, Greene J. Cost-sharing: a blunt instrument. Annu Rev Public Health 2009;30:293e311. 9. Becker D, Blackburn J, Morrisey M, Sen B, Kilgore M, Caldwell C, et al. Co- payments and the use of emergency department services in the children's health insurance program. Med Care Res Rev 2013;70:514. 10. Newhouse JP. Insurance experiment group. In: Free for all? Lessons from the RAND health insurance experiment. Cambridge (MA): Harvard University Press; 1993. 11. Brook RH, Ware Jr JE, Rogers WH, Keeler EB, Davies AR, Donald CA, Newhouse JP. Does free care improve adults' health? Results from a randomized controlled trial. N Engl J Med 1983;309:1426e34. 12. Hsu J, Price M, Brand R, Ray GT, Fireman B, Newhouse JP, Selby JV. Cost-sharing for emergency care and unfavorable clinical events: findings from the safety and financial ramifications of ED copayments study. Health Serv Res 2006 Oct;41(5):1801e20. 13. Madden JM, Soumerai SB, Lieu TA, Mandl KD, Zhang F, Ross- Degnan D. Effects of a law against early postpartum discharge on newborn follow-up, adverse events, and HMO expenditures. N Engl J Med 2002;347:2031e8. 14. Petrou P, Vandoros S. Cyprus in crisis: recent changes in the pharmaceutical market and options for further reforms without sacrificing access or quality of treatment. Health Policy May 2015;119(5):563e8. 15. Petrou Panagiotis, Vandoros Sotiris. Healthcare reforms in Cyprus 2013-2017: does the crisis mark the end of the healthcare sector as we know it? Health Policy 2018;122:75e80. 16. Petrou P. Financial crisis as a reform mediator in Cyprus's health services. Eurohealth Euro Observ 2014;20(4). 17. Petrou Panagiotis. Crisis as a serendipity for change in Cyprus' healthcare services. 2015. p. 805e7. 18. Petrou P. An interrupted time -series analysis to assess impact of introduction of co -payment on emergency room visits in Cyprus. Appl Health Econ Health Policy 2015;13:515e23. 19. Van den Heede Koen, Van de Voorde Carine. Interventions to reduce emergency department utilisation: a review of reviews. Health Policy 2016;120:1337e49. 20. Eichler K, Hess S, Chmiel C, Karin B€ogli, Patrick Sidler, Oliver Senn, et al. Sustained health-economic effects after reorganisation of a Swiss hospital emergency centre: a cost comparison study. Emerg Med J 2014;31:818e23. 21. Mehrotra A, Liu H, Adams J, Wang M, Lave J, Thygeson M, et al. Comparing costs and quality of care at retail clinics with that of other medical settings for 3 common illnesses. Ann Intern Med 2009;151(5):321e8 (PubMed: 19721020). 22. Martin BC. Emergency medicine versus primary care: a case study of three prevalent, costly, and non-emergent diagnoses at a community teaching hospital. J Health Care Finance 2000;27(2):51e65. 23. Pines MJ. Emergency department crowding is associated with poor care for patients with severe pain. Ann Emerg Med 2008;51(1):1e5. January. 24. Guttman A, Schull MJ, Vermeulen MJ, Stukel TA. Association between waiting times and short term mortality and hospital admission after departure from emergency department: population based cohort study from Ontario, Canada. BMJ 2011;342. 2011 Jun 1. 25. Ministry of Finance. Memorandum of understanding on specific economic policy conditionality. Republic of Cyprus: Nicosia; 2013. 26. Paphos Hospital, Ministry of Health https://www.moh.gov.cy/ moh/pgh/pgh.nsf/index_gr/index_gr?opendocument. 27. ACEP. Efficiency in the Emergency Department. Doing things faster without sacrificing quality. In: ACEP reference and resource guide; 2004. American College of Emergency Practicians. 28. Washington State Hospital Association. Potentially avoidable emergency room use. 2011. Report available at: http://wsha- archive.seattlewebgroup.com/files/62/ERReport2 (last accessed August 2015). 29. NYU Center for Health and Public Service Research ED ALGORITHM available at http://wagner.nyu.edu/faculty/ billings/nyued-background. 30. Gruneir A, Bell CM, Bronskill SE, Schull M, Anderson GM, Rochon PA. Frequency and pattern of emergency department visits by long-term care residents–a population-based study. J Am Geriatr Soc. Mar 2010;58(3):510e7. https://doi.org/10.1111/ j.1532-5415.2010.02736.x. 31. Barish R, Mcgauly R, Arnold T. Emergency room crowding: a marker of hospital health. Trans Am Clin Climatol Assoc 2012;123. 32. Ragin DF, Hwang U, Cydulka RK, Holson D, Haley Jr LL, Richards CF, Becker BM, Richardson LD. Reasons for using the emergency department: results of the EMPATH Study. Acad Emerg Med 2005;12(12):1158e66. 2005 Dec. Epub 2005 Nov 10. 33. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Therapeut 2002;27:299e309. 34. Pankratz A. Forecasting with dynamic regression models. New York, NY: Wiley; 1991. 35. Yaffee P. Introduction to time series analysis and forecasting. New York: ACADEMIC PRESS, INC; 1999. 36. Veney JE, Kaluzny AD. Trend analysis techniques and interpretation. In: Veney JE, Kaluzny AD, editors. Evaluation and decision making for health services. Ann Arbor, MI: Health Administration Press; 1998. 37. Biglan A, Ary D, Wagenaar AC. The value of interrupted time- series experiments for community intervention research. Prev Sci 2000;1:31e49. 38. Shadish W, Cook T, Campbell D. Experimental and quasi- experimental designs for generalized causal inference. Boston, Mass: Houghton Mifflin; 2002. 39. Bernal James Lopez, Cummins Steven, Gasparrini Antonio. Interrupted time series regression for the evaluation of public health interventions: a tutorial. Int J Epidemiol 1 February 2017;46(1):348e55. https://doi.org/10.1093/ije/dyw098. 40. SPSS IBM Corp. Released 2012. IBM SPSS statistics for windows, version 21.0. Armonk, NY: IBM Corp. 41. Evans T., Lerberghe VM (eds). The world health report 2008. Geneva, Switzerland: Now More Than Ever World Health Organization. 42. Weinick Robin M, Burns Rachel M, Mehrotra Ateev. How many emergency department visits could be managed at http://refhub.elsevier.com/S0033-3506(18)30406-2/sref3 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref3 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref3 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref3 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref4 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref4 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref4 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref5 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref5 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref5 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref5 https://www.ars.toscana.it/files/aree_intervento/indicatori_equita_qualita/equity_action/WP4_HIA_CASE_%20STUDY_Tuscany_2014.pd https://www.ars.toscana.it/files/aree_intervento/indicatori_equita_qualita/equity_action/WP4_HIA_CASE_%20STUDY_Tuscany_2014.pd https://www.ars.toscana.it/files/aree_intervento/indicatori_equita_qualita/equity_action/WP4_HIA_CASE_%20STUDY_Tuscany_2014.pd https://www.ars.toscana.it/files/aree_intervento/indicatori_equita_qualita/equity_action/WP4_HIA_CASE_%20STUDY_Tuscany_2014.pd http://refhub.elsevier.com/S0033-3506(18)30406-2/sref7 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref7 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref7 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref7 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref7 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref8 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref8 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref8 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref9 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref9 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref9 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref9 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref10 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref10 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref10 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref11 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref11 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref11 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref11 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref11 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref12 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref13 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref13 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref13 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref13 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref13 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref14 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref14 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref14 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref14 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref14 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref15 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref15 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref15 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref15 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref16 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref16 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref17 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref17 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref17 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref18 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref18 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref18 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref18 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref19 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref19 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref19 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref19 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref20 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref21 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref21 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref21 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref21 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref21 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref22 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref22 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref22 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref22 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref22 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref23 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref23 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref23 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref23 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref24 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref24 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref24 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref24 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref24 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref25 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref25 https://www.moh.gov.cy/moh/pgh/pgh.nsf/index_gr/index_gr?opendocument https://www.moh.gov.cy/moh/pgh/pgh.nsf/index_gr/index_gr?opendocument http://refhub.elsevier.com/S0033-3506(18)30406-2/sref27 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref27 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref27 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref27 http://wsha-archive.seattlewebgroup.com/files/62/ERReport2 http://wsha-archive.seattlewebgroup.com/files/62/ERReport2 http://wagner.nyu.edu/faculty/billings/nyued-background http://wagner.nyu.edu/faculty/billings/nyued-background https://doi.org/10.1111/j.1532-5415.2010.02736.x https://doi.org/10.1111/j.1532-5415.2010.02736.x http://refhub.elsevier.com/S0033-3506(18)30406-2/sref31 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref31 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref31 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref32 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref33 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref33 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref33 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref33 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref33 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref34 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref34 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref35 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref35 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref36 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref36 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref36 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref36 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref37 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref37 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref37 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref37 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref38 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref38 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref38 https://doi.org/10.1093/ije/dyw098 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref42 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref42 https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 0 e5 858 urgent care centers and retail clinics? Health Aff 2010 September;29(9):1630e6. 43. Pines JM, Hilton JA, Weber EJ, Alkemade AJ, Al Shabanah H, Anderson PD, et al. International perspectives on emergency department crowding December. Acad. Emerg. Med. 2011;18(12). 44. Franco SM, Mitchell CK, Buzon RM. Primary care physician access and gatekeeping: a key to reducing emergency department use. Clin Pediatr 1997;36(2):63e8. 45. Lowe RA, Localio AR, Schwarz DF, Williams S, Tuton LW, Maroney S, et al. Association between primary care practice characteristics and emergency department use in a medicaid managed care organization. Med Care 2005;43(8):792e800. 46. Flores-Mateo G, Violan-Fors C, Carrillo-Santisteve P, Peiro S, Argimon JM. Effectiveness of organizational interventions to reduce emergency department utilization: a systematic review. PLoS One 2012;7. e35903. [Electronic Resource]. 47. Kim J, Ko S, Yang B. The effects of patient cost sharing on ambulatory utilization in South Korea. Health Policy 2005;72:293e300. 48. Kentikelenis A. Bailouts, austerity and the erosion of health coverage in Southern Europe and Ireland. Eur J Public Health 2015;25(3):365e6. 49. Castellana C. Impact of the economic crisis on the Italian public healthcare expenditure working paper. Management School in Clinical Engineering. Available at: http://arxiv.org/pdf/1205. 2863v1 (last assessed December 2014). 50. Liu T, Sayre MR, Carleton SC. Emergency medical care: types, trends, and factors related to nonurgent visits. Acad Emerg Med 1999 Nov;6(11):1147e52. 51. Diserens L�eonard, Egli Lukas, Fustinoni Sarah, Santos- Eggimann Brigitte, Staeger Philippe, Hugli Olivier. Emergency department visits for non-life-threatening conditions: evolution over 13 years in a Swiss urban teaching hospital Swiss. Med Wkly 2015;145:w14123. 52. Jayaprakash N, O'Sullivan R, Bey T, Ahmed S, Lotfipour S. Crowding and delivery of healthcare in emergency departments: the european perspective. West J Emerg Med 2009 Nov;10(4):233e9. 53. Kellermann A, Weinick R. Emergency departments, medicaid costs, and access to primary care d understanding the link. N Engl J Med 2012;366:2141e3. http://refhub.elsevier.com/S0033-3506(18)30406-2/sref42 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref42 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref42 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref43 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref43 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref43 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref43 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref44 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref44 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref44 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref44 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref45 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref46 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref46 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref46 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref46 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref46 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref47 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref47 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref47 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref47 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref48 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref48 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref48 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref48 http://arxiv.org/pdf/1205.2863v1 http://arxiv.org/pdf/1205.2863v1 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref50 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref50 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref50 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref50 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref51 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref52 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref52 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref52 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref52 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref52 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref53 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref53 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref53 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref53 http://refhub.elsevier.com/S0033-3506(18)30406-2/sref53 https://doi.org/10.1016/j.puhe.2018.12.014 https://doi.org/10.1016/j.puhe.2018.12.014 Co-payments for emergency department visits: a quasi-experimental study Introduction Cyprus Methods Data Statistical analysis Results Discussion Limitations of this study Conclusions Author statements Ethical approval Funding Competing interests References Validity-of-screening-tools-for-dementia-and-mild-cognitive-impa_2019_Public ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 2 Available online at w Public Health journal homepage: www.elsevier.com/puhe Review Paper Validity of screening tools for dementia and mild cognitive impairment among the elderly in primary health care: a systematic review M.A. Abd Razak a,*, N.A. Ahmad a, Y.Y. Chan a, N. Mohamad Kasim a, M. Yusof b, M.K.A. Abdul Ghani c, M. Omar d, F.A. Abd Aziz a, R. Jamaluddin a a Institute for Public Health, Ministry of Health Malaysia, Jalan Bangsar, 50590 Kuala Lumpur, W.P. Kuala Lumpur, Malaysia b Women and Child Hospital Kuala Lumpur, Ministry of Health Malaysia, Jalan Dr Latiff, 50586 Kuala Lumpur, W.P. Kuala Lumpur, Malaysia c Klinik Rafeeq & Nurul, Sungai Rengit, 81620 Pengerang, Johor, Malaysia d Kuala Selangor Health District, Ministry of Health Malaysia, Jalan Semarak, 45000 Kuala Selangor, Selangor, Malaysia a r t i c l e i n f o Article history: Received 4 June 2018 Received in revised form 14 December 2018 Accepted 2 January 2019 Available online 1 March 2019 Keywords: Systematic review Dementia Mild cognitive impairment Screening tool Primary health care * Corresponding author. Tel.: þ603 22979421 E-mail addresses: aznuddin.ar@moh.go (Y.Y. Chan), noraida_kasim@moh.gov.my (N (M.K.A. Abdul Ghani), bentiomarbarawas@gm my (R. Jamaluddin). https://doi.org/10.1016/j.puhe.2019.01.001 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: This systematic review aims to provide updated and comprehensive evidence on the validity and feasibility of screening tools for mild cognitive impairment (MCI) and dementia among the elderly at primary healthcare level. Study design: A review of articles was performed. Methods: A search strategy was used by using electronic bibliographic databases including PubMed, Embase and CENTRAL for published studies and reference list of published studies. The articles were exported to a bibliographic database for further screening process. Two reviewers worked independently to screen results and extract data from the included studies. Any discrepancies were resolved and confirmed by the consensus of all authors. Results: There were three screening approaches for detecting MCI and dementia e screening by a healthcare provider, screening by a self-administered questionnaire and caretaker informant screening. Montreal Cognitive Assessment (MoCA) was the most common and preferable tool for MCI screening (sensitivity [Sn]: 81e97%; specificity [Sp]: 60 e86%), whereas Addenbrooke's Cognitive Examination (ACE) was the preferable tool for dementia screening (Sn: 79e100%; Sp: 86%). Conclusion: This systematic review found that there are three screening approaches for detecting early dementia and MCI at primary health care. ACE and MoCA are recom- mended tools for screening of dementia and MCI, respectively. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. ; fax: þ603 22823114. v.my (M.A. Abd Razak), drnoorani@moh.gov.my (N.A. Ahmad), chan.yy@moh.gov.my . Mohamad Kasim), muslimah_yusof@moh.gov.my (M. Yusof), drkamalariff@gmail.com ail.com (M. Omar), Fazlyazry.abdulaziz@moh.gov.my (F.A. Abd Aziz), rasidah.j@moh.gov. ic Health. Published by Elsevier Ltd. All rights reserved. mailto:aznuddin.ar@moh.gov.my mailto:drnoorani@moh.gov.my mailto:chan.yy@moh.gov.my mailto:noraida_kasim@moh.gov.my mailto:muslimah_yusof@moh.gov.my mailto:drkamalariff@gmail.com mailto:bentiomarbarawas@gmail.com mailto:Fazlyazry.abdulaziz@moh.gov.my mailto:rasidah.j@moh.gov.my mailto:rasidah.j@moh.gov.my http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.001&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 2 85 Introduction Globally, the estimated prevalence of dementia is around 5e7% in most regions, with the highest in Latin America (8.5%). The number is estimated to increase as the population of the world's elderly continues to increase.1 It is currently estimated that 35.6 million people live with dementia worldwide and that the number will double by 2030 and more than triple by 2050, with the majority living in developing countries.2 Dementia is a broad syndrome ranging from mild to severe cognitive impairment that significantly causes disability in older people.2 The cognitive impairment may include memory loss, difficulty in understanding or using words, inability to carry out motor activities despite adequate motor function and failure to identify or recognize objects.3,4 Routine clinical practice shows that the cognitive and functional changes of dementia are typically accompanied by changes in behaviour and personality, but these conditions have not become core criteria as they have been considered to lack sufficient diag- nostic specificity.5 About one in 10 people aged 65 years had dementia, and the prevalence increases by age. However, there was no gender difference for incidence of dementia.6 There are various types of dementia such as Alzheimer's disease (AD) dementia, vascular dementia, frontotemporal dementia and Lewy body dementia. AD is known as the most common cause of dementia, and it is an irreversible, pro- gressive brain disorder that mostly starts in those aged in their mid-60s.7 Vascular dementia commonly occurs due to blockage of blood vessels in the brain, leading to the death of tissues or infarction in the affected region. Frontotemporal dementia primarily affects regions of the brain governing planning, social behaviour and language perceptions.3,8 Lewy body dementia is characterized by the presence of Lewy bodies, protein in the cerebral cortex and brain stem.3 Some of these forms of dementia can be reversed through timely in- terventions. Thus, public health interventions to raise awareness of the importance of early screening for cognitive impairment among the elderly are necessary.3 Screening for dementia and early diagnosis among those who are at risk are important in managing the disease and ensuring preparedness among caregivers.9,10 Various screening tool modalities, including self-administered questionnaires, face-to-face assessment, telephone-based assessment and iPAD version of assessment, have been introduced for screening patients who have subjective memory complaints.11e13 Studies have also been carried out to evaluate the most suitable screening tools for dementia in primary care practice. These need to be brief, be easy to administer, be acceptable to the elderly and have high sensitivity and specificity.14 At present, one of the commonly used screening tools for dementia is the Mini-Mental State Examination (MMSE).15 However, the MMSE is known to have a long administration time and is difficult to interpret by general practitioners.14,16 As such, it is not the most efficient or feasible screen for use in primary care. Fortunately, other options exist, such as the General Practitioner Assessment of Cognition (GPCOG), the Memory Impairment Screen (MIS) and the Mini Cognitive Assessment Instrument (Mini-Cog).16 These tools have been validated; are brief and easy to administer and have a negative predictive value similar to the MMSE.16,17 Evaluating dementia-screening tests is a complex task. Reports of excellent sensitivity and specificity for a given in- strument must take into account whether performance may be inflated by high rates of dementia in the study sample, by high average severity of cognitive impairment among affected persons or by exclusion of subjects with demographic char- acteristics that compromise many screens.14 On the whole, effectiveness, freedom from biases irrelevant to dementia status, brevity and simplicity are the key characteristics of an ideal dementia-screening tool. It is not easy for clinicians to detect mild dementia because most patients present when they are in the moderate to severe stages.11,16 In addition, as discussed in a previous study,18 several barriers related to patients' caregivers and the healthcare system have been identified as contributing factors for missed or delayed diag- nosis of dementia in the primary care setting. Various reviews have been published on screening for MCI and dementia using various approaches and in various settings.19e23 A systematic review for the United States Pre- ventive Services Task Force based on data sources and searches until 10th December 2012 found that brief in- struments to screen for cognitive impairment can adequately detect dementia, but it is unclear whether the screening im- proves decision-making.24,25 In addition, a more recent sys- tematic review on cognitive assessment tools in Asia (databases searched between September 1989 and June 2014) reported that validated cognitive assessment tools in Asia are limited and subjected to cultural as well as educational bias.26 However, most of these reviews are based on articles pub- lished before 2013.20,21,23e25 Therefore, we carried out this systematic review to update current knowledge on the validity and feasibility of screening tools for dementia and mild cognitive impairments (MCIs) among the elderly at primary care level covering literature from 2012 to 2017. Methods The present systematic review was conducted based on the Cochrane Handbook for Systematic Review of Interventions guidelines27 and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE).28 The review included all validity studies assessing the sensitivity and specificity of various screening tools. We also looked for information on the feasibility of each tool, if available. The population considered in this study was the elderly aged 60 years and above who were screened at primary health care settings using three approaches e screening by a healthcare provider, screening by a self-administered questionnaire or by caregiver informant screening. Search strategy A search strategy was developed to identify studies for this review. The search strategy contained population, intervention https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 286 and outcome terms. The search terms ‘dementia’, ‘screening’ and ‘validity’ included Medical Subject Headings (MeSH) as well as title and abstract text searches. Searches were limited to elderly aged 60 years and above. Electronic searches for eligible articles were conducted for articles published in 2012 through 2017 in three databases: PubMed, Embase and CENTRAL. Then, we searched for reference lists of published studies and looked for further work done by correspondence authors. Inclusion and exclusion criteria Studies that validate any screening tools delivered by a healthcare provider in primary care, self-administered by patients or delivered by an informant as well as studies that were written in English only were included. Studies written in other languages were excluded. Screening and review process The studies identified through the search process were exported to a bibliographic database (EndNote version 7) for duplicates identification. Two reviewers (F.A.A.A. and R.J.) independently reviewed the titles, abstracts and keywords of electronic records for eligibility according to all inclusion criteria of this review. The initial screening results were Records identified through searching 3 databases (PubMed, CENTRAL, Embase) (n = 120) Records after duplicates removed (n = 116) Records screened (n = 116) Full text articles assessed for eligibility (n = 81) Studies included into systematic analysis (n = 30) noitacifitnedI Sc re en in g E lig ib ili ty In cl ud ed Fig. 1 e Study selection flow diagram (PRISMA). PRISMA, Prefer Analyses. compared and discussed among all reviewers. Where possible, full texts of screened titles and abstracts were ob- tained, and two reviewers (N.A.A. and C.Y.Y.) independently reviewed the full texts. The potential full texts were rescreened by other reviewers for inclusion in the final re- view by using a screening data form. Any disagreements were discussed and resolved among all reviewers. By using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA)29 method of screening, reviewers eliminated articles not related to the study. Fig. 1 illustrates the process of data collection and study selection methods. Two reviewers (M.A.A.R. and N.M.K.) independently per- formed data extraction by using a standardized data extrac- tion form. The data extraction form included variables such as research questions, study designs, participants, study outcomes, sensitivity (Sn), specificity (Sp) and feasibility of the screening tools. Any issues at this stage were resolved by discussion among all reviewers using our own appraisal format based on sensitivity, specificity and feasibility of the tools. The appraisal format has a maximum score of 25 with 5 points for each criterion: 1, Sn � 90%; 2, Sp � 80%; 3, acceptable by patients; 4, can be assessed by a non-physician; 5, short administration time of 15 min or less. Feasibility was ascertained based on three criteria e acceptability, judge- ment required and cost e as mentioned in the article Records excluded (n = 35) Full text articles excluded, with reason: aim to specific target group/ ethnics and not English version (n = 51) Additional records identified through other sources (clinical trial etc.) (n = 24) red Reporting Items for Systematic reviews and Meta- https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 Table 1 e Included studies characteristic. Number Tool Authors, Place, Year Setting Age eligibility, years Approaches 1 Montreal Cognitive Assessment-Basic (MoCA-B) Parunyou J et al., Bangkok, 201530 Hospital 55e80 Screening by a healthcare providerMontreal Cognitive Assessment (MoCA) Felicia C.G et al. Africa, 201431 Clinic �50 Yan H.D et al. Singapore, 201232 Clinic a Larner A.J, UK, 201233 Clinic 20e87 Freitas S et al. Portugal, 201334 Community 70.52 Roalf D.R et al. USA, 201335 Clinic 52e88 2 Short Portable Mental Status Questionnaire, SPMSQ Chetna M et al., Singapore, 201336 Clinic a 3 Memory, fluency and orientation (MEFO) Delgado D.C et al., Santiago, 201337 Clinic and community �60 4 Addenbrooke's Cognitive Examination III (ACE-III) Michael T.J et al., United Kingdom 201538 Community a Hsieh S et al., Australia, 201339 Community 66 ± 6.25 5 A Quick Test of Cognitive Speed (AQT-CF) Takahashi F et al. Japan, 201240 Hospital and community a 6 Saint Louis University Mental Status (SLUMS) Cummings-Vaughn L.A et al. United States, 201441 Community �60 7 5 Object Test Papageorgiou S.G et al. Greece, 201342 Hospital a 8 Brief Neuropsychological Battery (BNB)Semantic Fluency Serna A et al. 201543 Community a 9 The Subjective Memory Complaint Clinical (SMCC) compare to MMSE and CDT Ramlall et al. 201344 Community �60 10 Cognitive Abilities Screening Instrument-Short (CASI-S) Martins de Oliveira G et al., 201545 Clinic �60 11 Rapid Cognitive Screen (RCS) Malmstrom T.K et al., 201546 Hospital, Clinic 65e92 60e90 12 Cognitive Performance Scale (CPS) which is generated from five items of the interRAI Acute Care Wellens N.I.H et al., 201347 Hospital �75 13 Literacy Independent Cognitive Assessment Shim Y.S et al. 201548 Hospital, community �60 14 Brief Interview for Mental Status (BIMS) Mansbach W.E et al. United States 201449 Clinic �60 15 Brief Cognitive Assessment Tool (BCAT) Clinic �60 16 Modified Mini-Mental State Examination (3MS) Holsinger T et al., USA 201250 Clinic �65 17 Mini-Cog Clinic �65 18 Memory Impairment Screen (MIS) Clinic �65 19 Memory Function 2 administered to the participant (MF-2) Clinic �65 20 Virtual Reality (VT) technology: Virtual supermarket (VSM) Zygouris S et al., Greece, 201451 Clinic >56 Self-administered

screening21 Virtual Reality Day-Out-Task (VR-DOT) Tarnanas I et al. Switzerland, 201352 Clinic >60
22 Computerized Cognitive Screening Tests (CCS) Scanlon L et al. Ireland, 201553 Hospital �55
23 Computerized Assessment of Mild Cognitive Impairment

(CAMCI)

Tierney M.C et al., Canada, 201454 Clinic �65

24 Cognitive Assessment for Dementia, iPad version (CADi) Onoda K et al., Japan, 201355 Hospital, community �65
25 Revised Cognitive Assessment for Dementia, iPad version

(CADi-2)

Onoda K et al., Japan, 201456 Clinic 78.1 ± 4.4
76.0 ± 3.0

26 Dementia Risk Assessment (DRA) Brandt J et al., USA, 201357 Community �50
27 Participant-rated (p-AD8) Chin R et al., Singapore, 201358 Community 66.7 ± 10.08
28 Informant Questionnaire on Cognitive Decline in the Elderly

individuals (IQCODE)

Li F et al., China, 201259 Hospital, community �55 Screening by a
caretaker informant

a Age was not mentioned.

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
8
4
e
9
2

8
7

https://doi.org/10.1016/j.puhe.2019.01.001

Table 2 e Validity and feasibility of screening tools to detect dementia and MCI.

Number Tool Authors, Place, Year MCI Dementia Feasibility

Cut-off point Sn/Sp (%) Cut-off point Sn/Sp (%) Acceptability Judgement Duration
(minutes)

1 Montreal Cognitive Assessment-Basic

(MoCA-B)

Parunyou J et al. Bangkok, 201530 25/30 86/86 þþ þþ Yes Yes 15e21 m

Montreal Cognitive Assessment (MoCA) Felicia C.G et al. Africa, 201431 �24 95/63 �22 96/88 Yes Yes 10e15 m
Yan H.D et al. Singapore, 201232 � 19 83/86 19/20 83/86
Larner A.J, UK, 201233 �26 97/60 þþ þþ
Freitas S et al. Portugal, 201334 <22 81/77 þþ þþ Raolf D.R et al. USA, 201335 25 84/79 þþ þþ 2 Short Portable Mental Status Questionnaire, SPMSQ Chetna M et al. Singapore, 201336 �5 78/75 þþ þþ Yes Yes 10e15 m 3 Memory, fluency and orientation (MEFO) Delgado D.C et al. Santiago, 201337 <70 81/63 <7 86/96 Yes Yes 10e15 m 4 Addenbrooke's Cognitive Examination III (ACE-III) Michael T.J et al. United Kingdom 201538 þþ þþ <81 79/96 Yes Yes 15 m Hsieh S et al. Australia, 201339 þþ þþ 88 100/96 Yes Yes 15 m 5 A Quick Test of Cognitive Speed (AQT-CF) Takahashi F et al. Japan, 201240 þþ þþ 71/72 85/76 Yes Yes 3e5 m 6 Saint Louis University Mental Status (SLUMS) Cummings-Vaughn L.A et al. United States, 201441 �26 74/65 �20 93/96 Yes Yes 7 m 7 5 Object Test Papageorgiou S.G et al. Greece, 201342 3/5 per trial 23/98 þþ þþ Yes Yes <5 m 8 Brief Neuropsychological Battery (BNB) Semantic Fluency Serna A et al. 201543 11.5 62/67 10.5 79/76 Yes Yes 31 m 9 The Subjective Memory Complaint Clinical (SMCC) compare to MMSE and CDT Ramlall et al. 201344 þþ þþ >0
� 24
�5

90.9/45.7 63.6/

76 44.4/88.9

Yes No NA

10 Cognitive Abilities Screening Instrument-

Short (CASI-S)

Martins de Oliveira G et al., 201545 þþ þþ 22/23 93/81 Yes Yes NA

11 Rapid Cognitive Screen (RCS) Malmstrom T.K et al., 201546 �7 87/70 �5 89/94 Yes Yes <3 m �7 69/82 �5 92/94 12 Cognitive Performance Scale (CPS) which is generated from five items of the interRAI Acute Care Wellens NIH et al., 201347 �2 � 1 51/95 73/68 þþ þþ NA NA NA 13 Literacy Independent Cognitive Assessment Shim Y.S et al. 201548 202/203 187/188 209/210 76/72.7 76/ 70.3 75.5/71.4 þþ þþ Yes Yes 20 m 14 Brief Interview for Mental Status (BIMS) Mansbach W.E et al. United States 201449 þþ þþ <13 66/88 Yes Yes 3 m 15 Brief Cognitive Assessment Tool (BCAT) þþ þþ <36 99/81 Yes Yes 10e15 m 16 Modified Mini-Mental State Examination (3MS) Holsinger T et al., USA 201250 þþ þþ <83 86/79 Yes Yes 17 m 17 Mini-Cog þþ þþ <3 76/73 Yes Yes 3 m 18 Memory Impairment Screen (MIS) þþ þþ <5 43/93 Yes Yes 4 m 19 Memory Function 2 administered to the participant (MF-2) þþ þþ Both Yes 38/87 Yes Yes <2 m p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e 9 2 8 8 https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 2 0 V ir tu a l R e a li ty (V T ) te c h n o lo g y : V ir tu a l s u p e rm a rk e t (V S M ) Z y g o u ri s S e t a l. , G re e c e , 2 0 1 4 5 1 þþ 8 2 .4 /9 5 .2 þþ þþ Y e s Y e s >
1
2
m

2
1

V
ir
tu

a
l
R
e
a
li
ty

D
a
y
-O

u
t-
T
a
s
k
(V

R
-D

O
T
)

T
a
rn

a
n
a
s
I
e
t
a
l.
S
w
it
z
e
rl
a
n
d
,

2
0
1
3
5
2

2
0

1
0
0
/7
4
.2

þþ
þþ

N
o

Y
e
s

N
A

2
2

C
o
m
p
u
te
ri
s
e
d
C
o
g
n
it
iv
e
S
c
re
e
n
in
g
T
e
s
ts

(C
C
S
)

S
c
a
n
lo
n
L
e
t
a
l.
Ir
e
la
n
d
,
2
0
1
5
5
3

þþ
þþ

< 4 9 4 /6 0 Y e s Y e s 3 m 2 3 C o m p u te ri z e d A s s e s s m e n t o f M il d C o g n it iv e Im p a ir m e n t (C A M C I) T ie rn e y M .C e t a l. , C a n a d a , 2 0 1 4 5 4 �2 8 0 /7 4 þþ þþ N o Y e s 3 0 m 2 4 C o g n it iv e A s s e s s m e n t fo r D e m e n ti a , iP a d v e rs io n (C A D i) O n o d a K e t a l. , Ja p a n , 2 0 1 3 5 5 þþ þþ �7 .6 9 0 /8 4 N o Y e s 1 0 m 2 5 R e v is e d C o g n it iv e A s s e s s m e n t fo r D e m e n ti a , iP a d v e rs io n (C A D i- 2 ) O n o d a K e t a l. , Ja p a n , 2 0 1 4 5 6 þþ þþ þþ 8 5 /8 1 e 9 6 /9 3 N o Y e s 1 0 e 4 0 m 2 6 D e m e n ti a R is k A s s e s s m e n t (D R A ) B ra n d t J e t a l. , U S A , 2 0 1 3 5 7 þþ þþ < 2 9 6 8 /6 7 Y e s Y e s N A 2 7 P a rt ic ip a n t- ra te d (p -A D 8 ) C h in R e t a l. , S in g a p o re , 2 0 1 3 5 8 �1 8 5 /7 4 þþ þþ Y e s N A N A 2 8 In fo rm a n t Q u e s ti o n n a ir e o n C o g n it iv e D e c li n e in th e E ld e rl y in d iv id u a ls (I Q C O D E ) L i F e t a l. , C h in a , 2 0 1 2 5 9 3 .1 9 9 7 .9 /7 1 .4 þþ þþ Y e s Y e s 1 0 m þþ C u t- o ff p o in t, S n a n d S p w e re n o t a p p li c a b le fo r th e s tu d y . M C I, m il d c o g n it iv e im p a ir m e n t; N A , n o t a v a il a b le ; S n , s e n s it iv it y ; s p , s p e c ifi c it y . p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 2 89 published by The Canadian Review of Alzheimer's Disease and Other Dementias.11 Acceptability was assessed using the question ‘Are the items on the test acceptable to patients?’. Judgement required was assessed using the question ‘Can the test be interpreted by non-physicians?’. Cost was assessed using the question ‘What is the cost of the time and staffing required to administer the test?’.11 Results In total, 144 articles published in the five-year period were retrieved. After removing duplicate articles, only 116 remained. These remaining articles were screened for title and abstract, and another 35 articles were excluded due to language, article types and studies of non-primary care set- tings. After evaluating the full text, only 30 articles were included in this review.30e59 This review describes the validity of screening tools for dementia and MCI among the elderly at primary healthcare level. Three types of screening approaches were found: (1) screening by healthcare providers; (2) screening by a self- administered questionnaire; and (3) screening by informa- tion from caretakers. Out of the 30 articles included in this review, 21 studies were based on screening by healthcare providers,30e50 eight studies were of self-administered screening51e58 and only one study was of caretaker infor- mant screening.59 In terms of the setting where the studies were undertaken, 12 studies were conducted in primary care clinics, four in hospital-based clinics, eight at community- based settings and the rest at two or more different settings. The results are summarized in Table 1. Dementia Seventeen of the 30 articles were aimed at describing screening tools to detect dementia among the elderly. These articles covered 19 screening tools for detecting dementia. The Mon- treal Cognitive Assessment (MoCA) and Addenbrooke's Cogni- tive Examination III (ACE-III) were described in more than one article. Comparing both screening tools, ACE-III reported higher sensitivity and specificity than the MoCA in detection of de- mentia.31,32,38,39 Other screening tools reported with high sensitivity and specificity were the Saint Louis University Mental Status (SLUMS), Rapid Cognitive Screen (RCS) and Brief Cognitive Assessment Tool (BCAT).41,46,49 However, these screening tools were less sensitive than ACE-III (Sn ¼ 100%). Most of the screening tools for detecting dementia were re- ported to be feasible for use in community-based screening. Mild cognitive impairment A total of 19 articles within the five years of review described 14 different screening tools for detecting MCI. The MoCA was the most common tool used. There were six studies using MoCA with different cut-off points and different results for sensitivity, specificity and feasibility.30e35 Besides its feasibility, the MoCA was reported with the highest sensitivity and specificity ranges (Sn ¼ 81e97%; Sp ¼ 60e86%). Other studies described different tools as listed in Table 2. The Virtual Reality Day-Out-Task (VR- https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 290 DOT) and the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) were among the most sensitive tools to detect MCI.52,59 However, they were less specific than the MoCA. Finally, the 5 Objects Test, Rapid Cognitive Screen (RCS), Cognitive Performance Scale (CPS) and virtual reality technol- ogy: virtual supermarket (VT-VSM) were among the most spe- cific tools but have low sensitivity.46,47,51 Discussion This systematic review evaluated the validity of various screening tools for dementia and MCIs used on the elderly at primary care level. We categorized our selection of the tools by three different approaches in screening e screening by healthcare providers, by a self-administered questionnaire or by an informant. Based on our review, the ACE-III proved to be the ideal screening tool for detecting dementia based on test accuracy and predictive ability. The ACE-III was observed to be having the highest sensitivity and specificity (Sn/Sp ¼ 100%/96%).39 The ACE-III is the latest version of the Addenbrooke's Cogni- tive Examination Revised (ACE-R).38,39 Reviews based on arti- cles published before 2015 found that the ACE-R is the best screening tool in terms of sensitivity and specificity.60,61 The ACE-III was compared favourably with the ACE-R with extremely high correlation between the scores.39 Sensitivity and specificity of ACE-III remain high using previously rec- ommended cut-off scores.39 For detecting MCI at primary care settings, the MoCA appeared to be a better screening tool than others. With a high sensitivity of 83e97%, the MoCA is considered the screening tool of choice for MCI screening at primary care level. A screening tool should be highly sensitive, but not necessarily specific, to ensure high yield. Those noted as positive from screening should be referred to a second stage for confirma- tion or diagnosis of MCI. A review by Tsoi et al.60 also supports the use of the MoCA as a screening tool for MCI at primary care settings. By using the MMSE as a gold standard, the MoCA reported to have better performance than other MCI screening tests, with 89% sensitivity and 75% specificity.60 However, an earlier review published in 2010 found that different screening tools have different advantages.62 For instance, the MoCA has high sensitivity, the Rowland Uni- versal Dementia Assessment Scale (RUDAS) ignores the in- fluence of culture and education and cognitive drawing test (CDT) is more feasible.62 In the primary care setting, Mini- Cog, MIS or General practitioner assessment of cognition (GPCOG) were the screening tools of choice.62 Another re- view published in 2009 found inconclusive evidence about screening tools that fulfilled the criteria for MCI screening.63 The limitation of this review is that we did not include grey literature and articles in non-English language which may be able to provide a broader scope of information relevant to this review. Nevertheless, this systematic review provides the updated information on dementia or MCI screening tools for use at primary care settings based on the most recent pub- lished literature. The information will be useful to healthcare providers in planning their services towards the aims of early detection and treatment. Conclusion This review found that the ACE-III is a better screening tool for detection of dementia and that the MoCA is the preferred tool for screening of MCIs. This update concurred with previous reviews and was able to illustrate that both tools are still relevant to be used as screening tools for detection of MCI and dementia at primary care level. Author statements Acknowledgements The authors would like to thank the Director General of Health Malaysia, for his kind support and permission to publish this article. Ethics approval This study was approved by the Medical Research Ethical Committee of the National Institute of Health, Ministry of Health Malaysia. Funding None declared. Competing interests The authors declare that they have no competing interests. r e f e r e n c e s 1. Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. The global prevalence of dementia: a systematic review and metaanalysis. Alzheimer's Dementia 2013;9:63e75. 2. World Health Organization. Dementia a public health priority [Internet]. Geneva; London: World Health Organization; Alzheimer’s disease international; 2012 [cited 2014 Oct 24]. Available from: http://whqlibdoc.who.int/publications/2012/ 9789241564458_eng . 3. Chapman DP, Williams SM, Strine TW, Anda RF, Moore MJ. Dementia and its implication for public health. Prev Chronic Dis 2006;3(2):1e13. 4. Kaplan HI, Sadock BJ, Grebb JA. Kaplan and Sadock's synopsis of psychiatry: behavioral sciences, clinical psychiatry. 7th ed. Baltimore (MD): Williams & Wilkins; 1994. 5. Chertkow H, Feldman HH, Jacova C, Massoud F. Definitions of dementia and predementia states in Alzheimer's disease and vascular cognitive impairment: consensus from the Canadian conference on diagnostic of dementia. Alzheimer's Res Ther 2013;5(1):1e8. 6. Ruitenberf A, Ott A, van Swieten JC, Hoffman A, Breteler MM. Incidence of dementia: does gender make a difference. Neurobiol Aging 2001;22(4):575e80. 7. National Institute on Aging, Alzheimer's Disease Fact Sheet, https://www.nia.nih.gov/health/alzheimers-disease-fact- sheet. Retrieved 28th August 2018. 8. Kertesz A, Munoz DG. Frontotemporal dementia. Med Clin North Am 2002;86:501e18. http://refhub.elsevier.com/S0033-3506(19)30001-0/sref1 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref1 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref1 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref1 http://whqlibdoc.who.int/publications/2012/9789241564458_eng http://whqlibdoc.who.int/publications/2012/9789241564458_eng http://refhub.elsevier.com/S0033-3506(19)30001-0/sref3 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref3 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref3 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref3 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref4 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref4 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref4 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref4 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref5 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref6 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref6 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref6 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref6 https://www.nia.nih.gov/health/alzheimers-disease-fact-sheet https://www.nia.nih.gov/health/alzheimers-disease-fact-sheet http://refhub.elsevier.com/S0033-3506(19)30001-0/sref8 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref8 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref8 https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 2 91 9. Cordell CB, Borson S, Boustani M, Chondosh J, Reuben D, Verghese J, et al. Alzheimer's association recommendations for operationalizing the detection of cognitive impairment during the Medicare Annual Welness Visit in a primary care setting. Alzheimer's Dementia 2013;9(2):141e50. 10. Borson S, Frank L, Bayley PJ, Boustani M, Dean M, Lin PJ, et al. Improving dementia care: the role of screening and detection of cognitive impairment. Alzheimer's Dementia 2013;9(2):151e9. 11. Moorhouse P. Screening for dementia in primary care. Can Rev Alzheimer's Dis Other Dementias 2009;12:8e13. 12. Castanho TC, Amorim L, Zihl J, Palha JA, Sousa N, Santos NC. Telephone-based screening tool for mild cognitive impairment and dementia in aging studies: a review of validated instruments. Front Aging Neurosci 2014;6(16):1e17. 13. Galvin JE, Roe CM, Powlishta KK, Coats MA, Muich SJ, Grant E, et al. The AD8; A brief informant interview to detect dementia. Neurology 2005;65(4):559e64. 14. Lorentz WJ, Scanlan JM, Borson S. Brief screening tests for dementia. Can J Psychiatr 2002;47(8):723e33. 15. Society A. Alzheimer's Society online information. Symptoms and diagnosis. Diagnosing dementia. [Internet]. [cited 2015 Jan 5]. Available from: www.alzheimer.org.uk/site/scripts/ documents.php?categoryID¼200346. 16. Brodaty H, Low LF, Gibbson L, Burns K. What is the best dementia screening instrument for general practitioners to use? Am J Geriatr Psychiatr 2006;14(5):391e400. 17. Milne A, Culverwell A, Guss R, Tuppen J, Whelton R. Screening for dementia in primary care: a review of the use, efficacy and quality if measure. Int Psychogeriatr 2008;20(5):911e26. 18. Bradford A, Kunik ME, Schulz P, Williams SP, Singh H. Missed ad delayed diagnosis of dementia in primary care: prevalence and contributing factors. Alzheimer Dis Assoc Disord 2009;23(4):306e14. 19. Paddick S-M, Gray WK, McGuire J, Richardson J, Dotchin C, Walker RW. Cognitive screening tools for identification of dementia in illiterate and low-educated older adults, a systematic review and meta-analysis. Int Psychogeriatr 2017. https://doi.org/10.1017/S1041610216001976. 20. Jackson TA, Naqvi SH, Sheehan B. Screening for dementia in general hospital inpatients: a systematic review and meta- analysis of available instruments. Age Ageing 2013;42:689e95. 21. Martin S, Kelly S, Khan A, Cullum S, Dening T, Rait G, et al. Attitudes and preferences towards screening for dementia: a systematic review of the literature. BMC Geriatr 2015;15(66):1e12. 22. Chen HH, Sun FJ, Yeh TL, Liu HE, Huang HL, Kuo BIT, et al. The diagnostic accuracy of the Ascertain Dementia 8 questionnaire for detecting cognitive impairment in primary care in the community, clinics and hospitals: a systematic review and meta-analysis. Fam Pract 2018;35(3):239e46. 23. Yokomizo JE, Simon SS, de Campos Bottino CM. Cognitive screening for dementia in primary care: a systematic review. Int Psychogeriatr 2014;26(11):1783e804. 24. Lin JS, O'Connor E, Rossom RC, Perdue LA, Eckstrom E. Screening for cognitive impairment in older adult: a systematic review for the U.S. Preventive Services Task Force. Ann Intern Med 2013;159(9):601e12. 25. Boustani M, Peterson B, Harris R, Lux LJ, Krasnov C, Sutton SF, et al. “Screening for Dementia”. Systematic Evidence Review 2003:20. 26. Rosli R, Maw PT, Gray WK, Subramanian P, Chin AV. Cognitive assessment tools in Asia: a systematic review. Int Psychogeriatr 2016;28(2):189e210. 27. Higgins JPT, Deeks JJ. Chapter 7: selecting studies and collecting data. In: Higgins JPT, Green S, editors. Cochrane Handbook for systematic reviews of interventions version 5.1.0 (updated March 2011). The Cochrane Collaboration; 2011. Available from: www.cochrane-handbook.org. 28. Vandenbroucke JP, von Elm E, Altman DG, Gøtzche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. PLoS Med 2007;18(6):805e35. 29. Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6(7):e1000097. pmed.1000097. 30. Julayanont P, Tangwongchai S, Hemrungrojn S, Tunvirachaisakul C, Phanthumchinda K, Hongsawat J, et al. The Montreal Cognitive Assessment-basic: a screening tool for mild cognitive impairment in illiterate and low-educated elderly adults. Am Geriatr Soc 2015;63:2550e4. 31. Felicia CG, Angela VA, Eric M, Olga A, Lavezza Z, Veronique K. Validity of the Montreal Cognitive Assessment as a screen for mild cognitive impairment and dementia in African American. J Geriatr Psychiatr Neurol 2014;27(3):199e203. 32. Dong YH, Lee WY, Basri NA, Collinson SL, Merchant RA, Venketasubramanian N, et al. The Montreal Cognitive Assessment is superior to Mini-Mental State Examination in detecting patients at higher risk of dementia. Int Psychogeriatr 2012;24(11):1749e55. 33. Larner AJ. Screening utility of the Montreal Cognitive Assessment (MoCA): in place of -or as well as- the MMSE. Int Psychogeriatr 2012;24(3):391e6. 34. Freitas S, Simoes MR, Alves L, Santana I. Montreal cognitive assessment validation study for mild cognitive impairment and Alzheimer disease. Alzheimer Dis Assoc Disord 2013;27(1):37e43. 35. Raolf DR, Moberg PJ, Xie SX, Wolk DA, Moelter ST, Arnold SE. Comparative accuracies of two common screening instruments for the classification of Alzheimer's disease, mild cognitive impairment and healthy aging. Alzheimer's Dementia 2013;9(5):529e37. 36. Chetna M, Angelique C, David M, Dennis S, Adeline C, Young KD. Diagnostic performance of short portable mental status questionnaire for screening dementia among patients attending cognitive assessment clinics in Singapore,. Ann Acad Med Singapore 2013;42:315e9. 37. Delgado DC, Guerrero BS, Troncoso PM, Araneda YA, Slachevsky CA, Behrens PMI. Memory, fluency, and orientation (MEFO): a five-minute screening test for cognitive decline. Neurologia 2013;28(7):400e7. 38. Michael TJ, Jonathan JE. An investigation of the utility of the Addenbrooke's Cognitive Examination III in the early detection of dementia in Memory Clinic patientsaged over 75 years. Dement Geriatr Cognit Disord 2015;40:222e32. 39. Hsieh S, Schubert S, Hoon C, Mioshi E, Hodges JR. Validation of the Addenbrooke's cognitive examination III in frontotemporal dementia and Alzheimer's disease. Dement Geriatr Cognit Disord 2013;36:242e50. 40. Takahashi F, Awata S, Sakuma N, Inagaki H, Ijuin M. Reliability and validity of a Quick Test of Cognitive Speed for detecting early-stage dementia in elderly Japanese. Psychogeriatrics 2012;12:75e82. 41. Cummings-Vaughn LA, Chavakula NN, Malmstrom TK, Tumosa N, Morley JE, Cruz-Oliver DM. Veterans affairs Saint Louis University mental status examination compared with the Montreal Cognitive Assessment and the short test of mental status. Am Geriatr Soc 2014;62:1341e6. 42. Papageorgiou SG, Economou E, Routsis C. The 5 objects test: a novel, minimal-language, memory screening test. J Neurol 2013;261(2):422e31. 43. Serna A, Contador I, Bermejo-Pareja F, Mitchell AJ, Fern�andez-Calvo B, Ramos F, et al. Accuracy of a brief neuropsychological battery for the diagnosis of dementia and http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref9 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref10 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref11 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref12 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref13 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref13 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref13 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref13 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref14 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref14 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref14 http://www.alzheimer.org.uk/site/scripts/documents.php?categoryID=200346 http://www.alzheimer.org.uk/site/scripts/documents.php?categoryID=200346 http://www.alzheimer.org.uk/site/scripts/documents.php?categoryID=200346 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref16 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref17 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref18 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref18 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref18 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref18 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref18 https://doi.org/10.1017/S1041610216001976 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref20 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref21 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref21 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref21 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref21 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref21 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref22 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref23 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref23 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref23 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref23 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref24 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref24 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref24 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref24 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref24 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref25 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref26 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref26 http://www.cochrane-handbook.org http://refhub.elsevier.com/S0033-3506(19)30001-0/sref28 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref28 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref28 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref28 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref28 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref29 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref30 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref31 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref32 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref33 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref34 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref34 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref34 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref34 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref34 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref35 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref36 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref37 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref37 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref37 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref37 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref37 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref38 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref39 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref40 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref41 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref42 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 8 4 e9 292 mild cognitive impairment: an analysis of the NEDICES Cohort. J Alzheim Dis 2015;48:163e73. 44. Ramlall S, Chipps J, Bhigjee AI, Pillay BJ. The sensitivity and specificity of subjective memory complaints and the subjective memory rating scale, deterioration cognitive observe, mini-mental state examination, six-item screener and clock drawing test in dementia screening. Dement Geriatr Cognit Disord 2013;36:119e35. 45. Martins de Oliveira G, Yokomizo JE, Vinholi e Silva LdS, Saran LF, Bottino CMC, Yassuda MS. The applicability of the cognitive abilities screening instrument-short (CASI-S) in primary care in Brazil. Int Psychogeriatr 2016;28(1):93e9. 46. Malmstrom TK, Voss VB, Cruz-Oliver DM, Cummings- Vaughn LA, Tumosa N, Grossberg GT, et al. The Rapid Cognitive Screen (RCS): a point-of-care screening for dementia and mild cognitive impairment. J Nutr Health Aging 2015;19(7):741e4. 47. Wellens NIH, Flamaing J, Tournoy J, Hanon T, Moons P, Verbeke G, et al. Convergent validity of the cognitive performance scale of the interRAI acute care and the mini- mental state examination. Am J Geriatr Psychiatr 2013;21(7):636e45. 48. Shim YS, Ryu HJ, Lee DW, Jeong JH, Choi SH, Han SH, et al. Literacy Independent Cognitive Assessment: assessing mild cognitive impairment in older adults with low literacy skills. Psychiatr Invest 2015;12(3):341e8. 49. Mansbach WE, Mace RA, Clark KM. Differentiating levels of cognitive functioning: a comparison of the Brief Interview for Mental Status (BIMS) and the Brief Cognitive Assessment Tool (BCAT) in a nursing home sample. Aging Ment Health 2014;18(7):921e8. 50. Holsinger T, Plassman BL, Stechuchak KM, Burke JR, Coffman CJ, Williams JW. Screening for cognitive impairment: comparing the performance of four instruments in primary care. J Am Geriatr Soc 2012;60:1027e36. 51. Zygouris S, Giakoumis D, Votis K, Doumpoulakis S, Ntovas K, Segkouli S, et al. Can a virtual reality cognitive training application fulfill a dual role? Using the virtual supermarket cognitive training application as a screening tool for mild cognitive impairment. J Alzheim Dis 2015;44:1333e47. 52. Tarnanas I, Schlee W, Tsolaki M, Muri R, Mosimann U, Nef T. Ecological validity of virtual reality daily living activities screening for early dementia: longitudinal study. JMIR Serious Games 2013;1(1). https://doi.org/10.2196/games.2778, 2013. 53. Scanlon L, O'Shea E, O'Caoimh R, Timmons S. Usability and validity of a battery of computerised cognitive screening tests for detecting cognitive impairment. Gerontology 2016;62(2):247e52. 54. Tierney MC, Naglie G, Upshur R, Moineddin R, Charles J, Jaakkimainen RL. Feasibility and validity of the self- administered computerized assessment of mild cognitive impairment with older primary care patients. Alzheimer Dis Assoc Disord 2014;28(4):311e9. 55. Onoda K, Hamano T, Nabika Y, Aoyama A, Takayoshi H, Nakagawa T, et al. Validation of a new mass screening tool for cognitive impairment: cognitive assessment for dementia, iPad version. Clin Interv Aging 2013;8:353e60. 56. Onoda K, Yamaguchi S. Revision of the cognitive assessment for dementia, iPad version (CADi2). PLoS One 2014;9(10):1e6. 57. Brandt J, Sullivan C, Burrell II LE, Rogerson M, Anderson A. Internet-based screening for Dementia risk. PLos One 2013;8(2):1e7. 58. Chin R, Ng A, Narasimhalu K, Kandiah N. Utility of the AD8 as a self-rating tool for cognitive impairment in Asian population. Am J Alzheimer's Dis Other Dementias 2013;28(3):284e8. 59. Li F, Jia XF, Jia J. The Informant Questionnaire on cognitive decline in the elderly individuals in screening mild cognitive impairment with or without functional impairment. J Geriatr Psychiatr Neurol 2012;25(4):227e32. 60. Tsoi KKF, Chan JCC, Hirai HW, Wong SYS, Kwok TCY. Cognitive test to detect dementia: a systematic review and meta-analysis. JAMA Intern Med 2015;175(9):1450e8. 61. Lischka AR, Mendelsohn M, Overend T, Forbes D. A systematic review of screening tools for predicting the development of dementia. Can J Aging 2012;31(3):295e311. 62. Ismail Z, Rajji TK, Shulman KI. Brief cognitive screening Instruments: an update. Int J Geriatr Psychiatr 2010;25:111e20. 63. Lonie JA, Tierney KM, Ebmeier KP. Screening for mild cognitive impairment: a systematic review. Int J Geriatr Psychiatr 2009;24:902e15. http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref43 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref44 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref45 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref45 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref45 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref45 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref45 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref46 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref47 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref48 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref48 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref48 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref48 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref48 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref49 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref50 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref50 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref50 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref50 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref50 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref51 https://doi.org/10.2196/games.2778, 2013 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref53 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref53 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref53 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref53 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref53 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref54 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref55 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref55 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref55 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref55 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref55 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref56 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref56 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref56 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref57 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref57 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref57 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref57 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref58 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref58 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref58 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref58 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref59 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref59 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref59 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref59 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref59 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref60 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref60 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref60 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref60 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref61 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref61 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref61 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref61 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref62 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref62 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref62 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref63 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref63 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref63 http://refhub.elsevier.com/S0033-3506(19)30001-0/sref63 https://doi.org/10.1016/j.puhe.2019.01.001 https://doi.org/10.1016/j.puhe.2019.01.001 Validity of screening tools for dementia and mild cognitive impairment among the elderly in primary health care: a systemat ... Introduction Methods Search strategy Inclusion and exclusion criteria Screening and review process Results Dementia Mild cognitive impairment Discussion Conclusion Author statements Acknowledgements Ethics approval Funding Competing interests References Social-gradients-in-health-and-social-care-costs--Analysis-of-l_2019_Public- ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4 Available online at w Public Health journal homepage: www.elsevier.com/puhe Themed Papere Original Research Social gradients in health and social care costs: Analysis of linked electronic health records in Kent, UK W. Jayatunga a,*, M. Asaria b, A. Belloni c, A. George d, T. Bourne d, Z. Sadique a a London School of Hygiene and Tropical Medicine Keppel St, Bloomsbury, London, UK b Centre for Health Economics, University of York, Heslington, York, UK c Public Health England, Wellington House, 133-155 Waterloo Road, London, UK d Kent County Council, Sessions House, County Hall, Maidstone, Kent, UK a r t i c l e i n f o Article history: Received 16 April 2018 Received in revised form 8 October 2018 Accepted 4 February 2019 Available online 12 March 2019 Keywords: Inequality Deprivation Cost Utilisation Expenditure Healthcare Social care Economics * Corresponding author. 3 Constable Mews, B E-mail address: wikumj@gmail.com (W. J https://doi.org/10.1016/j.puhe.2019.02.007 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: Research into the socio-economic patterning of health and social care costs in the UK has so far been limited to examining only particular aspects of healthcare. In this study, we explore the social gradients in overall healthcare and social care costs, as well as in the disaggregated costs by cost category. Study design: We calculated the social gradient in health and social care costs by cost category using a linked electronic health record data set for Kent, a county in South East England. We performed a cross-sectional analysis on a sample of 323,401 residents in Kent older than 55 years to assess the impact of neighbourhood deprivation on mean annual per capita costs in 2016/17. Methods: Patient-level costs were estimated from activity data for the financial year 2016/17 and were extracted alongside key patient characteristics. Mean costs were calculated for each area deprivation quintile based on the index of multiple deprivation of the neighbourhood (lower super output area) in which the patient lived. Cost subcategories were analysed across primary care, secondary care, social care, community care and mental health. Results: The mean annual per capita cost increased with deprivation across each depriva- tion quintile, with a cost of £1205 in the most affluent quintile, compared with £1623 in the most deprived quintile, a 35% cost increase. Social gradients were found across all cost subcategories. Conclusions: Health inequalities in the population older than 55 years in Kent are associated with healthand socialcare costs of £109m, equivalent to15% ofthe estimated total expenditure in this age group. Such significant costs suggest that appropriate interventions to reduce socio- economic inequalities have the potential to substantially improve population health and, depending on how much investment they require, may even result in cost savings. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. romley, BR1 3BF, UK. Te ayatunga). ic Health. Published by E l.: þ447872959152. lsevier Ltd. All rights reserved. mailto:wikumj@gmail.com http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.007&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4 189 Introduction Health inequalities have been described as ‘the systematic dif- ferences in the health of people occupying unequal positions in so- ciety’, for example, due to differences in income, education, occupation, material resources and social status.1 Reducing these inequalities has become a key policy objective both in the UK2 and internationally,3 but despite this, health in- equalities remain persistent and progress in reducing them has been a challenge.4,5 Despite the vast literature demonstrating the existence of health inequalities, there has been less research into their impact on healthcare costs in England. A recent study of na- tional hospital data by Asaria et al.6 found that inpatient costs in England in 2011/12 were 31% higher for patients in the most deprived quintile than for those in the most affluent quintile and estimated that the total annual cost associated with this inequality was £4.8bn. Another study of inpatient hospital costs by Kelly et al.7 found a 35% difference in costs between the most and least deprived quintiles, in patients older than 65 years. A study by Charlton et al.8 on primary care data in the UK found that deprivation was associated with greater morbidity and increased healthcare costs. No studies were found exploring this relationship on services outside of sec- ondary care and primary care. There has also been research on the relationship between deprivation and healthcare utilisation, from which the im- pacts on costs can be reasonably inferred. Reviews of the literature by Dixon et al.,9 Goddard and Smith10 and Cookson et al.11 conclude that deprived groups tend to consume more healthcare due to greater health needs. However, these in- equities vary by service: in general, poorer populations tend to use more general practitioner (GP) services, relative to need, than affluent groups but are less likely to be referred on for specialist elective care. Uptake of health promotion and pre- ventative services was also found to be lower in areas of high deprivation.10,11 In multiple studies, deprivation has been found to be a strong predictor of accident and emergency (A&E) attendance and hospital admission.12e15 The studies' authors suggest many possible reasons for this: increased need for healthcare, less capability for self-care, lack of awareness or under- standing of the most appropriate health services and lower uptake of preventative services. This demonstrates the importance of looking at impacts between different services because they may be linked: lower use of preventative ser- vices may lead to higher use of emergency services. For example, one study showed that deprived populations had higher A&E attendance but lower use of the National Health Service (NHS) telephone line, ‘NHS Direct’.16 Goddard and Smith's10 review describes the difficulties in capturing the impacts of deprivation across the wide range of complemen- tary and substitute services involved in long-term care, such as social care, due to the complexity of different providers involved and differing funding streams. At present, social care in the UK is funded from local authority budgets rather than via the NHS. Given the policy goals of the NHS to better integrate care between these different sectors,17 it would be informative to assess the system-wide association between deprivation and costs. We found no literature on the socio-economic patterning of social care or community care costs. However, given that it is well established that there is a higher preva- lence of multimorbidity and chronic long-term conditions in deprived populations,18e20 we would expect this to be reflected in higher community care and social care costs in deprived groups. Similarly, there was also no literature on the associa- tion between deprivation and the cost of mental health ser- vices. Again, we know that the prevalence of mental health conditions is associated with deprivation,18,20,21 and so we would also expect a social gradient in mental health expendi- ture with higher costs for those living in more deprived areas. The difficulties in analysing system-wide impacts can be overcome through the analysis of linked electronic health records. The Kent Integrated Dataset (KID) is a ‘whole-popu- lation’ database, developed by Kent's local authority public health team since 2014, which links patient-level data across primary, secondary, community, mental health and social care while anonymising personal data.22 The database in- cludes data for all residents of Kent and from most of the healthcare and social care providers in the area, linked by means of the patients' NHS number as a common identifier. This study evaluates the association between socio- economic deprivation and annual per capita costs of health and social care in Kent. Previous studies at a patient level have tended to focus on a particular type of cost, such as hospital costs or primary care costs. The more comprehensive nature of this study and the disaggregated analysis by cost category is important because there may be differential impacts of deprivation across care settings, and impacts on one part of the system may be compensated for by impacts on other parts of the system. Methods Patient data were extracted from the KID using Microsoft SQL Server. Age is known to be a key determinant of healthcare expenditure, with older people more likely to utilise health- care and social care services. Because of this, the study was restricted to people older than 55 years, as a group with high care costs overall. Therefore, the inclusion criteria were peo- ple older than 55 years and currently alive, with a registered address in Kent, as of 1st May 2017. From this population list (502,675), some were excluded because of gaps in cost data: 85 of 238 GP practices in Kent were not flowing activity data into the KID during the study period, and we therefore excluded patients registered to these practices. This resulted in a study sample of 323,401, which is 63% of the total population older than 55 years. The English Index of Multiple Deprivation (IMD) was used as an area-based measure of deprivation, by quintile, for each patient based on lower super output area (LSOA) of residence. LSOAs are a standardised geographical unit for reporting small area statistics whereby each LSOA has around 1500 residents that are relatively socially homogenous. There are 32,844 LSOAs in England and 902 in Kent. Kent is a large county in South East England, with 1.6 million residents from a wide spectrum of social backgrounds; certain areas in Kent https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4190 feature among the most deprived and the most affluent in the country. Areas of deprivation tend to be concentrated around urban centres and in particular the eastern coastal towns, whereas areas of affluence are often found in the more rural parts of the county. Healthcare costs for each patient in the financial year 2016/17 were estimated from utilisation activity, taken from providers across primary care, secondary care, community care, social care and mental health services. Unit costs in the database were calculated in various ways across these sectors of care.23 Primary care unit care costs were taken from the Personal Social Services Research Unit (PSSRU) manual of reference costs.24 Secondary care unit care costs were taken from the national tariff price for that activity. For community care and mental care, which are commissioned by block contracts, unit costs were taken as the mean costs of activity; calculated by dividing the total sum of the contract by the throughput of activity. Social care costs were taken from Kent County Council's ‘SWIFT’ database which includes monthly billed invoices for each person receiving social care services. The database attempts to include all costs across health and social care in Kent, although there are some gaps, as described in Appendix 1. Mean total costs were calculated for each deprivation quintile using IBM SPSS version 23. Generalised linear model (GLM) regression was also performed to assess the role of age and gender as potential confounders, although this was found to have little impact on the relationship between mean costs and deprivation quintile (Appendix 2). Mean costs were also calculated across the cost subcategories. For each cost cate- gory, the total costs in the Kent population associated with deprivation were calculated using the formula: X4 i¼1 ðPQi*ðCQi � CQ5ÞÞ where � P ¼ population size � C ¼ mean cost � Qi ¼ deprivation quintile ‘i’ � Q5 ¼ most affluent quintile � P ¼ sum for deprivation quintiles 1 to 4. This gives, for each cost category, the hypothetical reduc- tion in healthcare and social care costs in Kent if the social Table 1 e Study population. Characteristic Study Population size 323,401 Age Mean age 68.9 ye Gender Male 152,77 Female 170,62 Deprivation Q1 e most deprived 29,64 Q2 52,72 Q3 77,18 Q4 87,39 Q5 e least deprived 76,44 gradient in costs was eliminated, that is, if the whole popu- lation older than 55 years had the same mean per capita costs as those living in the most affluent quintile of LSOAs. Results The study sample was 323,401 (Table 1). This compares to a whole population of 512,120 people in Kent older than 55 years and follows exclusion of patients registered to GP practices not flowing data to the KID. The sample was highly repre- sentative of the overall population of Kent, with very similar mean age, gender split and distribution among the depriva- tion quintiles. Costs increased with each deprivation quintile, with mean annual cost of £1623 for people living in the most deprived quintile compared with £1205 in the most affluent quintile. This difference of £418 represents a 35% increase in per capita costs between the least and most deprived quintiles (Table 2). A social gradient is observed across all cost categories (Fig. 1). Secondary care costs, as the largest component of per capita costs, increased by £141 between the least and most deprived quintiles, an increase of 27% (Table 2). Social care costs increased by £121 (47%), and primary care costs increased by £74 (26%). Mental health and community care are smaller components of overall per capita costs, but the cost increases (£44 and £37, respectively) represent steep social gradients (66% and 54%, respectively). Overall, the cost variation by deprivation is associated with about £111m of additional costs across Kent, representing 15% of the total healthcare and social care costs in the Kent pop- ulation older than 55 years (Table 3). In absolute terms, the largest of these additional costs by cost category are in sec- ondary care and social care (£37m and £39m, respectively). When looked at in relative terms, larger proportions of the overall costs in social care, community care and mental health are associated with deprivation (23%, 22% and 27%, respec- tively) than with secondary care (12%) and primary care (8%) (Table 3). Discussion The annual mean per capita cost was £1629 in the most deprived quintile compared with £1211 in the least deprived quintile. There was a clear social gradient in mean costs sample Kent population (older than 55 years) 512,120 ars 69.0 years 3 (47.2%) 241,652 (47.2%) 8 (52.8%) 270,468 (52.8%) 6 (9.2%) 53,414 (10.4%) 1 (16.3%) 82,405 (16.1%) 6 (23.9%) 128,380 (25.1%) 7 (27.0%) 138,546 (27.1%) 8 (23.6%) 109,375 (21.4%) https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 T a b le 2 e M e a n a n n u a l p e r ca p it a co s ts b y d e p ri v a ti o n q u in ti le a n d ca re s e ct o r. D e p ri v a ti o n S e c o n d a ry c a re c o s ts S o c ia l c a re c o s ts P ri m a ry c a re c o s ts C o m m u n it y c a re c o s ts M e n ta l h e a lt h c o s ts T o ta l c o s ts Q 5 £ 5 2 3 e e £ 2 5 8 e e £ 2 9 1 e e £ 6 7 e e £ 6 6 e e £ 1 ,2 0 5 e e Q 4 £ 5 9 1 £ 6 7 1 2 .9 0 % £ 3 2 8 £ 7 0 2 7 .0 0 % £ 3 0 6 £ 1 5 5 .3 0 % £ 8 2 £ 1 5 2 2 .2 0 % £ 9 0 £ 2 3 3 5 .0 0 % £ 1 3 9 7 £ 1 9 2 1 5 .9 3 % Q 3 £ 6 2 3 £ 9 9 1 9 .0 0 % £ 3 5 1 £ 9 2 3 5 .8 0 % £ 3 1 9 £ 2 8 9 .6 0 % £ 8 5 £ 1 8 2 6 .6 0 % £ 9 4 £ 2 8 4 1 .9 0 % £ 1 4 7 2 £ 2 6 7 2 2 .1 6 % Q 2 £ 6 1 3 £ 9 0 1 7 .2 0 % £ 3 8 8 £ 1 3 0 5 0 .2 0 % £ 3 2 3 £ 3 2 1 1 .0 0 % £ 1 0 8 £ 4 1 6 0 .1 0 % £ 1 0 8 £ 4 2 6 3 .2 0 % £ 1 5 4 0 £ 3 3 5 2 7 .8 0 % Q 1 £ 6 6 4 £ 1 4 1 2 7 .0 0 % £ 3 8 0 £ 1 2 1 4 7 .0 0 % £ 3 6 5 £ 7 4 2 5 .5 0 % £ 1 0 4 £ 3 7 5 4 .3 0 % £ 1 1 0 £ 4 4 6 5 .6 0 % £ 1 6 2 3 £ 4 1 8 3 4 .6 9 % B o ld v a lu e s re p re s e n t m e a n a n n u a l p e r c a p it a c o s ts . It a li c is e d v a lu e s re p re s e n t in c re a s e in m e a n c o s ts re la ti v e to Q 5 . % ¼ p e rc e n ta g e in c re a s e in m e a n c o s ts re la ti v e to Q 5 . Q 5 ¼ m o s t a ffl u e n t; Q 1 ¼ m o s t d e p ri v e d . p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4 191 across all deprivation quintiles and all cost subcategories. The results for secondary care, an increase in costs of 27% between the most and least deprived quintiles, are similar to those in the literature (secondary care being the only cost category for which comparable literature exists). One study of all-age na- tional hospital costs found an increase of 31% between the most and least deprived quintiles,6 and another study of na- tional hospital costs in patients aged 65 years or older found an increase of 35%.7 A key contribution of this analysis is in highlighting the even steeper socio-economic gradients in other care sectors, particularly mental health (66%), commu- nity care (54%) and social care (47%). Variation in healthcare costs could relate to a multitude of factors, such as health needs, access to services and demand/ utilisation of services.9e11 Given that the association between socio-economic deprivation and ill-health is already well- established,2 health needs are likely to be the most impor- tant of these factors. In Kent, it is known that the more deprived populations have higher social risks for poor health, higher prevalence of diagnosed conditions and higher rates of premature mortality.25 This study demonstrates the cost im- plications of these health inequalities to the health and social care system in Kent. Populations with high prevalence of chronic long-term conditions and multimorbidity are likely to have high care needs with regards to both social care and community care, which may explain the steep social gradients observed. Social care, unlike the other categories, is means-tested rather than universally available, so the cost gradient here may partly be due to a greater proportion of deprived populations being eligible to receive state-funded care. Mental illness is also known to be strongly associated with deprivation,18,20,21 perhaps relating to the social circum- stances of those living in deprived areas, such as financial hardship and difficulties with accommodation and employ- ment. The steep social gradients in mental health costs (66%) suggest that targeted interventions in deprived areas are needed to improve population mental health, perhaps through addressing these wider social determinants of mental illness. Secondary care costs are the largest component of per capital health and care costs, although the socio-economic gradient is less than in other sectors in relative terms. The literature indicates that deprived populations have higher use of emergency services and lower utilisation of elective and specialist services, and affluent populations tend to consume more preventative care and present at an earlier stage of illness.11 This study did not distinguish between elective and emergency care, and so expected gradients, which run in opposite directions, may be partially balancing out. Reducing demand for emergency services is a key policy objective for the NHS, and this could be achieved in deprived populations by ensuring comprehensive preventative care both in primary care and specialist elective care. The strengths of this study include its very large sample size which was representative of the Kent population. As Kent is a large region of the country with areas of both deprivation and affluence, the findings may be broadly generalisable to the rest of the country. Another strength is the range and breadth of healthcare activity included, from different care https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 Fig. 1 e Mean annual per capita costs by deprivation quintile for each sector of care. IMD, Index of Multiple Deprivation. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4192 sectors, which distinguishes this analysis from existing liter- ature on the topic. The main limitation of this study is that use of an area- based measure of deprivation risks the ecological fallacy; just because someone lives in a deprived area does not mean they themselves are deprived, and vice versa. However, indi- vidual measures of deprivation cannot be easily linked to routine health service data. Furthermore, this is a cross- Table 3 e Costs attributable to deprivation in the Kent populat Cost category Secondary care Social care Costs associated with deprivation £36.9m £38.7m Total costs £305.m £171.m Proportion of costs associated with deprivation 12.1% 22.6% sectional analysis, in which deprivation status is based on the area in which someone is currently living. This is a snapshot measure, which does not account for the fact that people may have recently moved house into or out of deprived areas. Another factor that could not be analysed from the routine data available was patient utilisation of private healthcare services, which would act as a substitute for NHS care. Roughly 11% of the UK population has some form of ion older than 55 years. Primary care Community care Mental health Total £12.3m £9.7m £12.6m £110.8m £161.3m £44.1m £46.5m £727.9m 7.6% 22.1% 27.1% 15.2% https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4 193 health insurance,26 and because more affluent populations are more likely to have insurance, this study's findings of lower NHS costs in affluent groups may in part be a reflection of this. On the other hand, only patients registered to a GP are included in the analysis. This means that vulnerable groups who are less likely to be registered in primary care (such as asylum-seekers, ex-offenders and the homeless) may be under-represented, despite the fact that these groups have high health needs.27 This might therefore lead to an under- estimate of the costs associated with deprivation. The findings of this study suggest that socio-economic in- equalities are associated with around 15% of overall healthcare and social care costs in those older than 55 years in Kent, ranging from 7.6% for overall primary care costs to 27.1% for overall mental care costs. If the Kent population is represen- tative of the national picture and if the relationship applies to other age groups and costs more broadly, this would mean that inequalities are associated with £674m of the £8.88bn spent in primary care,28 £2.63bn of the £9.72bn spent on mental health29 and £8.64bn of the £71.4bn spent in secondary care.30 While reducing inequalities in health is often seen as a moral imperative, the results of this study indicate that it may also result in significant cost savings on healthcare and social care systems. However, this would depend on the level of public expenditure required to reduce health inequalities (assuming that effective interventions exist), and fully elimi- nating social gradients is probably unrealistic. Nonetheless, the findings suggest that health resources could be better redistributed to address health inequalities; preventative in- terventions targeted towards populations in deprived areas might reduce the onset of ill-health in these groups, leading to savings to both the NHS and local authorities. Other upstream interventions that address the social determinants of health should be explored to tackle the primary causes of deprivation in the first place. Public health professionals could use the results of this study to make a stronger economic case for policy action to reduce inequalities in health. Author statements Acknowledgements Thanks to Gerard Abi-Aad for hosting this analysis at the Kent Public Health Observatory in Kent County Council. Ethical approval This study was approved by London School of Hygiene and Tropical Medicine MSc Research Ethics Committee, and also Kent County Council. Funding None declared. Competing interests None declared. r e f e r e n c e s 1. McCartney G, Collins C, Mackenzie M. What (or who) causes health inequalities: theories, evidence and implications? Health Policy 2013 Dec;113(3):221e7. 2. Marmot MG. Fair Society, Healthy Lives: The marmot review. [Internet]. London: UCL; 2010 [cited 2017 Aug 17]. Available from: http://www.ucl.ac.uk/gheg/marmotreview/ FairSocietyHealthyLives. 3. WHO Commission on Social Determinants of Health. World health organization. In: Closing the gap in a generation: health equity through action on the social determinants of health: commission on Social Determinants of Health final report. Geneva, Switzerland: World Health Organization, Commission on Social Determinants of Health; 2008. 4. Mackenbach JP. The persistence of health inequalities in modern welfare states: the explanation of a paradox. Soc Sci Med 2012 Aug;75(4):761e9. 5. Newton JN, Briggs ADM, Murray CJL, Dicker D, Foreman KJ, Wang H, et al. Changes in health in England, with analysis by English regions and areas of deprivation, 1990e2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015 Dec 5;386(10010):2257e74. 6. Asaria M, Doran T, Cookson R. The costs of inequality: whole- population modelling study of lifetime inpatient hospital costs in the English National Health Service by level of neighbourhood deprivation. J Epidemiol Community Health 2016 Oct;70(10):990e6. 7. Kelly E, Stoye G, Vera-Hernandez M. Public hospital spending in England: evidence from National Health Service administrative records. Fisc Stud 2016;37:433e59. 8. Charlton J, Rudisill C, Bhattarai N, Gulliford M. Impact of deprivation on occurrence, outcomes and health care costs of people with multiple morbidity. J Health Serv Res Pol 2013 Oct;18(4):215e23. 9. Dixon A, Le Grand J, Henderson J, Murray R, Poteliakhoff E. Is the British National Health Service equitable? The evidence on socioeconomic differences in utilization. J Health Serv Res Pol 2007 Apr 1;12(2):104e9. 10. Goddard M, Smith P. Equity of access to health care services. Soc Sci Med 2001 Nov 1;53(9):1149e62. 11. Cookson R, Propper C, Asaria M, Raine R. Socio-economic inequalities in health care in England. Fisc Stud 2016;37(3e4):371e403. 12. Scantlebury R, Rowlands G, Durbaba S, Schofield P, Sidhu K, Ashworth M. Socioeconomic deprivation and accident and emergency attendances: cross-sectional analysis of general practices in England. Br J Gen Pract 2015 Oct 1;65(639):e649e54. 13. Majeed A, Bardsley M, Morgan D, O'Sullivan C, Bindman AB. Cross sectional study of primary care groups in London: association of measures of socioeconomic and health status with hospital admission rates. BMJ 2000 Oct 28;321(7268):1057e60. 14. Baker R, Bankart MJ, Rashid A, Banerjee J, Conroy S, Habiba M, et al. Characteristics of general practices associated with emergency-department attendance rates: a cross-sectional study. BMJ Qual Saf 2011 Nov 1;20(11):953e8. 15. Petrou S, Kupek E. Socioeconomic differences in childhood hospital inpatient service utilisation and costs: prospective cohort study. J Epidemiol Community Health Lond 2005 Jul;59(7):591. 16. Shah SM, Cook DG. Socio-economic determinants of casualty and NHS Direct use. J Public Health 2008 Mar 1;30(1):75e81. 17. Five Year Forward View [Internet]. NHE England. 2014 [cited 2017 Aug 22]. Available from: https://www.england.nhs.uk/ wp-content/uploads/2014/10/5yfv-web . http://refhub.elsevier.com/S0033-3506(19)30031-9/sref1 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref1 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref1 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref1 http://www.ucl.ac.uk/gheg/marmotreview/FairSocietyHealthyLives http://www.ucl.ac.uk/gheg/marmotreview/FairSocietyHealthyLives http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref3 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref4 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref4 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref4 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref4 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref5 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref6 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref7 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref7 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref7 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref7 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref8 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref8 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref8 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref8 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref8 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref9 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref9 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref9 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref9 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref9 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref10 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref10 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref10 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref11 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref11 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref11 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref11 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref11 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref12 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref12 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref12 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref12 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref12 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref13 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref14 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref14 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref14 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref14 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref14 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref15 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref15 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref15 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref15 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref16 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref16 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref16 https://www.england.nhs.uk/wp-content/uploads/2014/10/5yfv-web https://www.england.nhs.uk/wp-content/uploads/2014/10/5yfv-web https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 8 8 e1 9 4194 18. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet 2012 Jul 7;380(9836):37e43. 19. Dalstra J a A, Kunst AE, Borrell C, Breeze E, Cambois E, Costa G, et al. Socioeconomic differences in the prevalence of common chronic diseases: an overview of eight European countries. Int J Epidemiol 2005 Apr 1;34(2):316e26. 20. McLean G, Gunn J, Wyke S, Guthrie B, Watt GC, Blane DN, et al. The influence of socioeconomic deprivation on multimorbidity at different ages: a cross-sectional study. Br J Gen Pract 2014 Jul 1;64(624):e440e7. 21. Brilleman SL, Purdy S, Salisbury C, Windmeijer F, Gravelle H, Hollinghurst S. Implications of comorbidity for primary care costs in the UK: a retrospective observational study. Br J Gen Pract 2013 Apr;63(609):e274e82. 22. Caincross L, Sinclair C. Transforming social care through the use of information and technology. Local Government Association; 2016. 23. George A, Bourne Tom. The kent integrated dataset [Internet]. Kent County Council; 2017 [cited 2018 Apr 6]. Available from: https://www.kpho.org.uk/__data/assets/pdf_file/0004/74146/ Kent-Integrated-Dataset-August-2017 . 24. Curtis L. Unit Costs of Health and Social Care 2016 V. Sources of information. Chapter 16: Inflation Indices. [Internet]. PSSRU; 2016. Available from: http://www.pssru.ac.uk/project-pages/unit- costs/unit-costs-2016/. 25. Jayatunga W, Kennard R. Mind the gap: health inequalities action plan for Kent analytical report [Internet]. Kent Public Health Observatory; 2016 [cited 2017 Sep 5]. Available from: http:// www.kpho.org.uk/__data/assets/pdf_file/0011/58835/Mind- the-Gap-Analytical-Report-D2 . 26. The King’s Fund. The UK private health market [Internet]. 2014 [cited 2017 Sep 6]. Available from: https://www.kingsfund.org. uk/sites/default/files/media/commission-appendix-uk- private-health-market . 27. Davies AR, Chitnis X, Bardsley M. Hospital activity and cost incurred because of unregistered patients in England: considerations for current and new commissioners. J Public Health 2013 Dec 1;35(4):590e7. 28. NHS Digital. NHS payments to general practice, England. 2016/17 [Internet]. [cited 2018 Mar 21]. Available from: https://digital. nhs.uk/catalogue/PUB30089. 29. NHS England. Mental health five year forward view dashboard Q1 and Q2 2017/18 - summary [Internet]. 2018 [cited 2018 Mar 21]. Available from: https://www.england.nhs.uk/publication/ mental-health-five-year-forward-view-dashboard/. 30. ONS. UK health accounts: 2015 - office for national statistics [Internet]. 2017 [cited 2018 Mar 27]. Available from: https:// www.ons.gov.uk/peoplepopulationandcommunity/ healthandsocialcare/healthcaresystem/bulletins/ ukhealthaccounts/2015. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.puhe.2019.02.007. http://refhub.elsevier.com/S0033-3506(19)30031-9/sref18 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref18 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref18 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref18 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref18 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref19 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref19 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref19 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref19 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref19 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref20 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref20 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref20 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref20 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref20 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref21 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref21 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref21 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref21 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref21 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref22 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref22 https://www.kpho.org.uk/__data/assets/pdf_file/0004/74146/Kent-Integrated-Dataset-August-2017 https://www.kpho.org.uk/__data/assets/pdf_file/0004/74146/Kent-Integrated-Dataset-August-2017 http://www.pssru.ac.uk/project-pages/unit-costs/unit-costs-2016/ http://www.pssru.ac.uk/project-pages/unit-costs/unit-costs-2016/ http://www.kpho.org.uk/__data/assets/pdf_file/0011/58835/Mind-the-Gap-Analytical-Report-D2 http://www.kpho.org.uk/__data/assets/pdf_file/0011/58835/Mind-the-Gap-Analytical-Report-D2 http://www.kpho.org.uk/__data/assets/pdf_file/0011/58835/Mind-the-Gap-Analytical-Report-D2 https://www.kingsfund.org.uk/sites/default/files/media/commission-appendix-uk-private-health-market https://www.kingsfund.org.uk/sites/default/files/media/commission-appendix-uk-private-health-market https://www.kingsfund.org.uk/sites/default/files/media/commission-appendix-uk-private-health-market http://refhub.elsevier.com/S0033-3506(19)30031-9/sref27 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref27 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref27 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref27 http://refhub.elsevier.com/S0033-3506(19)30031-9/sref27 https://digital.nhs.uk/catalogue/PUB30089 https://digital.nhs.uk/catalogue/PUB30089 https://www.england.nhs.uk/publication/mental-health-five-year-forward-view-dashboard/ https://www.england.nhs.uk/publication/mental-health-five-year-forward-view-dashboard/ https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthcaresystem/bulletins/ukhealthaccounts/2015 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthcaresystem/bulletins/ukhealthaccounts/2015 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthcaresystem/bulletins/ukhealthaccounts/2015 https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthcaresystem/bulletins/ukhealthaccounts/2015 https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 https://doi.org/10.1016/j.puhe.2019.02.007 Social gradients in health and social care costs: Analysis of linked electronic health records in Kent, UK Introduction Methods Results Discussion Author statements Acknowledgements Ethical approval Funding Competing interests References Appendix A. Supplementary data Planning-area-specific-prevention-and-intervention-programs-for_2019_Public- ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 9 Available online at w Public Health journal homepage: www.elsevier.com/puhe Original Research Planning area-specific prevention and intervention programs for HIV using spatial regression analysis S. Das a,*, J.J. Li b, A. Allston a, M. Kharfen c a Strategic Information Division, HIV/AIDS, Hepatitis, STD, and TB Administration (HAHSTA), District of Columbia Department of Health, 899 North Capitol St. NE / Fourth Floor, Washington, DC 20002, USA b George Washington University, Milken Institute School of Public Health, Department of Epidemiology and Biostatistics, 950 New Hampshire Ave NW, Washington, DC 20052, USA c HIV/AIDS, Hepatitis, STD and TB Administration (HAHSTA), District of Columbia Department of Health, Government of the District of Columbia 899 N. Capitol St., NE/ Fourth Floor, Washington, DC 20002, USA a r t i c l e i n f o Article history: Received 13 April 2018 Received in revised form 26 November 2018 Accepted 2 January 2019 Available online 25 February 2019 Keywords: HIV STIs Spatial variation Geographically weighted regression District of Columbia E-mail addresses: Suparna.das@dc.gov (S fen@dc.gov (M. Kharfen). https://doi.org/10.1016/j.puhe.2019.01.009 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objective: The study was conducted to inform area-based prevention intervention programs and plan resource allocation to reduce new infections in the District of Columbia (DC), United States of America. Study design: The analysis used spatial regression to evaluate the spatial heterogeneity of the new HIV rate and its association with sexually transmitted infection repeaters (STIR- EPs) and socio-economic as well as demographic characteristics. The HIV and STIREP data were obtained from the DC Department of Health surveillance data (2010e2016). Other covariates were obtained from the American Community Survey, 2016. Methods: Ordinary least squares (OLS) and geographically weighted regression (GWR) were used to compare global and local relationships. GWR-computed robust results were compared with other spatial regression methods such as spatial lag or spatial error methods. Results: For the OLS model, age, high school dropouts (NHSD), and the black population had an association with new HIV diagnoses (HIVDVi). The results from the GWR model demonstrate spatial variations of association of STIREPs; mean age of each block group; and percentage of female population, NHSD, unemployment, and poverty with HIVDVi. Akaike information criterion (AICc) value for the global model was 2770.99, and R2 was 0.54 (54%). The R2 and AICc of the GWR model was 0.81 (81%) and 2580.84, respectively, where the latter showed a 0.27 (27%) increase in R2 and a decreased AICc. Conclusion: These results will assist in planning HIV prevention and intervention strategies. These results will also be used for targeted testing, planning pre-exposure prophylaxis, and access to health care. The results will help plan resource allocation to community-based providers for prevention intervention programs and fund public health programs such as condom distribution, mobile vans, and youth-based sex education. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. * Corresponding author. Tel.: þ(202) 671 4943. . Das), jessicajli@gwmail. ic Health. Published by E gwu.edu (J.J. Li), Adam.Allston@dc.gov (A. Allston), Michael.Khar- lsevier Ltd. All rights reserved. mailto:Suparna.das@dc.gov mailto:jessicajli@gwmail.gwu.edu mailto:Adam.Allston@dc.gov mailto:Michael.Kharfen@dc.gov mailto:Michael.Kharfen@dc.gov http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.009&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.009 https://doi.org/10.1016/j.puhe.2019.01.009 https://doi.org/10.1016/j.puhe.2019.01.009 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 942 Introduction The Joint United Nations Program on HIV and AIDS (UNAIDS) launched the 90-90-90 plan in 2014, which aimed to end the AIDS epidemic by 2020 across the world. Reducing new transmission of HIV was one of the recommended goals for effective implementation of prevention and treatment pro- grams under the UNAIDS plan. Since 2010, the annual number of new HIV infections (all ages) in the world declined by 16% to 1.8 million,1 which is far less than the rate recommended by the United Nations General Assembly to reach the Fast Track target of 500,000 new infections by 2020.2 In the United States of America (US), the number of new HIV infections fell by 18% from 2008 to 2014; this was the first substantial drop in new infections in the country in two decades.3 The District of Columbia (DC) has experienced a constant decrease of new HIV infections over the last decade.4 However, after these large declines, there is an indication of a plateau effect in re- ductions, which has been seen globally. The DC has proposed a plan (90/90/90/50)4 concurrent with the UNAIDS plan, which seeks to reduce new HIV infections by 50% by 2020. An essential part of the DC's 90/90/90/50 plan was the imple- mentation of area-specific intervention programs based on risk clusters to reduce new HIV diagnoses (HIVDVi). To aid the area-specific prevention programs and optimize resource allocation, it is imperative to identify factors that may have an association with new HIV infections. Although there are studies that have evaluated various factors that may have an association with HIV, there has been little research that ex- amines the spatial dimension of these associations, and this analysis fills that gap. The analysis is recommended for the health department aiming to inform HIV prevention programs based on their local surveillance data. HIV and its relationship with the social determinants of health and sexually transmitted infections (STIs)5e9 is well- documented research. However, these relationships are most often analyzed using national and provincial data by applying global regression. By using global regression models, researchers implicitly rely on the assumption of a stationary relationship, which means parameter estimates describe an invariant relationship across space,10 thus masking any local area variation, which in turn leads to misinterpretation of the underlying spatial patterns.10e13 Therefore, public health researchers have begun to counter stationary assumptions by using spatial regression modeling such as geographically weighted regression (GWR) which al- lows for spatial variations in parameter estimates.14 These localized spatial models allow researchers to recognize the association between local patterns of disease prevalence and the covariates. The spatial models help in planning preven- tion efforts,13 planning resource allocation,15,16 inferring gaps in service provision,16 understanding biases of surveillance data,17 adapting services, and targeting interventions.12 It is critical for planners to have an understanding of the local variations of the intensity of the epidemic, and spatial methods make this feasible. In this study, we sought to answer the following questions: (1) do the relationships between HIVDVi and sexually trans- mitted infection repeaters (STIREPs) vary across places?; (2) do the relationships between HIVDVi and block group (BG)elevel socio-economic characteristics vary across places?; and (3) how do local relationships explain health outcomes when compared with global analysis? Methods and data Data sources The DC has an estimated 2016 population of 659,009 according to the American Community Survey (ACS) estimates. Administrative and statistical divisions of the DC include eight wards, 179 census tracts, and 450 Block Groups (BGs). The US Census Bureau BGs are statistical divisions of census tracts, which are defined to contain between 600 and 3000 people, and are used to present data and control block numbering. The DC cartographic boundary shapefile of BGs was obtained from the Master Address File/Topologically In- tegrated Geographic Encoding and Referencing database of the DC Office of the Chief Technology Officer for analysis (Fig. 1). HIV data HIV surveillance data used in this analysis are collected routinely by the HIV/AIDS, Hepatitis, sexually transmitted disease (STD), and tuberculosis (TB) Administration (HAHSTA) within the DC Department of Health (DOH). Newly diagnosed HIV cases reported from providers and laboratories were collected and managed in the Enhanced HIV/AIDS Reporting System. The total number of newly diagnosed HIV infections from 2010 to 2016 used for this analysis was 4237. The cases were geo-coded using Maptitude and then aggregated by BGs. The incidence rates for HIVDVi for each BG were calculated using Equation (1). HIVDVi ¼ ðTNIi=POPULATIONiÞ � 1000 (1) where HIVDVi is the rate of the diagnosed cases from 2010 to 2016 and TNIi is the total number of HIV diagnoses in the DC BG from 2010 to 2016. The population was obtained from the ACS 2016 estimate for each BG. STI data Repeaters are individuals who acquire more than one non- viral STI in a specified period,17 which, for this analysis, was 6 months or more. The STI data were regularly collected in the Sexually Transmitted Diseases Management Information System (STD*MIS) which were later transferred to the DC Public Health Information System (DCPHIS). Providers and laboratories report STI data which are routinely collected in the DCPHIS. The data for this study were obtained from the DCPHIS. There were 54,266 incidences identified using patient ID as well as the first name, last name, date of birth, and date of diagnosis between 2010 and 2016, of which 12,347 repeaters were identified using an algorithm of first name, last name, date of birth, and date of diagnosis. Repeaters are individuals who had more than any one of the reportable STIs diagnosed https://doi.org/10.1016/j.puhe.2019.01.009 https://doi.org/10.1016/j.puhe.2019.01.009 Fig. 1 e The boundary map of the District of Columbia. The red lines demarcate the block groups, and the black lines, the ward boundaries. (For interpretation of the references to color, the reader is referred to the Web version of this article.) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 9 43 between 2010 and 2016. The geographic coordinates associ- ated with each case of infection were assigned using Mapti- tude Geographic Information System software. In this study, three of the reportable STIsdchlamydia, gonorrhea, and syphilisdwere combined. The model was applied using each of the STIREPs separately in the ordinary least squares (OLS) model, but the variance inflation factor (VIF) indicated multi- collinearity. Thus, the STIREPs were combined into the STI repeat incidence rate variable to be included as an explanatory variable in the study. The incidence rates of STIREPs in each BG were calculated using Equation (1) as well. Other variates The other covariates used in the study were obtained from the ACS for each BG. The ACS is conducted each year to provide up-to-date information about the social and economic https://doi.org/10.1016/j.puhe.2019.01.009 https://doi.org/10.1016/j.puhe.2019.01.009 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 944 variables of communities. The census is conducted once every 10 years to provide an official count of the entire US popula- tion to the US Congress. The study aims to inform policies in the DC; thus, current estimates of population from the ACS were used instead of the US census data. There were six covariates used in the model for each BG: (1) mean age of people (AGE); (2) percentage of the female popu- lation (FEM); (3) percentage of people who are black (BLACK); (4) percentage of people who are high school dropouts (NHSD); (5) percentage of the unemployed (UNEM); and (6) percentage of people below the poverty line (POV). The variables were selected based on the literature18e20 and VIF as well as cor- relation association analysis of the independent variables. Statistical analyses Global regression The global linear regression (OLS) was computed using R software, version 3.3.4. yi ¼ a þ bxi þ εi (2) where a and b are the intercept and slope of the true regres- sion line, respectively. Each observation,yi, may be viewed as the sum of a component that predicts the value of yi on the basis of the value of xi (using true coefficients a and b) and some random error ðεiÞ: Spatial regression Spatial regression analyses (GWR) were performed using spgwr package of R software, version 3.3.4 , to account for the spatial non-stationarity of these relationships. Rather than calculating global parameter estimates based on one regres- sion analysis, GWR extends OLS by allowing regression co- efficients to vary spatially within a study area. Unlike OLS, GWR assumes that relationships between exposure and health outcomes may vary over space; it generates a set of local regression coefficients for each observation point in the BG.21 An established form of the GWR model is described in the following paragraphs. yi ¼ bi0ðui; viÞ þ Xp k¼1 bikðui; viÞxik þ εi i ¼ 1; …::n; (3) where yi is the dependent variable at point i, bi0ðui; viÞ is the intercept parameter at point i, bikðui; viÞ is the local regression coefficient for the kth independent variable at point i, and ðui; viÞ is the coordinate of the ith point in the study area. In our study, the x and y coordinates are given in miles through geographic coordinate system World Geodetic Sys- tem 1984 projection system. GWR requires assigning a specific x and y coordinate to each observation, i.14,21,22 GWR used equation (3) to generate separate regression equations for each observation, with a search window also known as spatial kernel,23 deciding which adjacent observa- tions were involved in the regulation of each regression. The spatial kernel also defined how the adjacent observations were weighed based on Gaussian distance decay.12,23 These functions give most weight to the observations that were nearest to the one at the center. The weights were based on the supposition that the nearby observations are more related to each other than the distant ones (Tobler's law). The spatial kernel area was based on a method of calibration to choose an ‘optimal’ bandwidth.23 This study applied adaptive kernel where the size would vary depending on the neighbors. Because the regression equation was calibrated indepen- dently for each observation, a separate parameter estimate, t- value, and goodness of fit were calculated for each observa- tion.23 Local values of t-statistics were calculated by dividing each local regression coefficient by the corresponding local standard error. These values were mapped for the graphical interpretation of the spatially varying association. Akaike information criterion (AICc) values reported for both global regression and GWR models were used to compare the performance of the models.23 The model with the lower AICc was considered to have a better fit. Spatial autocorrelation of standardized residuals was checked for both global regression and GWR models, using Moran's I in R. The global spatial clus- tering of HIVDVi was also checked using global Moran's I. Global Moran'sIisatoolthatmeasuresspatialautocorrelationbasedon both feature locations and feature values simultaneously.24 Results Descriptive analyses In total, 23% of the population were female compared with 73% male (the ramianing data were missing). Blacks or African Americans represented the majority of the new cases (approximately 69%) diagnosed in the DC. New diagnoses decreased from 20% in 2010 to 8% in 2016. The global Moran's I of HIVDVi is 0.0802 (P < 0.001) and demonstrates spatial auto- correlation, which makes GWR a relevant statistical analysis (Table 1). Global regression analyses The results of the OLS multiple regression analysis indicated that STIREPs had an association with HIVDVi (b ¼ 0.118, 95% confidence interval [CI]: 0.107, 0.129) (Table 2). NHSD also showed an association with HIVDVi (b ¼ 0.482, 95% CI: 0.091, 0.872), and UNEM (b ¼ 0.269, 95% CI: 0.099, 0.439). FEM (b ¼ �0.071, 95% CI: �0.143, �0.0002) had a negative associa- tion with HIVDVi. R 2 for the OLS model was 0.54, and the AICc, was 2770.99 (Table 2). Local regression analyses The summary results showed that R2 and AICc of the GWR model was 0.81 and 2580.84, respectively, thus a better fit than the OLS global regression model. The global Moran's I of the GWR residuals is 0.03 (P > 0.1), showing no spatial autocorre-
lation. The lack of spatial autocorrelation of the residuals in-

dicates that the residuals are independent and normally

distributed.

Fig. 2(aeg) showed the maps of coefficients of each

explanatory variables and the corresponding t-values. The

pseudo t-values for each variable were mapped to represent

the fitting level for each specific variable under GWR analysis.

https://doi.org/10.1016/j.puhe.2019.01.009

Table 1 e Characteristics of new HIV diagnoses (HIVDVi)
in the District of Columbia (2010e2016).

Variables Number (N) Percent (%)

Year of diagnosis

2010 883 21

2011 724 17

2012 709 17

2013 604 14

2014 481 11

2015 489 12

2016 347 8

Total 4237

Age group at diagnosis

0e14 years 17 1

15e24 years 882 21

25e49 years 2569 60

�50 years 769 18
Total 4237

Race/ethnicity

White 606 14

Black 2946 69

Hispanic 378 8

Other 110 2

Unknown 197 1

Total 4237

Sex at birth

Male 3103 73

Female 976 23

Missing 158 3

Total 4237

Global Moran’s I of
new HIV diagnoses

0.0802***

*P � 0.05, **P � 0.01, ***P � 0.001.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 9 45
The t-values indicated that the parameter estimations in these

areas were reliable.

AGE coefficients showed higher coefficient values in the

southern parts of the DC in the BGs within wards 6 and 8

extending into few BGs in ward 7. Few BGs in ward 4 showed
Table 2 e Global regression results of HIVDVi with explanatory

OLS glob

Estimate Standard error

Intercept 5.648 1.898

STIREPs 0.118 0.006

AGE (ages 15 to 24 years) �0.022 0.020
ELSEAGE (ages below

15 and above 24 years) (REF)

e e

FEM �0.071 0.036
NHSD 0.482 0.199

BLACK 0.000 0.009

UNEM 0.269 0.086

POV 0.046 0.049

R2 0.54

AICc 2770.99

Global Moran’s I standard residuals �0.011
AGE, mean age of people; AICc, Akaike information criterion; BLACK, bla

geographically weighted regression; HIVDVi, new HIV diagnoses; NHSD, h

the poverty line; STIREP, sexually transmitted infection repeater; UNEM,

*P � 0.05, **P � 0.01, ***P � 0.001.
higher coefficients as well. Lowest values were located in the

BGs within ward 7 (Fig. 2a) for AGE. FEM coefficients had

higher values in wards 1 and 7; few BGs in wards 4 and 5

showed higher values as well. Lowest FEM values were located

in the BGs of ward 8 (Fig. 2b). NHSD showed clustering of

higher coefficient values in the BGs within wards 7 and 8 and

in few BGs within ward 1 (Fig. 2c). BLACK had the lowest co-

efficients in few BGs of ward 7 (Fig. 2d). POV had higher co-

efficients in wards 4 and 8 and the lowest values in ward 7

(Fig. 2e). UNEM had higher coefficients clustered in wards 2

and 8 (Fig. 2f). STIREP (Fig. 2g) coefficients showed higher

values clustering in ward 4 and the lowest values clustered in

the southern BGs of the DC.

Supplement Fig 1 showed the maps of the locally weighed

R2 between the observed and fitted values, which indicated

how well the GWR model replicated the local HIVDVi around

the covariates selected for this analysis. It is evident that the

value of R2 was not homogeneously distributed all over the

DC, and the overall GWR model fitted best in the BGs of wards

1, 5, and 8 (R2 > 0.83).
The map of the intercept term represented the distributions

of HIVDVi when the covariates equaled zero (Supplement Fig. 2).

It was observed higher intercept values clustered in wards 7 and

8 and in few BGs within ward 5.
Discussion

This analysis provided further indications that the relation-

ships of HIVDVieSTIREP and the socio-economic variables are

spatially non-stationary in the DC. From the GWR model, it is

clear that the intensity and directions of the influence of

STIREPs and other covariates on HIVDVi are spatially hetero-

geneous. The result can thus be adapted to plan for area-

specific prevention and intervention strategies to control new

HIV infections.

The 90/90/90/50 plan suggests treatment as prevention,

which means preventing new infections by increasing the
variables.

al regression results GWR results

T-value Lower limit
95% CI

Upper limit
95% CI

b range

2.976** 1.9182 9.3770 �5.726, 28.168
20.895*** 0.107 0.129 0.0373, 0.4327

�1.119 �0.0617 0.0169 �0.348, 0.1912
e e e e

�1.970* �0.1426 �0.0002 �0.327, 0.1669
2.424* 0.0911 0.8720 �0.751, 1.881

�0.046 �0.0173 0.0165 �0.306, 0.0703
3.121** 0.0997 0.4389 �0.126, 1.466
0.930 �0.0511 0.1427 �0.462, 0.408

0.81

2580.84

�0.007
ck population; CI, confidence interval; FEM, female population; GWR,

igh school dropouts; OLS, ordinary least squares; POV, people below

unemployment.

https://doi.org/10.1016/j.puhe.2019.01.009

Fig. 2 e Spatial mapping of coefficients and the corresponding t-values. The dependent variable was new HIV diagnoses

(HIVDVi) in the District of Columbia (DC) from 2010 to 2016. (a) Age (15e24 years) coefficients and the t-values (AGE), (b) female

coefficients and the t-values (FEM), (c) high school dropouts coefficients and the t-values (NHSD), (d) black population

coefficients and the t-values (BLACK), (e) Poverty coefficients and the t-values (POV), (f) unemployment coefficients and the t-

values (UNEM), and (g) STI coefficients and the t-values (STIREP). AGE, mean age of people; BLACK, black population; FEM,

female population; NHSD, high school dropouts; POV, people below the poverty line; UNEM, unemployment; STIREP,

sexually transmitted infection repeater.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 946
number of people who reach viral suppression and are un-

likely to pass on the virus has the potential to reduce new

infections by 50% by 2020.25 In 2012, the Food and Drug

Administration approved a treatment regimen that is widely
used to prevent HIV infection among high-risk HIV negatives.

This treatment is called pre-exposure prophylaxis, or PrEP,

and is the first novel strategy introduced since the start of the

epidemic that is explicitly targeted at preventing sexual HIV

https://doi.org/10.1016/j.puhe.2019.01.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 9 47
transmission. Previous studies have shown that PrEP has been

successful in preventing HIV acquisition.26 The challenge is

access to PrEP; the results of this analysis will be used to plan

for providing PrEP through community-based organizations or

health department clinics to areas which are vulnerable.

The analysis found higher coefficients of STIREPs in the

northern parts of the DC, although the t-values were more

than 1.96 in all wards. It is well known that STIs increase the

risk of HIV,27 and controlling for repeat STIs has always

been considered a significant step to prevent new HIV

infection transmission.28 Repeat STIs double the risk of new

HIV infection and can be used to identify risk in HIV-

negative individuals.25 The high-risk areas which show

significant association of STIs and HIV will be targeted for

providing interventions, such as STI treatment, PrEP and

postexposure prophylaxis use, and elimination of needle

sharing through syringe access programs, to interrupt HIV

transmissions.

Women showed higher values clustered in central DC and

ward 7. There are disparities in the degree of variations which

calls for a deeper understanding of the burden of HIV di-

agnoses among women in the DC. Although PrEP for women

has been an essential step in the community, it is also

important to pay attention to the women’s ability to negotiate
safe sexual practices, particularly condom use.29 Apart from

condom use, attention needs to be paid to the association of

HIV risk with awareness, outlooks, peer encouragements, and

perception of self-worth among women.29e31

Unlike OLS, GWR showed a spatial association of AGE with

HIVDVi. According to Center for Disease Control and Preven-

tion, United States (CDC) (2015), 22% of all HIVDVi in the US

were between ages 15 and 24 years.32 Of any age group, youths

with HIV are least likely to be linked to care and have a sup-

pressed viral load (i.e., having a low level of the virus in the

body, which helps the person to stay healthy and reduce the

risk of transmitting HIV to others). Addressing HIV in youths

required them to be provided with information and tools

needed to reduce their risk, make healthy decisions, and get

treatment and care if needed.32

The global regression model also showed a positive associ-

ation of HIVDVi with lack of high school education (NHSD). The

spatial regression analysis found spatial heterogeneity in the

association of NHSD with HIVDVi in southern DC and few BGs

spread across wards 1 and 5. High schooling years have shown

to decrease the risk of HIV.33 On a worldwide level, it has been

observed that lower education and lack of sexual health edu-

cation in school is related to a higher risk of HIV epidemics.34,35

The DOH formed a collaboration across healthcare providers,

researchers, district government agencies, community orga-

nizations, and young people to develop the 2016e2020 Youth

Sexual Health Plan to address sexual health among younger

people. Programs need to be extended to include initiatives that

support students absent from in-school sessions. The DOH will

continue to work in concert with public education efforts and

school-based sexual health education.

Unemployment and poverty show spatial variation in as-

sociation with HIVDVi, which points out areas where poverty

and unemployment may need to be addressed. Employment

benefits in the past have been shown to improve HIV health

outcomes,36e38 and the DC provides employment assistance,
financial services, utility assistance, and rental assistance to

those in need and has special programs for HIV patients.

There are also studies that have shown that employment

benefits HIV health outcomes. Employed people were 39%

more likely to have achieved optimal adherence to antiretro-

viral medications (reaching better than 95% adherence),

which in turn helps curtail new infections.39 However, it also

needs to be acknowledged that employment generation does

not always lead to poverty alleviation, particularly in areas

where disease stems from disparity.37 Many researchers have

often described HIV as a ‘disease of poverty’.38 Thus, health

departments should consider poverty and its impact on HIV-

related health outcomes.

Geographical heterogeneity was detected by the GWR

method in the relationship of new HIV infections with STIs

and other economic and demographic variables. The analysis

used GWR because conventional regression analysis, OLS,

cannot discriminate the spatial variation in relationships if

geographical non-stationarity exists. The results of adjusted

R2 and AICc indicated GWR was a better model to explain the

dataset. GWR analysis and its application to HIV and STI

research have been gradually gaining significance, and this

analysis is a significant step towards that.

Conclusion

As a methodology, GWR has some limitations. Non-linear terms

cannot be added to the GWR models, and the model inferences

cannot be conducted in this model.12 However, despite the

limitations, the article is an important contribution toward

understanding the underlyingetiologythat may have animpact

on new HIV infections. The article also opens avenues such as

access to PrEP, HIV risk among women, and space-based net-

works for research that await to be explored. Future research

using Bayesian additive regression models, which are based on

Markov chain Monte Carlo algorithms, for parameter estima-

tions and inferences to overcome the mentioned problems is

underway.
Author statements

Acknowledgments

The authors would like to thank the District of Columbia

Department of Health for constant support.

Ethical approval

The DC surveillance receives electronic reports from providers

and laboratories for HIV, gonorrhea, chlamydia, and syphilis.

The studies published from the District of Columbia Depart-

ment of Health do not require any IRB clearance unless they

have to identify information. Our article does not have any

identifying information; the cases/events are aggregated by

census tracts, thus geo-masked.

Funding

The authors did not receive any funding for research.

https://doi.org/10.1016/j.puhe.2019.01.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 948
Competing interests

The authors do not have any known competing interests. The

authors do not have anything to declare.
r e f e r e n c e s

1. Unaids. Ending Aids Progress Towards the 90-90-90 Targets.
Global Aids Update; 2017. p. 198. Available from: http://www.
unaids.org/sites/default/files/media_asset/Global_AIDS_
update_2017_en .

2. Unaids. 90-90-90 An ambitious treatment target to help end the
AIDS epidemic. 2014. p. 40. http://WwwUnaidsOrg/Sites/
Default/Files/Media_Asset/90-90-90_En_0Pdf.

3. CDC. HIV incidence: estimated annual infections in the U.S.,
2008e2014. Atlanta, GA. 2017.

4. HAHSTA DC Department of Health. Annual Epidemiology &
Surveillance Report. 2017.

5. Marmot M. Social determinants of health inequalities. Lancet
2005;365(9464):1099e104. Available from: http://www.
sciencedirect.com/science/article/pii/S0140673605711466.

6. Takarinda KC, Madyira LK, Mhangara M, Makaza V,
Maphosa M, Rusakaniko S, et al. Factors associated with HIV
testing in the 2010-11 Zimbabwe Demographic and Health Survey.
DHS Working Papers No. 110 (Zimbabwe Working Papers No. 11).
Rockville, Maryland, USA: ICF International; 2014. Available
from: http://dhsprogram.com/pubs/pdf/WP110/WP110 .

7. Ward H, R€onn M. The contribution of STIs to the sexual
transmission of HIV. Curr Opin HIV AIDS 2010;5(4):305e10.

8. Weber JN, McCreaner A, Berrie E, Wadsworth J, Jeffries DJ,
Pinching AJ, et al. Factors affecting seropositivity to human T
cell lymphotropic virus type III (HTLV-III) or
lymphadenopathy associated virus (LAV) and progression of
disease in sexual partners of patients with AIDS. Genitourin
Med 1986;62(March):177e80. Available from: http://www.ncbi.
nlm.nih.gov/entrez/query.fcgi?
cmd¼Retrieve&db¼PubMed&dopt¼Citation&list_
uids¼3015772.

9. Piot P, Taelman H, Bila Minlangu K, Mbendi N, Ndangi K,
Kalambayi K, et al. Acquired Immunodeficiency Syndrome in
a Heterosexual Population in Zaire. Lancet 1984;324:65e9.

10. Clary C, Lewis DJ, Flint E, Smith NR, Kestens Y, Cummins S.
The local food environment and fruit and vegetable intake: A
geographically weighted regression approach in the ORiEL
study. Am J Epidemiol 2016;184(11):837e46.

11. John Tukey. Statistical mapping: What should not be mapped?
Collect Work. Belmont, CA: Wadsworth; 1988.

12. Fotheringham AS, Brunsdon C, Charlton M. Geographically
Weighted Regression: The Analysis of Spatially Varying
relationships. John Wiley & Sons; 2003.

13. Weir SS, Boerma T. From People to Places : Overview of the PLACE
Method and Lessons Learned Background : Evaluation, vol. 4; 2002.
p. 1e6.

14. Fotheringham A. Stewart et al. Geographically Weighted
Regression : The Analysis of Spatially Varying Relationships.
John Wiley & Sons, Incorporated,;

15. Auerbach JD, Stall R, Herrick A, Guadamuz T E, Friedman M S,
Fenton KA, et al. HIV Prevention. HIV Prevention. Elsevier; 2009
[cited 2016 Mar 23]. Available from: http://www.sciencedirect.
com/science/article/pii/B9780123742353000170.

16. Wilson DP, Blower SM. Designing equitable antiretroviral
allocation strategies in resource-constrained countries. PLoS
Med 2005;2(2):0132e41.

17. Montana LS, Mishra V, Hong R, Montana LS. Comparison of
HIV prevalence estimates from antenatal care surveillance
and population-based surveys in Sub-Saharan Africa . DHS
Working Papers No 47. Calverton, Maryland, USA : Macro
International 2008;84(Suppl 1):i78e84. Available from: http://
dhsprogram.com/pubs/pdf/WP47/WP47 .

18. Delpierre C, Cuzin L, Lauwers-Cances V, Datta GD, Berkman L,
Lang T. Unemployment as a risk factor for AIDS and death for
HIV-infected patients in the era of highly active antiretroviral
therapy. Sex Transm Infect 2008;84:183e6.

19. Krueger LE, Wood RW, Diehr PH, Maxwell CL. Poverty and HIV
seropositivity: the poor are more likely to be infected. AIDS
1990;4(8). Available from: http://journals.lww.com/
aidsonline/Fulltext/1990/08000/Poverty_and_HIV_
seropositivity__the_poor_are_more.15.aspx.

20. Gregson S, Waddell H, Chandiwana S. School education and
HIV control in sub-Saharan Africa: from discord to harmony?.
John Wiley & Sons, Ltd. J Int Dev 2001;13(4):467e85. Available
from: https://doi.org/10.1002/jid.798.

21. Gebreab SY, Diez Roux AV. Exploring racial disparities in CHD
mortality between blacks and whites across the United States: A
geographically weighted regression approach. Health and place18
5. Elsevier; 2012. p. 1006e14. Available from: https://doi.org/
10.1016/j.healthplace.2012.06.006.

22. Brunsdon C, Fotheringham S, Charlton M. Geographically
weighted regression- modelling spatial non-stationarity. 1998.
p. 431e43.

23. Tsiko R. Geographically Weighted Regression of
Determinants Affecting Women’s Access to Land in Africa.
Geosciences 2016;6:16. Available from: http://www.mdpi.com/
2076-3263/6/1/16.

24. Anselin L. Local Indicators of Spatial Association – Lisa. Geogr
Anal 1995;27(2):93e115.

25. HAHSTA DC DEPT OF HEALTH, DC Appleseed WA partnership.
90-90-90-50 Plan. 2016.

26. McCormack S, Dunn DT, Desai M, Dolling DI, Gafos M,
Gilson R, et al. Pre-exposure prophylaxis to prevent the
acquisition of HIV-1 infection (PROUD): Effectiveness results
from the pilot phase of a pragmatic open-label randomised
trial. McCormack et al. Open Access article distributed
under the terms of CC BY Lancet 2016;387(10013):53e60.
Available from:, https://doi.org/10.1016/S0140-6736(15)
00056-2.

27. Malek R, Mitchell H, Furegato M, Simms I, Mohammed H,
Nardone A, et al. Contribution of transmission in HIV-positive
men who have sex with men to evolving epidemics of
sexually transmitted infections in England: an analysis using
multiple data sources, 2009e2013. Euro Surveill 2015;20(15).
Available from: http://www.eurosurveillance.org/content/10.
2807/1560-7917.ES2015.20.15.21093.

28. Lattimore S, Thornton A, Delpech V, Elford J. Changing
patterns of sexual risk behavior among London gay men:
1998-2008. Sex Transm Dis 2011;38(3):221e9.

29. Pulerwitz J, Amaro H, Jong W De, Gortmaker SL, Rudd R.
Relationship power, condom use and HIV risk among women
in the USA. AIDS Care 2002;14(May):789e800. Available from:
http://www.tandfonline.com/doi/abs/10.1080/
0954012021000031868.

30. Gielen AC, Faden RR, O’Campo P, Kass N, Anderson J. Women’s
protective sexual behaviors: a test of the health belief model. AIDS
education and prevention, vol. 1. United States: official
publication of the International Society for AIDS Education;
1994 Feb;6. p. 1e11.

31. Cynthia A, Vanoss B. Gender , culture , and power : Barriers to
HIV-prevention strategies for women. 1996.

32. Centre for Disease Control. HIV Among Youth. Factsheet
[Internet]. 2015 (December 2015):2. Available from: http://
www.cdc.gov/hiv/pdf/risk_youth_fact_sheet_final .

33. Gant Z, Gant L, Song R, Willis L, Johnson AS. A census tract-
level examination of social determinants of health among

http://www.unaids.org/sites/default/files/media_asset/Global_AIDS_update_2017_en

http://WwwUnaidsOrg/Sites/Default/Files/Media_Asset/90-90-90_En_0Pdf

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref3

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref4

http://www.sciencedirect.com/science/article/pii/S0140673605711466

http://dhsprogram.com/pubs/pdf/WP110/WP110

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref7

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=3015772

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref9

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref10

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref11

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref12

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref13

http://www.sciencedirect.com/science/article/pii/B9780123742353000170

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref16

http://dhsprogram.com/pubs/pdf/WP47/WP47

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref18

http://journals.lww.com/aidsonline/Fulltext/1990/08000/Poverty_and_HIV_seropositivity__the_poor_are_more.15.aspx

https://doi.org/10.1002/jid.798

https://doi.org/10.1016/j.healthplace.2012.06.006

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref22

http://www.mdpi.com/2076-3263/6/1/16

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref24

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref25

https://doi.org/10.1016/S0140-6736(15)00056-2

http://www.eurosurveillance.org/content/10.2807/1560-7917.ES2015.20.15.21093

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref28

http://www.tandfonline.com/doi/abs/10.1080/0954012021000031868

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref30

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref31

http://www.cdc.gov/hiv/pdf/risk_youth_fact_sheet_final

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref33

https://doi.org/10.1016/j.puhe.2019.01.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 4 1 e4 9 49
black/African American men with diagnosed HIV infection,
2005-2009-17 US areas. PLoS One 2014;9(9):2005e9.

34. Behrman JA. The effect of increased primary schooling on adult
women’s HIV status in Malawi and Uganda: Universal Primary
Education as a natural experiment. Social Science and Medicine, vol.
127. Elsevier Ltd; 2015. p. 108e15. Available from: https://doi.
org/10.1016/j.socscimed.2014.06.034.

35. De Neve JW, Fink G, Suburamanian S, Moyo S, Bor J. Length of
secondary schooling and risk of HIV infection in Botswana:
evidence from a natural experiment. Lancet Global Health
2015;3(8):470e7.

36. Nachega JB, Uthman OA, Peltzer K, Richardson LA, Mills EJ,
Amekudzi K, et al. Association between antiretroviral therapy
adherence and employment status: systematic review and
meta-analysis. Bull World Health Organ
2015;93(October):29e41. Available from: http://www.
pubmedcentral.nih.gov/articlerender.fcgi?
artid¼4271680&tool¼pmcentrez&rendertype¼abstract.
37. Gutierrez C, Carlo O, Pierella P, Pieter S. Does Employment
Generation Really Matter for Poverty Reduction? Policy Research
Working Paper 4432 [Internet] (December). 2007. p. 4e33.
Available from:, http://elibrary.worldbank.org/doi/pdf/10.
1596/1813-9450-4432.

38. Gillies P, Tolley K, Wolstenholme J. Is AIDS a disease of
poverty? AIDS Care 1996;8(3).

39. Nachega JB, Uthman O a, Mills EJ, Peltzer K. The Impact of
Employment on HIV Treatment Adherence. 2013.
Appendix A. Supplementary data

Supplementary data to this article can be found online at

https://doi.org/10.1016/j.puhe.2019.01.009.

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref33

https://doi.org/10.1016/j.socscimed.2014.06.034

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref35

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4271680&tool=pmcentrez&rendertype=abstract

http://elibrary.worldbank.org/doi/pdf/10.1596/1813-9450-4432

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref38

http://refhub.elsevier.com/S0033-3506(19)30009-5/sref39

https://doi.org/10.1016/j.puhe.2019.01.009

Planning area-specific prevention and intervention programs for HIV using spatial regression analysis
Introduction
Methods and data
Data sources
HIV data
STI data
Other variates
Statistical analyses
Global regression
Spatial regression

Results
Descriptive analyses
Global regression analyses
Local regression analyses

Discussion
Conclusion

Author statements
Acknowledgments

Ethical approval
Funding
Competing interests
References
Appendix A. Supplementary data

The-economics-of-prevention_2019_Public-Health

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 9 e1 5 0
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Guest Editorial
The economics of prevention
Investing in non-communicable disease (NCD) prevention and

management has been described as a global development

imperative in order to reduce premature NCD mortality by

one-third by 2030.1 This case for investing in prevention and

wider public health is often made but in reality resources

often do not follow the rhetoric.2

For example, in the United Kingdom, the National Health

Service (NHS) Five-Year Forward View promised a ‘radical

upgrade to prevention’, something we have yet to see take

place in any systematic sense.3 The latest NHS Long-Term

Plan similarly promises a stronger focus on prevention, and

the health and social care system will have to grasp the op-

portunity this time round.4 An environment of significant cuts

in public health budgets, and local government budgets more

generally, does not help to create an environment in which a

focus on prevention can flourish.

Making the case for investing in prevention is at face value a

straightforward onedthe popular perception is that ‘preven-

tion is better than cure’, as reiterated in the UK Secretary of

State’s recent pronouncement for England set out in the Pre-
vention Vision.5 The economic case for prevention needs to be

made not just in relation to the health and social care system

but also the wider economy, appealing not just to public health

professionals but also to key decision makers across the board.

In this special section, Devaux et al.6 remind us of this ‘eco-

nomic dividend’ that can even extend to modest savings in

heathcare expenditure if we get prevention and early inter-

vention right. But more broadly, effectively tackling the key risk

factors is likely to convey significant benefits to the wider

economy in terms of enhanced productivity and the informal

care system. The Prevention Vision has at its heart the aspi-

ration to gain extra life years in good health, as well as a desire

to reduce the gap between rich and poor, more of which later.

A myth that has hopefully been exploded in recent years is

that there is limited or poor evidence on the economic value of

preventive interventions. In the UK, the National Institute for

Health and Care Excellence (NICE) has been evaluating public

health interventions for nearly 15 years, and Owen and Fischer7

remind us that a wide range of preventive interventions are

cost-effective, with a small but significant proportion demon-

strating the potential for cost savings. Craig and Robinson8

reinforce the need for us to harness the growing evidence

base on the economics of prevention, given the scope both to
improve population health and reduce health inequalities.

They point out, however, that political will is critical, no doubt

heavily influenced by the Scottish experience where the

importance of the social determinants of health is to the fore in

the strategy to tackle health inequalities.9

Alongside the wider determinants and arguments for

tackling health inequalities, Jayatunga et al.10 helpfully

remind us that there is a social gradient in health and social

care costs, so that there is in effect an ‘excess cost’ associated

with health inequalities. Tackling these may therefore have

the potential to reduce system costs, even if this is not the

primary objective of concern here.

Despite the growing evidence base, Penny Reeves high-

lights the often overlooked issue of how effective and cost-

effective interventions are implemented in practicedarguing

that we should also be focusing on the economic evaluation of

public health implementation interventions. The process of

implementation itself consumes considerable resources and

therefore has a significant opportunity cost that needs to be

recognized (as well as the similarly often overlooked equity

impact of different interventions).

This theme is echoed in Kelly’s11 piece on cognitive biases
in public health thinking, which he argues distort the evi-

dence base at our disposal. Critically, he points out that ‘pol-

icy, practice and interventions frequently repeat things which

have previously been shown to have had little or no effect’: in

short, there is a lack of focus on the mechanics of prevention

and how to change behaviour. Encouragingly, Kelly considers

that ‘health economics has potentially an important role to

play in developing the ideas that will overcome the problems

attaching to the cognitive biases.’ Importantly, however, he

reminds us of the traditional focus of economics on individual

utility maximization, and even behavioural economics largely

operates within the same paradigm. The ‘call to arms’ is for us

to enhance our knowledge of population dynamics and

strengthen the evidence base on how to implement policies

that benefit all segments of the population.

Fiscal policies at least have the potential to bring about

significant change at the population level, and Ludbrook12

calls for the smarter design of those policies. Convention-

ally, the rationale for such policies is to correct for production

or consumption externalities, with the emphasis on taxes/

raising prices to ‘correct’ for perceived market failures. She

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.03.009&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.03.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 9 e1 5 0150
argues for a greater emphasis on subsidies and taxes in order

to promote healthier choices. This was an important theme in

Public Health England’s recently published evidence review on
the scope for fiscal and pricing policies to improve population

health and reduce health inequalities.13

So where does all this leave us? Firstly, alongside the

health arguments, there is a financial imperative to shift the

focus to prevention; otherwise, we simply have a health and

social care system that will continue to run faster and faster to

stand still. Secondly, we have a profoundly strong and

growing evidence base for prevention but a track record of

failing to convince key decision makers of the wider economic

value of prevention. There is a tremendous opportunity to

change this but it will need courage and, dare we say, the odd

taking of risks in order to really make the seismic shift needed

to orientate the system more toward prevention being a cen-

tral theme. The challenge of tackling climate change (another

major public health priority) shows starkly the tendency of

society to focus on the here and now, imposing more and

more cost, and potentially severe adverse health impacts, on

current and future generations.

Thirdly, we are reminded here of the requirement to focus

not just on what interventions need to be implemented but on

how effective and lasting change can be brought about. If we

do not do so, then past failures will be blindly repeated.

Economists have a continuing role in questioning what ob-

jectives we are trying to achieve here and what trade-offs

exist. This includes being clear not just about the benefits

(health or wider economic) that we are trying to bring about

but the distribution of those benefits in terms of outcomes to

particular population groups. Such issues are inevitably po-

litical at their heart, but there is much that can be learned

through an economic lens, as evidenced by the range and

depth of the contributions to this special section.
r e f e r e n c e s

1. Nugent R, Bertram MY, Jan S, Niessen LW, Sassi F,
Jamison DT, et al. Investing in non-communicable disease
prevention and management to advance the Sustainable
Development Goals. Lancet 2018 19;391(10134):2029e35.

2. WHO Europe. The case for investing in public health: a public
health summary report for EPHO 8. World Health Organisation;
2014. Available online at: http://www.euro.who.int/__data/
assets/pdf_file/0009/278073/Case-Investing-Public-Health.
pdf. [Accessed 8 March 2019].

3. National Health Service. Five-year forward view. October 2014.
Available at: https://www.england.nhs.uk/wp-content/
uploads/2014/10/5yfv-web . [Accessed 8 March 2019].
4. National Health Service. The NHS long-term plan. January 2019.
Available at: https://www.longtermplan.nhs.uk/publication/
nhs-long-term-plan/. [Accessed 8 March 2019].

5. Department of Health and Social Care. Prevention is better than
cure – our vision to help you live well for longer. Available at:
https://assets.publishing.service.gov.uk/government/uploads/
system/uploads/attachment_data/file/753688/Prevention_is_
better_than_cure_5-11 . [Accessed 8 March 2019].

6. Devaux M, Lerouge A, Ventelou B, Goryakin Y, Feigl A, Vuik S,
et al. Assessing the potential outcomes of achieving the WHO
global NCDs targets for risk factors by 2025: is there also an
economic dividend? Publ Health 2019;169:173e9.

7. Owen L, Fischer A. The cost-effectiveness of public health
interventions examined by NICE from 2005 to 2018. Publ
Health 2019;169:151e62.

8. Craig N, Robinson M. Towards a preventative approach to
improving health and reducing health inequalities: a view
from Scotland. Publ Health 2019;169:195e200.

9. NHS Health Scotland. Health inequalities policy review for the
scottish ministerial task force on health inequalities. June 2013.
Available at: http://www.healthscotland.scot/media/1538/
health-inequalities-policy-review-march-2014-english .
[Accessed 12 March 2019].

10. Jayatunga W, Asaria M, Belloni A, George A, Bourne T,
Saddique Z. Social gradients in health and social care costs:
analysis of linked electronic health records in Kent, UK. Publ
Health 2019;169:188e94.

11. Kelly MP. Cognitive biases in public health and how
economics and sociology can help overcome them. Publ
Health 2019;169:163e72.

12. Ludbrook A. Fiscal measures to promote healthier choices: an
economic perspective on price based interventions. Publ
Health 2019;169:180e7.

13. Pimpin L, Sassi F, Corbould E, Freibel R, Webber L. Fiscal and
pricing policies to improve public health: a review of the evidence.
Public Health England; 2018. Available at: https://www.gov.
uk/government/publications/fiscal-and-pricing-policies-
evidence-report-and-framework. [Accessed 8 March 2019].

B. Ferguson*

A. Belloni

Public Health England, UK

*Corresponding author.

E-mail address: brian.ferguson@phe.gov.uk (B. Ferguson)

https://doi.org/10.1016/j.puhe.2019.03.009
0033-3506/© 2019 The Royal Society for Public Health. Published

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref1

http://www.euro.who.int/__data/assets/pdf_file/0009/278073/Case-Investing-Public-Health

https://www.england.nhs.uk/wp-content/uploads/2014/10/5yfv-web

https://www.longtermplan.nhs.uk/publication/nhs-long-term-plan/

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/753688/Prevention_is_better_than_cure_5-11

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref6

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref7

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref8

http://www.healthscotland.scot/media/1538/health-inequalities-policy-review-march-2014-english

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref10

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref11

http://refhub.elsevier.com/S0033-3506(19)30083-6/sref12

https://www.gov.uk/government/publications/fiscal-and-pricing-policies-evidence-report-and-framework

mailto:brian.ferguson@phe.gov.uk

https://doi.org/10.1016/j.puhe.2019.03.009

The economics of prevention
References

Corrigendum-to–Effect-of-forest-bathing-on-physiological-and-psy_2019_Publi

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 2 0 1
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Corrigendum
Corrigendum to “Effect of forest bathing on
physiological and psychological responses in
young Japanese male subjects” [Public Health 125
(2) (February 2011), 93e100]
J. Lee a,*, B.-J. Park b, Y. Tsunetsugu c, T. Ohira c, T. Kagawa c,
Y. Miyazaki a

a Centre for Environment, Health and Field Sciences, Chiba University, 6-2-1 Kashiwanoha, Kashiwa City,

Chiba Prefecture 277-0882, Japan
b Department of Environment & Forest Resources, Chungnam National University, Daejeon, South Korea
c Forestry and Forest Products Research Institute, Tsukuba, Japan
The authors would like to highlight to the readers there was an error in POMS data used as a psychological index in our published

paper. The error was found in page 98 of issue 123(vol. 2) and is as follows:

– Error: Fig. 5(page 98) and its explanation (Line 23 left column e Line 10 right column, page 96). There are no changes in the main

result and conclusion.

Correct Figure 5 can be found below.

The authors would like to apologise for any inconvenience caused.
Fig. 5. Scores of subjective evaluation for ‘comfortableeuncomfortable’ (top), ‘soothingeawakening’ (upper middle) and

‘refreshed’ feeling (lower middle) at three measurement periods, and subscale T-scores for the Profile of Mood States (POMS;

bottom) at before stimuli and after stimuli periods in the forest and urban environments. Be, before stimuli; Af, after stimuli; T-A,

tension-anxiety; D, depression-dejection; A-H, anger-hostility; V, vigour; F, fatigue; C, confusion; TMD, total mood disturbance,

calculated by adding the scores of the six subscales where V is negatively weighted. n ¼ 12, mean ± standard error; *P < 0.05; **P < 0.01; P values were obtained by Wilcoxon signed rank test. DOI of original article: https://doi.org/10.1016/j.puhe.2010.09.005. * Corresponding author. E-mail address: lohawi@gmail.com (J. Lee). https://doi.org/10.1016/j.puhe.2019.03.002 0033-3506/© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. https://doi.org/10.1016/j.puhe.2010.09.005 mailto:lohawi@gmail.com http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.03.002&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.03.002 https://doi.org/10.1016/j.puhe.2019.03.002 Corrigendum to “Effect of forest bathing on physiological and psychological responses in young Japanese male subjects” [Pub ... Planning-of-births-and-maternal--child-health--and-nutritional_2019_Public-H ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 Available online at w Public Health journal homepage: www.elsevier.com/puhe Original Research Planning of births and maternal, child health, and nutritional outcomes: recent evidence from India M.J. Rana a,*, A. Gautam b,1, S. Goli a,2, Uttamacharya b,3, T. Reja b,4, P. Nanda c,5, N. Datta b,6, R. Verma b,7 a Centre for the Study of Regional Development, School of Social Sciences, Jawaharlal Nehru University (JNU), New Delhi, India b International Center for Research on Women (ICRW), New Delhi, India c Bill & Melinda Gates Foundation (BMGF), India Country Office, New Delhi, India a r t i c l e i n f o Article history: Received 8 August 2018 Received in revised form 22 November 2018 Accepted 30 November 2018 Available online 14 February 2019 Keywords: Planning of births Maternal and child health Nutrition Family planning India * Corresponding author. Tel.: þ91 875045740 E-mail addresses: jranajnu@gmail.com (M icrw.org ( Uttamacharya), treja@icrw.org (T. icrw.org (R. Verma). 1 Tel.:þ91 9958591007 (mobile). 2 Tel.: þ91 7042181232 (mobile). 3 Tel.: þ91 8750844827 (mobile). 4 Tel.: þ91 9892404598 (mobile). 5 Tel.: þ91 9871298452 (mobile). 6 Tel.: þ91 8527885518 (mobile). 7 Tel.:þ91 9810595578 (mobile). https://doi.org/10.1016/j.puhe.2018.11.019 0033-3506/© 2018 The Authors. Published by under the CC BY license (http://creativecom a b s t r a c t Objectives: In an effort to provide recommendation for maximizing synergy between maternal, infant, and young children's nutrition and family planning in India, this study makes a comprehensive assessment of the effects of the planning of births in terms of timing, spacing and limiting childbearing on maternal and child health outcomes. Study design: This study used the latest National Family Health Survey data of India that is globally known as the Demographic and Health Survey. A robust two-stage systematic random sampling was used for selecting representative samples for measuring de- mographic and health indicators. Methods: Maternal and child health outcomes are measured by body mass index (grouped as normal, underweight, and overweight) and anemia for mothers, and stunting, under- weight, anemia, and under-five mortality for the children. Logistic regression and Cox proportional hazard models were applied. Results: Women with a higher number of births and among those with first-order births with fewer than 2 years between marriage and first birth, the risk of being underweight and having anemia was significantly higher compared with their counterparts. In addition, the probability of being underweight and risk of stunting, anemia, and mortality was higher among the children from women with a higher number of births and with fewer than 3 years of spacing between births than that of their counterparts. 5 (mobile). .J. Rana), agautam@icrw.org (A. Gautam), sirispeaks2u@gmail.com (S. Goli), uttamacharya@ Reja), Priya.Nanda@gatesfoundation.org (P. Nanda), ndatta@icrw.org (N. Datta), rverma@ Elsevier Ltd on behalf of The Royal Society for Public Health. This is an open access article mons.org/licenses/by/4.0/). mailto:jranajnu@gmail.com mailto:agautam@icrw.org mailto:sirispeaks2u@gmail.com mailto:uttamacharya@icrw.org mailto:uttamacharya@icrw.org mailto:treja@icrw.org mailto:Priya.Nanda@gatesfoundation.org mailto:ndatta@icrw.org mailto:rverma@icrw.org mailto:rverma@icrw.org http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2018.11.019&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 http://creativecommons.org/licenses/by/4.0/ p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 15 Conclusions: The findings from this study support the importance of birth planning in improving maternal, child health, and nutritional outcomes. The proper planning of births could help to achieve the Sustainable Development Goal-3 of good health and well-being for all by 2030 in India, where a significant proportion of women still participate in early marriages, early childbearing, and a large number of births with close spacing. © 2018 The Authors. Published by Elsevier Ltd on behalf of The Royal Society for Public Health. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Introduction Maternal, child health, and nutritional outcomes have been key public health issues in developing countries including India. The United Nations has targeted improving maternal, child health, and nutritional outcomes under its Sustainable Development Goal-3 (SDG-3), focusing on the lagging regions by 2030.1 The integration of family planning with maternal, newborn, child health, and nutrition services at the program- and policy-level is considered to have natural synergy that benefits women as well as their children. Family planning and maternal, newborn, and child health and nutritional out- comes service integration has shown promising improvement in large variety of health care, processes, and outcomes.2 In the global context, the existing evidence indicates that family planning can have a significant influence on achieving key maternal, newborn, and child health and nutritional out- comes.3,4 However, a substantial evidence gap continues to persist in the developing countries. Family planning affects maternal and child health and nutritional outcomes in myriad direct and indirect ways. The integration of family planning with maternal and child health and nutrition services is not a process that occurs in a single episode; instead, it is a continuous process of timing, spacing, and limiting of births. Family planning helps couples plan childbearing regarding timing, spacing, and limiting preg- nancy and childbirths.3,4 Earlier studies have analyzed the effects of each of these components of the planning of births, such as timing (age at first birth), spacing (birth interval), and limiting (number of children) on maternal, newborn, child health, and nutritional outcomes independently.5e10 The timing of the first birth is assessed based on the timing of marriage, timing of the first birth after marriage, and the gap between the date of the marriage and the first birth. The timing of the first birth is related to maternal, newborn, child health, and nutritional outcomes.11 The timing of the first birth at a maternal adolescent age adversely affects the women's health because they are in a critical period of physical growth that is hindered by pregnancy and childbearing.12e17 However, even if marriage takes place at an early age, postponement of childbearing until women become physically and psycholog- ically capable will result in better pregnancy and delivery outcomes.12e17 The use of family planning helps in the post- ponement of childbearing to optimal ages.17 On the other hand, a shorter birth spacing and a higher number of births due to repeated childbearing as a consequence of poor acceptance of family planning, higher unmet need for family planning, and more unintended births in turn lead to poor maternal, newborn, and child health and nutritional outcomes.7e9,18,19 Insufficient birth spacing and repeated childbearing cause the recurrent loss of macronu- trients and micronutrients from the women's body during the pregnancy, delivery, and breastfeeding.15,17,20 Such un- planned childbearing elevates the risk of intrauterine growth restriction (IUGR), low birth weight (LBW), premature birth, and small birth size. The poor pregnancy and delivery out- comes for the baby make them vulnerable to reduced physical growth and add to the risk of mortality during their childhood.2,15,20 On the other hand, studies that used direct measures and indirect proxies of family planning such as unintended births, in particular when based on cross-sectional data from de- mographic and health surveys (DHSs),21,22 lack a compre- hensive outline of empirical evidence on the pathways of the influence of family planning on maternal, newborn, child health and on nutritional outcomes. Moreover, unintended births are not only just a result of not having access to family planning or the failure of it but also due to other social or cultural reasons.23 Recent evidence suggests that the rate of unintended births has been falling and the fertility is declining in India.24e27 Therefore, an unintended birth is not a good proxy to predict family planning. Furthermore, limitations related to family planning questions in DHS data do not allow the direct linking of family planning to maternal, child health, and nutritional outcomes.4 Although, progress in age at first birth, birth order, and birth interval helps to achieve favorable for maternal, child health, and nutritional outcomes, but the best outcomes will be possible with the right combinations of all three compo- nents comprising a comprehensive framework of planning of births are not identified in the previous studies.5e10 For instance, many states of India overdrive to achieve replace- ment level fertility through female sterilizations which have led to certainly rapid decline in fertility, but at the same time, there is only moderate progress in age at first birth and hardly any improvement in birth interval.24e28 In a paradoxical sit- uation of declining fertility with stalling, unmet need for family planning demands a deeper understanding into plan- ning of births in India. Therefore, this article advances an argument for comprehensive strategy of planning of births (through appropriate timing, spacing, and limiting of births) rather than individual components in a context where the levels of contraceptive use is declining as evident from the http://creativecommons.org/licenses/by/4.0/ https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 516 recent National Family and Health Survey (NFHS).26,27 A few recent studies have used the intersectional axes of the con- tinuum process of planning of births (as a proxy of family planning outcomes) to predict differential maternal and child nutritional outcomes among women who adopted better family planning compared with that of their counterparts. The detailed theoretical framework showing linkages between family planning and maternal and health nutritional out- comes has been discussed elsewhere.3,4 Using the framework prepared in the context of South Asia, this study aims to provide comprehensive empirical evidence showing the ef- fects of the intersectional axes of the planning of births on the select key maternal, child health and nutritional outcomes using a recent nationally representative large-scale database available in India. Methods The data for the current study have been taken from the fourth round of the NFHS conducted during 2015e16. Sampling design and sample size The data consisting of essential health and family welfare indicators have been collected by interviewing 699,686 women. The women were selected using two-stage system- atic random sampling. However, in the analyses of this study, only those women who had at least one child were included. Thus, the final sample accounts for 437,501 and 461,141 for body mass index (BMI) and anemia, respectively. The final sample size in cases of child undernutrition and anemia an- alyses account for 223,011 and 207,594, respectively. The childhood mortality analysis was based on a total sample of 526,868 live births delivered in the 10 years preceding the date of the survey. Outcome variables We have considered maternal, child nutritional status and childhood mortality as outcome indicators for the study. BMI and anemia have been considered as indicators of maternal health and nutritional outcomes, while child health outcomes have been measured by stunting, underweight, anemia, and mortality. Maternal BMI has been categorized into three groups as per WHO guidelines viz. undernourished (<18.5 kg/ m2), normal (18.5e24.9 kg/m2), and obese (�25 kg/m2). The children with less than �2 standard deviations of height-for- age and weight-for-age have been grouped into stunted and underweight, respectively. Any anemia, including severe and less severe, has been considered for both the women and their children. The under-five mortality rates have been estimated indirectly from the survival rates of the life tables of last 10 years of birth history. Predictors The predictor variable, the intersectional axes of the planning of births, has been created using the continuum process of timing, spacing, and limiting of births, namely, the interval between marriage and first birth (IBMFB), the interval between a birth and a subsequent birth (IBBSB), and birth order. The birth order has been categorized into 1, 2, 3, and > 3. The
IBMFB has been grouped into <2 years, 2e3 years, and >3 years
for the first birth order; whereas, for birth order >1, the IBBSB
has been divided into <3 years and �3 years. Altogether, nine intersectional axes of the planning of births have been created. The socio-economic, demographic, and other asso- ciated confounders (region) have been considered for the multivariate analyses. The region is divided into four cate- gories: Uttar Pradesh, Bihar, the Empowered Action Groups (EAG) states, and others (the remaining states of India). Statistical analyses We used both bivariate and multivariate statistical analyses to establish the association between the planning of births and maternal, newborn, and child health and nutritional out- comes. The multinomial logistic regression and multiple classification analysis conversion model was applied to esti- mate the adjusted association between the planning of births and maternal BMI, while separate binary logistic regression models were used to assess the effect of the planning of births on child stunting, underweight, and women and child's ane- mia. As a postestimation of the regression models, the pre- dicted probabilities have been estimated and converted into percentages for ease of interpretation. Furthermore, the assessment of the effects of the planning of births on under- five mortality was carried out using the Cox-proportional hazard regression model. Results Characteristics of the participants The univariate sample distribution by outcome and predictor variables has been displayed in Table 1 for the women and children separately. The estimation of nutritional status and anemia levels in women suggests that approximately 18% are underweight, while 52% are anemic. Approximately, 38%, 35%, and 58% of the children are stunted, underweight, and anemic, respectively. The sample distribution for women varies from a minimum of 5% to a maximum of 19% in different axes of the planning of births (Table 1). The per- centage of the total sample for the children for different intersectional axes of the planning of births ranges from a low of 7% to a high of 18% (Table 1). Overall, across the sample, the percentage of women having shorter (<3 years) birth spacing is higher than the percentage of those who have had the longer spacing (>3 years).

Women’s health outcomes

Table 2 shows the percentage of underweight, normal, and

obese mothers by the axes of the planning of births adjusted

for the other socio-economic and demographic confounders.

As the interest of this study is only the underweight women,

hereafter, normal and obese categories of women will not be

discussed. The results suggest that the probability of being

https://doi.org/10.1016/j.puhe.2018.11.019

Table 1 e Descriptive statistics of the outcome variables and predictor of this study, 2015e16.

Variables Body mass index Women anemia Child undernutrition Child anemia Under-five mortality

n % (95% CI) N % (95% CI) n % (95% CI) n % (95% CI) n % (95% CI)

BMI

Underweight 78,880 18.2 (18.1e18.3) e e e e e e e e

Normal 2,56,214 58.7 (58.5e58.8) e e e e e e e e

Obese 1,02,407 23.1 (23.0e23.3) e e e e e e e e

Women anemia e e 2,40,773 52.2 (52.1e52.4) e e e e e e

Child stunting e e e e 86,239 38.1 (37.9e38.3) e e e e

Child underweight e e e e 77,723 34.5 (34.3e34.7) e e e e

Child anemia e e e e e e 1,19,560 57.6 (57.4e57.8) e e

Planning of births

Order 1 & <2 years of IBMFB 32,512 7.6 (7.6e7.7) 38,184 8.3 (8.2e8.4) 37,926 16.2 (16.1e16.4) 34,638 16.7 (16.5e16.8) 82,663 15.7 (15.6e15.8) Order 1 & 2e3 years of IBMFB 19,764 4.7 (4.6e4.8) 22,970 5.0 (4.9e5.0) 22,668 9.7 (9.6e9.8) 21,097 10.2 (10.0e10.3) 49,936 9.5 (9.4e9.6) Order 1 & >3 years of IBMFB 21,950 5.3 (5.2e5.3) 24,688 5.4 (5.3e5.4) 21,001 9.1 (9.0e9.2) 19,624 9.5 (9.3e9.6) 49,103 9.3 (9.2e9.4)
Order 2 & <3 years of IBBSB 82,169 18.8 (18.7e19.0) 86,361 18.7 (18.6e18.8) 40,845 18.3 (18.2e18.5) 38,224 18.4 (18.2e18.6) 94,918 18.0 (17.9e18.1) Order 2 & >3 years of IBBSB 65,262 14.7 (14.6e14.8) 67,269 14.6 (14.5e14.7) 28,405 13.4 (13.2e13.5) 26,169 12.6 (12.5e12.7) 61,216 11.6 (11.5e11.7)
Order 3 & <3 years of IBBSB 59,675 13.6 (13.5e13.7) 61,603 13.4 (13.3e13.5) 20,662 9.3 (9.2e9.5) 19,476 9.4 (9.3e9.5) 53,563 10.2 (10.1e10.2) Order 3 & >3 years of IBBSB 42,541 9.5 (9.5e9.6) 43,535 9.4 (9.4e9.5) 15,560 7.3 (7.2e7.5) 14,414 6.9 (6.8e7.1) 35,253 6.7 (6.6e6.8)
Order >3 & <3 years of IBBSB 70,349 16.0 (15.9e16.1) 72,224 15.7 (15.6e15.8) 20,757 9.5 (9.4e9.6) 19,673 9.5 (9.4e9.6) 63,044 12.0 (11.9e12.1) Order >3 & >3 years of IBBSB 43,279 9.7 (9.6e9.8) 44,307 9.6 (9.5e9.7) 15,187 7.2 (7.0e7.3) 14,279 6.9 (6.8e7.0) 37,172 7.1 (7.0e7.1)

Total 4,37,501 100 4,61,141 100 2,23,011 100 2,07,594 100 5,26,868 100

n, unweighted sample size; IBMFB, interval between marriage and first birth; IBBSB, interval between birth and subsequent birth; CI, confidence interval; BMI, body mass index.

Upper and lower limit of confidence interval have been shown in the parentheses.

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
4
e
2
5

1
7

https://doi.org/10.1016/j.puhe.2018.11.019

Table 2 e Results from multivariate regression analysis: adjusted percentages of underweight, normal, obese and anemic
women by the selected factors in India, 2015e16 [% (95% CI)].

Variables Body mass indexa Anemiab

Underweight Normal® Obese

Planning of births

Order 1 & <2 years of IBMFB® 19.4 (19.3e19.6) 58.4 (58.4e58.5) 22.2 (22.0e22.3) 51.8 (51.7e51.8) Order 1 & 2e3 years of IBMFB 18.9 (18.7e19.1)* 58.9 (58.8e59.0) 22.2 (22.0e22.4) 52.3 (52.2e52.4) Order 1 & >3 years of IBMFB 18.9 (18.8e19.1) 55.3 (55.2e55.4) 25.7 (25.5e26.0)*** 52.4 (52.3e52.5)*
Order 2 & <3 years of IBBSB 16.8 (16.7e16.9) 55.8 (55.7e55.8) 27.4 (27.3e27.5)** 53.4 (53.3e53.4)*** Order 2 & >3 years of IBBSB 14.1 (14.0e14.2) 54.0 (54.0e54.1) 31.9 (31.7e32.0)* 52.8 (52.8e52.8)***
Order 3 & <3 years of IBBSB 18.1 (18.0e18.2)* 56.7 (56.6e56.7) 25.2 (25.1e25.4) 54.4 (54.4e54.5)*** Order 3 & >3 years of IBBSB 17.4 (17.3e17.5)*** 55.0 (54.9e55.1) 27.6 (27.4e27.8) 54.9 (54.8e54.9)***
Order >3 & <3 years of IBBSB 20.7 (20.6e20.8)*** 57.4 (57.3e57.4) 21.9 (21.8e22.0) 54.6 (54.6e54.7)*** Order >3 & >3 years of IBBSB 22.3 (22.1e22.4)*** 56.9 (56.8e56.9) 20.9 (20.7e21.0)** 55.5 (55.4e55.5)***

Age at marriage in years

<15® 19.3 (19.2e19.4)*** 56.5 (56.5e56.6) 24.2 (24.0e24.3) 54.5 (54.5e54.6) 15e19 19.7 (19.6e19.7)*** 56.6 (56.6e56.7) 23.7 (23.7e23.8)*** 54.3 (54.3e54.3)** 20e24 15.2 (15.1e15.2)*** 55.8 (55.7e55.8) 29.1 (29.0e29.2)*** 52.2 (52.2e52.2)*** 25e29 11.9 (11.8e12.1)*** 53.6 (53.4e53.7) 34.5 (34.3e34.7)*** 50.0 (49.9e50.1) 30þ 11.8 (11.4e12.1)* 55.7 (55.5e56.0) 32.5 (31.9e33.1)*** 47.6 (47.4e47.8) Not reported 18.0 (17.8e18.1)*** 55.4 (55.3e55.5) 26.6 (26.4e26.8)*** 56.5 (56.4e56.5)*** Current age in years 15e19® 36.8 (36.6e37.1)*** 57.0 (56.8e57.1) 6.2 (6.1e6.3) 60.7 (60.6e60.8)*** 20e24 29.9 (29.8e30.0)*** 59.5 (59.5e59.6) 10.6 (10.5e10.6)*** 57.2 (57.2e57.3)*** 25e29 21.9 (21.8e21.9)*** 59.8 (59.8e59.8) 18.3 (18.3e18.4)*** 53.8 (53.7e53.8)*** 30e34 17.0 (17.0e17.1)*** 57.2 (57.1e57.2) 25.8 (25.7e25.9)*** 52.1 (52.1e52.1)*** 35e39 14.4 (14.3e14.4)*** 55.9 (55.9e56.0) 29.7 (29.6e29.8)*** 53.1 (53.0e53.1)*** 40e44 14.0 (13.9e14.0)*** 52.7 (52.7e52.8) 33.3 (33.2e33.4)*** 53.5 (53.5e53.6)*** 45e49 13.3 (13.2e13.4)*** 52.1 (52.0e52.2) 34.6 (34.5e34.7)*** 52.7 (52.6e52.7)*** Place of residence Urban® 9.5 (9.5e9.6) 50.9 (50.8e50.9) 39.6 (39.5e39.7) 51.3 (51.2e51.3) Rural 22.3 (22.2e22.3)*** 58.9 (58.8e58.9) 18.8 (18.8e18.9)*** 54.8 (54.8e54.8)** Religion Hindu® 18.8 (18.8e18.9) 56.7 (56.7e56.7) 24.5 (24.4e24.5) 54.2 (54.2e54.2) Muslim 15.7 (15.6e15.7)*** 53.8 (53.7e53.9) 30.5 (30.4e30.7)*** 51.2 (51.2e51.3)*** Christian 10.6 (10.5e10.8)*** 55.1 (54.9e55.3) 34.2 (33.9e34.6)*** 49.0 (48.9e49.1)*** Others 12.5 (12.3e12.7)*** 54.1 (54.0e54.3) 33.4 (33.1e33.7)** 53.3 (53.2e53.3) Caste Others® 12.7 (12.6e12.8) 53.4 (53.3e53.4) 33.9 (33.8e34.0) 50.2 (50.2e50.2) SC 20.5 (20.4e20.5)* 58.0 (58.0e58.1) 21.5 (21.4e21.6)*** 56.2 (56.2e56.3)*** ST 29.2 (29.1e29.3)*** 58.7 (58.6e58.7) 12.1 (12.0e12.2)*** 60.3 (60.3e60.4)*** OBC 17.5 (17.5e17.6)** 56.3 (56.2e56.3) 26.2 (26.1e26.3)*** 52.9 (52.9e52.9)*** Do not know/not reported 15.7 (15.6e15.8)** 56.1 (56.0e56.3) 28.2 (27.9e28.4)** 52.2 (52.2e52.3)* Women's education Illiterate® 23.8 (23.8e23.9) 58.5 (58.4e58.5) 17.7 (17.7e17.8) 56.3 (56.3e56.4) Primary 19.3 (19.2e19.4)*** 56.4 (56.3e56.4) 24.3 (24.2e24.5)*** 54.5 (54.5e54.6)** Secondary 14.7 (14.6e14.7)*** 54.7 (54.6e54.7) 30.6 (30.6e30.7)*** 52.2 (52.2e52.2)*** Higher 7.5 (7.4e7.5)*** 53.5 (53.4e53.6) 39.0 (38.9e39.2)*** 47.9 (47.8e47.9)*** Partner's education Illiterate® 24.5 (24.4e24.7) 57.9 (57.8e58.0) 17.6 (17.4e17.8) 57.9 (57.8e58.0) Primary 20.8 (20.6e21.0) 56.8 (56.7e56.9) 22.4 (22.1e22.7) 55.1 (55.0e55.2) Secondary 16.2 (16.1e16.3) 54.3 (54.2e54.4) 29.5 (29.3e29.7)** 53.3 (53.2e53.3) Higher 8.9 (8.8e9.1) 52.3 (52.1e52.5) 38.8 (38.4e39.1) 48.9 (48.8e49.0)* Not reported 18.1 (18.1e18.1)* 56.4 (56.4e56.4) 25.5 (25.4e25.5) 53.6 (53.6e53.6) Occupation Not working® 16.5 (16.4e16.6) 54.2 (54.1e54.2) 29.4 (29.2e29.5) 53.4 (53.4e53.4) White collar worker 7.5 (7.2e7.7)*** 52.1 (51.8e52.5) 40.4 (39.9e40.9) 49.6 (49.4e49.7) Agricultural worker 25.0 (24.8e25.2)*** 60.4 (60.3e60.5) 14.6 (14.4e14.7)** 57.3 (57.2e57.3) Service sector/manual worker 17.3 (17.0e17.5) 54.5 (54.3e54.7) 28.2 (27.9e28.6)* 53.7 (53.6e53.8) Do not know/not reported 18.1 (18.1e18.1) 56.4 (56.4e56.4) 25.5 (25.4e25.6) 53.6 (53.6e53.6) Wealth status Poorest® 33.2 (33.2e33.3) 59.6 (59.5e59.6) 7.2 (7.2e7.2) 59.1 (59.1e59.2) Poorer 25.2 (25.1e25.2)*** 60.7 (60.7e60.7) 14.1 (14.1e14.2)*** 55.8 (55.7e55.8)*** Middle 17.3 (17.2e17.3)*** 59.2 (59.2e59.2) 23.6 (23.5e23.6)*** 53.7 (53.7e53.7)*** Richer 10.9 (10.9e11.0)*** 53.7 (53.6e53.7) 35.4 (35.3e35.5)*** 51.4 (51.4e51.5)*** Richest 5.6 (5.6e5.6)*** 48.3 (48.2e48.4) 46.1 (46.1e46.2)*** 48.6 (48.6e48.7)*** p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 518 https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 Table 2 e (continued) Variables Body mass indexa Anemiab Underweight Normal® Obese Exposure to mass media No® 24.3 (24.3e24.4) 58.2 (58.2e58.3) 17.4 (17.4e17.5) 55.9 (55.8e55.9) Partial 14.8 (14.8e14.8)*** 55.3 (55.3e55.4) 29.9 (29.8e29.9)*** 52.5 (52.5e52.5) Full 11.3 (11.2e11.3)*** 53.1 (53.0e53.2) 35.6 (35.5e35.7)*** 51.4 (51.4e51.5) Regions Others® 14.3 (14.3e14.3) 54.0 (53.9e54.0) 31.7 (31.7e31.8) 53.8 (53.7e53.8) EAG 23.3 (23.2e23.3)*** 59.4 (59.4e59.5) 17.3 (17.2e17.4)*** 51.3 (51.3e51.3)*** Uttar Pradesh 19.1 (19.0e19.2)*** 58.1 (58.1e58.2) 22.8 (22.6e22.9)*** 52.9 (52.8e52.9)*** Bihar 26.6 (26.5e26.7)*** 58.7 (58.6e58.8) 14.7 (14.6e14.8)*** 60.8 (60.8e60.8)*** Total 18.0 (18.0e18.1) 56.2 (56.2e56.2) 25.8 (25.7e25.8) 53.7 (53.6e53.7) Number of observations 4,37,501 4,61,141 Log pseudo likelihood �394224 �319963 Wald chi2 30295.36*** 2840.17*** CI, confidence interval; IBMFB, interval between marriage and first birth; IBBSB, interval between birth and subsequent birth; MCA, multiple classification analysis; EAG, Empowered Action Group. Upper and lower limit of confidence interval have been shown in the parentheses; Estimates are weighted with national women weight; *P < 0.10, **P < 0.05, ***P < 0.01. ® stands for reference category of the variable. a Estimates based on multinomial regression and MCA conversional model. b Estimates based on logistic regression model and MCA conversional model. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 19 underweight is higher among women with >3 births and <3 years IBBSB (20.7%, P<0.01) and >3 years of IBBSB (22.3%,
P < 0.01) compared with that for women with one birth and 2e3 years of IBMFB (18.9%, P < 0.10), three births, and <3 years of IBBSB (18.1%, P < 0.10) and 3 births and >3 years of IBBSB
(17.4%, P < 0.01). The EAG states (23.3%, P < 0.01), Uttar Pradesh (19.1%, P < 0.01), and Bihar (26.6%, P < 0.01) have a higher number of underweight mothers compared with that of the rest of India (14.3%). Along with the planning of births and place of residence (region), other factors such as demographic and socio-economic variables were also significantly associ- ated with women being underweight. Table 2 presents the percentage of women with anemia by the planning of births after controlling for the other socio- economic and demographic predictors reported in the previ- ous studies. The results show that the likelihood of anemia is significantly higher among the women with two births and <3 years of IBBSB (53.4%, P < 0.01), three births and <3 years of IBBSB (54.4%, P < 0.01), three births and >3 years of IBBSB
(54.9%, P<0.01), >3 births and <3 years of IBBSB (54.6%, P<0.01) and >3 births and >3 years of IBBSB (55.5%, P < 0.01) compared with that for one birth and >3 years of IBMFB (52.4%, P < 0.10). The risk of anemia was lower in the EAG states (51.3%, P < 0.01) and Uttar Pradesh (52.9%, P < 0.01) but was substantially higher in Bihar (60.8%, P < 0.01) than that in the other states of India (53.8%). Similar to previous studies, the socio-economic and demographic factors reported in Table 3 have also emerged as significant correlates of women's anemia, excluding the occupational status of the mothers. Child health outcomes The association between child health outcomes, notably stunting, underweight, anemia, and under-five mortality, and the planning of births including other confounders is presented in Table 3. The results show that the possibility of stunting is significantly higher among the children born to women with two births and <3 years of IBBSB (41.1%, P < 0.01), three births and <3 years of IBBSB (46.8%, P<0.01), >3 births and <3 years of IBBSB (53.1%, P < 0.01), and >3 births and >3 years of IBBSB
(47.3%, P < 0.01) than that of children born to women with one birth and >2 years of IBMFB (32.5%) and one birth and 2e3 years
of IBMFB (33.0%, P < 0.10). The EAG states (39.7%, P < 0.01), Uttar Pradesh (46.6%, P < 0.01), and Bihar (48.7%, P < 0.01) had a considerably higher rate of stunting than that of the rest of India (32.0%). Excluding the occupational status of the mother, other socio-economic and demographic characteristics of the chil- dren were significantly correlated with their nutritional outcomes. From Table 3, the results of childhood underweight sug- gests that the probability of being underweight is higher among children born to mothers with two births and <3 years of IBBSB (38.1%, P < 0.01), three births and <3 years of IBBSB (43.0%, P < 0.01), >3 births and <3 years of IBBSB (48.4%, P < 0.01), and >3 births and >3 years of IBBSB (43.5%, P < 0.01) than that of children born to mothers with one birth and <2 years of IBMFB (30.7%). The rate of being underweight was higher among children living in EAG states (38.8%, P < 0.05), Uttar Pradesh (39.9%, P < 0.01), and Bihar (44.2%, P < 0.01) compared with that of the other states of India. On the line of previous studies, the other socio-economic factors, barring the occupational statuses of mothers, were significantly correlated with childhood underweight. The estimates showing the association between the ane- mia level of the children and the planning of births is pre- sented in Table 3. The results suggest that the probability of being anemic was higher among the children born to mothers with one birth and <3 years of IBMFB (57.5%, P < 0.01), two births and <3 years of IBBSB (60.2%, P < 0.01), three births and <3 years of IBBSB (63.5%, P < 0.01), three births and >3 years of
IBBSB (59.4%, P<0.01), >3 births and <3 years of IBBSB (64.9%, P < 0.01), and >3 births and >3 years of IBBSB (62.6%, P < 0.01) than that of children born to mothers with one birth and <2 years of IBBSB (54.1%), two births and >3 years of IBBSB (55.8%,

https://doi.org/10.1016/j.puhe.2018.11.019

Table 3 e Results from multivariate regression analysis: adjusted percentages of childhood undernutrition, anemia, and
hazard ratios of under-five mortality by selected factors in India, 2015e16.

Variables Undernutritiona Anemiaa [% (95% CI)] Mortalityb

[hazard ratio (95% CI)]
Stunting

[% (95% CI)]
Underweight
[% (95% CI)]

Planning of births

Order 1 & <2 years of IBMFB® 32.5 (32.0e33.0) 30.7 (30.3e31.2) 54.1 (53.6e54.7) e Order 1 & 2e3 years of IBMFB 33.0 (32.4e33.6)* 32.1 (31.5e32.7) 56.1 (55.4e56.7)* 0.81 (0.74e0.89)*** Order 1 & >3 years of IBMFB 35.7 (35.0e36.3) 33.5 (32.9e34.1) 57.5 (56.8e58.2)*** 0.88 (0.80e0.96)***
Order 2 & <3 years of IBBSB 41.1 (40.6e41.6)*** 38.1 (37.7e38.6)*** 60.2 (59.7e60.7)*** 0.96 (0.89e1.04) Order 2 & >3 years of IBBSB 32.2 (31.6e32.7) 30.1 (29.6e30.6) 55.8 (55.2e56.4)** 0.61 (0.55e0.67)***
Order 3 & <3 years of IBBSB 46.8 (46.1e47.5)*** 43.0 (42.3e43.7)*** 63.5 (62.8e64.2)*** 1.11 (1.02e1.21)** Order 3 & >3 years of IBBSB 37.9 (37.1e38.7) 35.1 (34.3e35.9) 59.4 (58.6e60.3)*** 0.62 (0.56e0.69)***
Order >3 & <3 years of IBBSB 53.1 (52.4e53.8)*** 48.4 (47.6e49.1)*** 64.9 (64.2e65.7)*** 1.37 (1.25e1.49)*** Order >3 & >3 years of IBBSB 47.3 (46.5e48.2)*** 43.5 (42.6e44.4)*** 62.6 (61.7e63.4)*** 0.72 (0.65e0.80)***

Age at marriage in years

<15® 46.4 (45.6e47.1) 42.4 (41.7e43.1) 60.3 (59.6e61.0) e 15e19 41.2 (41.0e41.5) 38.4 (38.2e38.7)** 60.0 (59.8e60.3) 0.97 (0.91e1.02) 20e24 33.8 (33.4e34.2) 31.7 (31.4e32.1) 57.1 (56.7e57.5)* 0.89 (0.83e0.95)*** 25e29 27.4 (26.6e28.3)** 25.5 (24.8e26.3) 51.1 (50.2e52.1) 0.89 (0.79e0.99)** 30þ 25.8 (23.7e28.1)*** 27.5 (25.3e29.8) 51.2 (48.6e53.8) 0.94 (0.76e1.16) Not reported 46.9 (44.2e49.5) 44.7 (42.1e47.4)** 58.2 (55.5e60.7) 1.29 (1.14e1.47)*** Current age in years 15e19® 32.8 (31.6e34.0) 33.5 (32.3e34.7) 64.8 (63.4e66.2) e 20e24 37.7 (37.3e38.1) 35.7 (35.4e36.1) 61.5 (61.2e61.9) 0.63 (0.53e0.76)*** 25e29 38.4 (38.0e38.7) 35.6 (35.2e35.9) 57.9 (57.5e58.2) 0.63 (0.53e0.76)*** 30e34 39.3 (38.8e39.8)* 36.2 (35.7e36.6) 56.1 (55.6e56.6)* 0.62 (0.51e0.74)*** 35e39 42.8 (41.9e43.6) 38.9 (38.1e39.7) 55.3 (54.4e56.1)*** 0.61 (0.50e0.74)*** 40e44 48.6 (46.9e50.2) 45.0 (43.4e46.7) 58.9 (57.2e60.5) 0.71 (0.58e0.87)*** 45e49 53.2 (50.1e56.3) 45.6 (42.5e48.8) 57.9 (54.8e60.9)** 0.76 (0.61e0.94)** Mother's BMI Normal® 37.4 (37.1e37.7) 33.8 (33.5e34.0) e e Underweight 45.9 (45.5e46.4)*** 47.9 (47.5e48.4)*** e 1.07 (1.02e1.12)*** Obese 26.7 (26.2e27.3)*** 21.7 (21.2e22.1)*** e 0.96 (0.90e1.02) Not reported 43.6 (43.0e44.2)*** 39.5 (38.9e40.2)*** e 1.76 (1.66e1.87)*** Mother's anemia Not anemic® e e 50.6 (50.3e51.0) e Anemic e e 64.8 (64.5e65.1)*** e Not reported e e 54.3 (50.7e57.8)* e Place of residence Urban® 31.3 (30.9e31.7) 29.6 (29.2e29.9) 56.1 (55.7e56.6) e Rural 41.5 (41.3e41.8)*** 38.6 (38.3e38.8)*** 59.6 (59.4e59.9)*** 0.99 (0.93e1.04) Religion Hindu® 38.8 (38.6e39.0) 36.6 (36.4e36.9) 58.9 (58.6e59.1) e Muslim 40.1 (39.6e40.6)** 35.2 (34.7e35.7)* 59.3 (58.7e59.8) 0.99 (0.93e1.05) Christian 30.1 (28.7e31.5)*** 27.3 (26.0e28.7)*** 45.9 (44.4e47.5)*** 1.43 (1.31e1.55)*** Others 33.4 (32.2e34.6) 31.5 (30.4e32.7) 58.8 (57.5e60.1)** 1.04 (0.93e1.16) Caste Others® 30.9 (30.5e31.4) 28.9 (28.4e29.3) 54.7 (54.2e55.2) e SC 43.1 (42.6e43.5)*** 39.6 (39.1e40.0)*** 60.8 (60.3e61.2)*** 1.19 (1.11e1.27)*** ST 44.2 (43.6e44.9)*** 45.4 (44.8e46.1)*** 63.8 (63.1e64.5)*** 1.26 (1.17e1.36)*** OBC 39.0 (38.7e39.3)*** 35.9 (35.6e36.2)*** 58.7 (58.4e59.0)*** 1.03 (0.97e1.10) Do not know/not reported 34.8 (33.8e35.8) 30.2 (29.2e31.1)* 53.1 (52.1e54.2)*** 1.02 (0.91e1.14) Mother's education Illiterate 51.1 (50.7e51.5) 47.2 (46.8e47.6) 65.0 (64.6e65.4) e Primary 43.8 (43.2e44.3)*** 40.3 (39.8e40.9)*** 60.7 (60.1e61.2)*** 0.93 (0.88e0.98)*** Secondary 33.0 (32.7e33.3)*** 31.4 (31.1e31.7)*** 55.8 (55.5e56.1)*** 0.72 (0.69e0.76)*** Higher 21.1 (20.6e21.6)*** 19.1 (18.6e19.6)*** 49.6 (48.9e50.3)*** 0.49 (0.43e0.57)*** Father's education Illiterate® 51.4 (50.2e52.6) 47.5 (46.3e48.8) 64.5 (63.2e65.7) e Primary 43.4 (42.1e44.8) 41.2 (39.8e42.5) 61.9 (60.5e63.2) 0.90 (0.78e1.02)* Secondary 35.9 (35.2e36.6)*** 33.2 (32.6e33.9)** 57.1 (56.4e57.8) 0.80 (0.71e0.89)*** Higher 23.9 (22.8e25.1)*** 22.4 (21.3e23.6)*** 51.1 (49.7e52.6) 0.60 (0.47e0.76)*** Not reported 38.8 (38.6e39.0) 36.2 (36.0e36.5) 58.7 (58.5e59.0) 0.81 (0.56e1.17) p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 520 https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 Table 3 e (continued) Variables Undernutritiona Anemiaa [% (95% CI)] Mortalityb [hazard ratio (95% CI)] Stunting [% (95% CI)] Underweight [% (95% CI)] Mother's occupation Not working® 36.7 (36.1e37.2) 34.0 (33.5e34.6) 58.1 (57.5e58.7) e White collar worker 27.5 (24.8e30.4) 23.7 (21.1e26.4) 49.5 (46.4e52.7) 0.94 (0.69e1.27) Agricultural worker 46.3 (44.8e47.8) 44.2 (42.8e45.7) 61.5 (60.1e63.0) 0.86 (0.77e0.97)** Service sector/manual worker 42.6 (40.7e44.6) 39.6 (37.7e41.5) 59.1 (57.1e61.1) 1.02 (0.87e1.18) Do not know/not reported 38.8 (38.6e39.0) 36.2 (36.0e36.5) 58.7 (58.5e59.0) 1.06 (0.73e1.53) Wealth status Poorest® 51.8 (51.4e52.2) 49.0 (48.6e49.4) 64.2 (63.7e64.6) e Poorer 43.8 (43.4e44.3)*** 40.8 (40.4e41.2)*** 59.9 (59.4e60.3)*** 0.86 (0.82e0.90)*** Middle 36.8 (36.3e37.2)*** 33.7 (33.3e34.2)*** 59.0 (58.6e59.5) 0.79 (0.74e0.84)*** Richer 29.5 (29.0e29.9)*** 27.7 (27.3e28.1)*** 54.5 (54.0e55.0)*** 0.72 (0.67e0.78)*** Richest 22.5 (22.0e23.0)*** 20.4 (20.0e20.9)*** 51.9 (51.3e52.5)*** 0.59 (0.53e0.65)*** Exposure to mass media No® 45.7 (45.4e46.1) 42.6 (42.3e42.9) 61.8 (61.5e62.2) e Partial 34.7 (34.4e35.0)** 32.5 (32.2e32.8)* 56.9 (56.5e57.2) 0.92 (0.89e0.97)*** Full 29.2 (28.7e29.8)*** 27.0 (26.4e27.6)*** 54.4 (53.8e55.1) 0.86 (0.78e0.94)*** Sex of the children Male® 39.2 (38.9e39.5) 36.4 (36.2e36.7) 58.5 (58.2e58.8) e Female 38.1 (37.8e38.4)*** 35.7 (35.4e36.0)*** 58.8 (58.5e59.2) 1.18 (1.13e1.22)*** Age of the children 1 year® 21.8 (21.4e22.2) 28.0 (27.6e28.5) 68.5 (67.9e69.2) e 2 years 43.0 (42.5e43.5)*** 35.4 (35.0e35.9)*** 70.6 (70.2e71.0)*** e 3 years 43.0 (42.5e43.5)*** 38.0 (37.6e38.5)*** 62.4 (61.9e62.8)*** e 4 years 43.7 (43.2e44.1)*** 38.7 (38.3e39.2)*** 52.2 (51.8e52.7)*** e 5 years 40.3 (39.8e40.8)*** 39.4 (39.0e39.9)*** 44.7 (44.3e45.2)*** e Regions Others® 32.0 (31.7e32.3) 30.7 (30.4e30.9) 55.9 (55.5e56.2) e EAG 39.7 (39.3e40.1)*** 38.8 (38.4e39.2)** 57.8 (57.3e58.2)*** 1.41 (1.34e1.49)*** Uttar Pradesh 46.6 (46.1e47.1)*** 39.9 (39.4e40.4)*** 63.4 (62.9e63.9)*** 2.02 (1.90e2.15)*** Bihar 48.7 (48.1e49.3)*** 44.2 (43.6e44.8)*** 63.6 (63.0e64.1)*** 1.13 (1.04e1.22)*** Total 38.7 (38.5e38.9) 36.1 (35.9e36.3) 58.7 (58.4e58.9) e Number of observations 2,23,011 2,23,011 2,07,594 5,07,265 Log pseudo likelihood �132135 �131640 �127078 �142369 Wald chi2 9130.37*** 7454.68*** 7011.4*** 4470.23*** CI, confidence interval; IBMFB, interval between marriage and first birth; IBBSB, interval between birth and subsequent birth; MCA, multiple classification analysis; EAG, Empowered Action Groups. Estimates are weighted with national women weight; Cox proportional hazard ratios has been presented for under-five child mortality; upper and lower limit of confidence interval have been shown in the parentheses; *P < 0.10, **P < 0.05, ***P < 0.01. ® stands for reference category of the variable. a Estimates based on logistic regression and MCA conversion model. b Estimates based on Cox proportional hazard regression model. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 21 P < 0.01), and one birth and 1e2 years of IBMFB (56.1%, P < 0.10). The EAG states (57.8%, P < 0.01) have lower preva- lence, while Uttar Pradesh (63.4%, P < 0.01) and Bihar (63.6%, P < 0.01) have higher prevalence of anemia among the chil- dren than that of the other states of India (55.9%). Child ane- mia was also significantly associated with other socio- economic and demographic characteristics, except for mother's occupational status and exposure to mass media, the father's educational level, and the sex of the child. However, in case of child anemia, the planning of births emerges as the most significant predictor, showing greater differences in the anemia levels compared with that of any of the socio- economic indicators. The KaplaneMeier estimates of survival during the child- hood are displayed in Fig. 1. The curves suggest that the probability of dying is highest among the children of mothers with >3 births and <3 years of IBBSB, followed by those with three births and <3 years of birth spacing, >3 births and <3 years of birth spacing, and two births and <3 years of birth spacing. There was not much variation in the probability of dying observed among the rest of the groups. The adjusted hazard ratios from the Cox proportional haz- ard regression analysis showing the association between under-five mortality and the planning of births after control- ling for other socio-economic and demographic characteristics are presented in Table 3. The results demonstrate that the risk of death is lower among children born to mothers with one birth and 2e3 years of IBMFB (hazard ratio [HR] 0.81, P < 0.01), one birth and >3 years of IBMFB (HR 0.88, P < 0.01), two births and >3 years of IBBSB (HR 0.61, P < 0.01), three births and <3 years of IBBSB (HR 0.62, P<0.01), and >3 births and >3 years of
IBBSB (HR 0.72, P < 0.01) compared with that for children born to mothers with one birth and >2 years of IBMFB. Compared to the
children born to mothers with one birth and >2 years of IBMFB,

https://doi.org/10.1016/j.puhe.2018.11.019

Fig. 1 e KaplaneMeier survival estimates of under-five children in India, 2015e16. Or 1 & <2 years ¼ Order 1 & <2 years of IBMFB; Or 1 & 2e3 years ¼ Order 1 & 2e3 Years of IBMFB; Or 1 & >3 years ¼ Order 1 & >3 Years of IBMFB; Or 2 & <3 years ¼ Order 2 & <3 Years of IBBSB; Or 2 & >3 years ¼ Order 2 & >3 Years of IBBSB; Or 3 & <3 years ¼ Order 3 & <3 Years of IBBSB; Or 3 & >3 years ¼ Order 3 & >3 Years of IBBSB; Or >3 & <3 years ¼ Order >3 & <3 Years of IBBSB; Or >3 & >3
years ¼ Order >3 & >3 Years of IBBSB. IBMFB, interval between marriage and first birth; IBBSB, interval between birth and
subsequent birth.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 522
the risk of child death was higher among those with three

births and <3 years of IBBSB (HR 1.11, P<0.01) and >3 births and
<3 years of IBBSB (HR 1.37, P < 0.01). The hazard of childhood death was higher in the EAG states (HR 1.41, P < 0.01), Uttar Pradesh (HR 2.02, P < 0.01), and Bihar (HR 1.13, P < 0.01) than that in the rest of the country. Except the place of residence, other socio-economic and demographic factors were significantly correlated with the under-five mortality. Discussion This study provides a comprehensive assessment of the ef- fects of the planning of births (by adopting family planning) on select key maternal, child health, and nutritional outcomes using the most recent national family health survey data. The findings suggest that the selected maternal, child health, and nutritional outcomes (viz. BMI and anemia level of women as well as the stunting, underweight, anemia, and mortality of children) significantly differed by the intersectional axes of the planning of births. In particular, the risk of childhood underweight is considerably higher among children born to women who have had three or more births and less than 3 years of birth spacing compared with risk for those with one birth and more than 2 years of spacing between marriage and their first birth. Similarly, among the women with more than one birth and shorter birth spacing, the likelihood of being anemic is higher than that among those with one birth and greater than 2 years of spacing between marriage and the first birth. Compared to the children with mothers who have longer birth spacing and a lower number of births, the likeli- hood of stunting, underweight, and anemia is considerably higher among those born to women with more than one birth and less than 3 years of birth spacing. The hazard of child death is substantially higher among those with three births or more and less than 3 years between births, while it is lower among those with one birth and more than 2 years between births, as well as those with more than two births and more than 3 years between births. Thus, this study has found that the planning of births is significantly associated with maternal, child health outcomes. The findings from the present study support the argu- ments put forward by the previous studies in global and South Asian contexts.3,5,14,16,18 In particular, early marriage, early childbearing, and lack of family planning lead to a shorter time between marriage and a first birth, which results in poor nutritional outcomes of the women.16,18 Furthermore, women with more births and shorter birth spacing have a greater risk of being underweight than their counterparts do.3 This study shows that the risk of anemia among women and children is higher for a greater number of births, while previous studies have suggested that the direction of the relationship between birth spacing and maternal and child anemia is not just one way; instead, the direction varies from positively significant to not significant.14 In the present study, among the women and their children from their first birth, the risk of being anemic https://doi.org/10.1016/j.puhe.2018.11.019 https://doi.org/10.1016/j.puhe.2018.11.019 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 23 was higher among those with more than 3 years of spacing between marriage and the birth than that among those with a gap of less than 2 years. Further research is needed for this contradictory findings. The childhood nutritional outcomes, notably stunting and underweight, were also associated with the planning of births for timing, spacing, and limiting, which is constituent with that reported previously.4 Similar to pre- vious studies,29,30 this study's findings also suggest that the children born to mothers with a greater number of births with shorter birth spacing have a greater risk of under-five death compared with the risk for those born to mothers as their first birth. The findings of this study are based on a more nuanced assessment of a larger number of indicators and re- strengthen the argument that the intersectional axes of the planning of births comprising the timing, spacing, and limiting of births have a biodemographic relationship with maternal and child health outcomes. The adolescent period of women is an important period of physical growth and devel- opment.5 Pregnancy and lactation during this period leads not only to the depletion of nutritional elements (fat, iron, and folate) among them but also in their children. This depletion of fat and iron among the mothers causes underweight and anemia among them and poor pregnancy and delivery out- comes among their children. These poor outcomes drive the children to be underweight, stunted, and anemic, which in turn results in childhood mortality. In similar findings, shorter birth spacing and repeated pregnancy and delivery (higher number of births) hinder women in recovering their body weights and other micronutrients. As a result, these women are more likely to experience poor pregnancy and delivery outcomes such as IUGR, LBW, premature birth, and small birth size, which are risk factors for adverse childhood growth and higher mortality. As mentioned earlier, the pathways through which family planning affect maternal, newborn, and child health and nutrition are both direct and indirect. By helping couples attain the number of children they want at the healthiest times in their lives, family planning can benefit mothers, in- fants, and children. The adequate spacing of births allows women's bodies to recover and restock vital nutrients and leads to better maternal, newborn, and child health and nutritional outcomes, such as healthy pregnancy outcomes and lower childhood mortality. Family planning can help ad- olescents to delay pregnancy until an ideal reproductive age (>18 years) and thus can improve their growth and develop-
ment and reduce the risk of poor nutritional and health out-

comes for their infants. A growing body of evidence shows

that intentional pregnancy can also influence nutritional

outcomes.4,6,7,9 The children from unintended pregnancies

have a higher risk of poor nutrition, underscoring the impor-

tant role of family planning.10,21

Family planning indirectly affects nutrition via its impact on

infant and young child feeding practices. When births are well

spaced, mothers have more time, energy, and resources to

adequately breastfeed and feed their young infants and chil-

dren, respectively. Research shows that when pregnancies are

planned and occur when women are older than 18 years,

breastfeeding practices improve, leading to improved nutrition

of the infants.2e4,11 When unplanned pregnancies are avoided,
women are less exposed to the risks of dying due to pregnancy

and childbirth. Since mothers play a crucial role in feeding their

families, reductions in maternal death can positively influence

infant and child nutrition. Finally, family planning can have an

indirect impact on nutrition by reducing unintended pregnan-

cies among adolescents, allowing them to stay in school

and complete more years of education. Research shows that

greater education among women leads to greater productivity,

empowerment, and control of resourcesdallowing them to

make better choices that ultimately benefit both them and their

children’s health and nutrition.4,6e10

This study has some limitations and strengths that must

be noted. We used cross-sectional data to draw the underlying

association between the planning of births and maternal and

child health outcomes, but for establishing a perfect causal

relationship, a longitudinal design of experiments is required.

As the purpose of this study is to link family planning to

maternal, newborn, and child health and nutritional out-

comes, both past and current use of contraception are not

appropriate for establishing this relationship. However, by

constructing a proxy variable for representing family planning

use in the form of the planning of births (the continuum

process of timing, spacing, and limiting), this study fills a

critical gap by providing timely empirical evidence on linking

family planning to maternal, newborn, and child health and

nutritional outcomes. The spacing between marriage and first

birth was included in the planning of births variable, which is

often overlooked in most studies on the process of linking

family planning to maternal, newborn, and child health and

nutritional outcomes, and it emerged as one of the essential

biodemographic factors of the latter. This study advances the

strengthening of strategies of integration of family planning

with maternal, newborn, and child health and nutrition in

India.

Conclusions

This study provides several cross-cutting implications for

clinical practice and health policymaking. The findings from

this study show that the planning of births has a bio-

demographic advantage in improving maternal and child

health outcomes. Given the evidence that more than one-

fourth of all adolescent girls experience child marriages in

India,2 a longer spacing between marriage and the first birth

could improve the maternal, child health outcomes, notably,

underweight of women and stunting, underweight, and

mortality of children. Family planning programs in India have

always been heavily skewed toward limiting methods, espe-

cially female sterilization, since it is cost effective for the

policymakers to control population growth,30 but appropriate

spacing between births also has an important effect on better

maternal and child health outcomes. Apart from limiting,

spacing methods of family planning must be emphasized for

better health outcomes. Fertility declines have almost reached

the replacement level in the country despite a lower level of

contraceptive use and a high unmet demand for family

planning;31 this does not mean in any way that the relevance

of family planning can be underestimated. The high unmet

demand for family planning in both spacing and limiting must

be addressed not only to meet population stabilization goals

https://doi.org/10.1016/j.puhe.2018.11.019

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 524
but also to accomplish the SDG goals of health for all.

Although the FP2020 vision document of the Ministry of

Health and Family Welfare aimed to reach an additional 48

million women and girls with family planning methods,32 the

findings from this study encourage the strategy of universal

coverage of contraceptive use in India for achieving the ho-

listic benefits of family planning, such as better maternal,

newborn, and child health and better nutritional outcomes.

The integration of family planning, maternal, newborn,

and child health and nutritional programs has multiple op-

portunities to provide family planning counseling and ser-

vices to the mothers during their postnatal care at the health

centers. This integration is also less time consuming and is

cost effective for the healthcare system; it also helps in

improving birth spacing and avoids unintended births. The

evidence shows that a few countries have harvested these

opportunities for providing family planning counseling and

services to this ‘captive audience’.33 In India, to a great extent,

the integration of family planning with maternal, newborn,

and child health and nutrition has not succeeded to the extent

that it was targeted under successive population policies due

to lack of a true integration strategy at the implementation

level, service delivery at the peripheral level, a shortage of

frontline health workers, and consequent overburdening of

them.34 United Nations has defined linkages as ‘policy, pro-

gramatic, services, and advocacy of bidirectional synergies’

between maternal, newborn, and child health and nutrition

and family planning.35 In contrast to linkages, which exist at

multiple levels, these organizations define integration at the

service delivery level only as ‘different kinds of services or

operational programs joined together to ensure and perhaps

maximize collective outcomes.’ Therefore, India needs to

revise their integration mechanism and eliminate disconnects

that hinder the delivery of these services.
Author statements

Ethical approval

This study used publicly available secondary sources of data.

Thus, it does not require ethical approval.

Funding

This study was funded by the Bill & Melinda Gates Foundation,

India Country Office, New Delhi, India. Grant Number:

OPP1142874.

Competing interest

The authors have no conflicts of interest to declare.
r e f e r e n c e s

1. United Nations sustainable development goals. United Nations.
2016. Available from: https://sustainable_development.un.
org/sdg3. [Accessed 16 February 2018].
2. Allison A, Anthony-Kouyate R, Blanchard H, Deller B,
Foreman M, Galloway R, et al. Maximizing synergy between
maternal, infant and young nutrition and pregnancy prevention: a
discussion paper prepared for maternal, infant and young child
nutrition (MIYCN) and family planning (FP) integration technical
meeting. 2010. Available from: www.k4health.org/sites/
default/files/Position%20Paper . [Accessed 30 March 2015].

3. Rana MJ, Goli S. Family planning and its association with
nutritional status of women: investigation in select South
Asian countries. Indian J Human Dev 2017;11(1):56e75.

4. Rana MJ, Goli S. Does planning of births affect childhood
undernutrition? Evidence from demographic and health
surveys of selected South Asian countries. Nutrition
2018;47(3):90e6.

5. Naik R, Smith R. Impacts of family planning on nutrition.
Washington (DC): Futures Group. Health Policy Project; 2015.

6. Hamdallah M, Foehringer Merchant H, Creel Harris N. The
added value of integrating family planning into community-based
services: learning from implementation. Arlington, VA:
Advancing Partners & Communities; 2017.

7. Borwankar R, Amieva S. Desk review of programs integrating
family planning with food security and nutrition. Washington, DC:
FHI 360/FANTA; 2015.

8. Ringheim K, Gribble J, Foreman M. Integrating family planning and
maternal and child health care: saving lives, money, and time (Policy
brief). Washington, DC: Population Reference Bureau; 2011.

9. Sebert KA, Gavin L, Galavotti C. The integration of family
planning with other health services: a literature review. Int
Perspect Sex Reprod Health 2010;36(4):189e96.

10. Tsui AO, McDonald-Mosley R, Burke AE. Family planning and
the burden of unintended pregnancies. Epidemiol Rev
2010;32(1):152e74.

11. Goli S, Rammohan A, Singh D. The effect of early marriages
and early childbearing on women’s nutritional status in India.
Matern Child Health J 2015;19(8):1864e80.

12. Rutstein SO. Further evidence of the effects of preceding birth
intervals on neonatal infant and under-five-years mortality and
nutritional status in developing countries: evidence from the
demographic and health surveys. 2008.

13. Pasricha SR, Black J, Muthayya S, Shet A, Bhat V, Nagaraj S,
Prashanth NS, Sudarshan H, Biggs BA, Shet AS. Determinants
of anemia among young children in rural India. Pediatrics
2010;126(1):e140e9.

14. Dewey KG, Cohen RJ. Does birth spacing affect maternal or
child nutritional status? A systematic literature review.
Matern Child Nutr 2007;3(3):151e73.

15. King JC. The risk of maternal nutritional depletion and poor
outcomes increases in early or closely spaced pregnancies. J
Nutr 2003;133(5):1732Se5S.

16. Rah JH, Christian P, Shamim AA, Arju UT, Labrique AB,
Rashid M. Pregnancy and lactation hinder growth and
nutritional status of adolescent girls in rural Bangladesh. J
Nutr 2008;132(8):1505e11.

17. Winkvist A, Rasmussen KM, Habicht JP. A new definition of
maternal depletion syndrome. Am J Public Health
1992;82(5):691e4.

18. Gold RB. Family planning and health care reform: the benefits
and challenges of prioritizing prevention. Guttmacher Policy
Rev 2009;12(1):19e24.

19. Bongaarts J. Does family planning reduce infant mortality
rates? Popul Dev Rev 1987;13(2):323e34.

20. DaVanzo J, Hale L, Razzaque A, Rahman M. The effects of
pregnancy spacing on infant and child mortality in Matlab,
Bangladesh: how they vary by the type of pregnancy outcome
that began the interval. Popul Stud 2008;62(2):131e54.

21. Singh A, Chalasani S, Koenig MA, Mahapatra B. The
consequences of unintended births for maternal and child
health in India. Popul Stud 2012;66(3):223e39.

https://sustainable_development.un.org/sdg3

http://www.k4health.org/sites/default/files/Position%20Paper

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref3

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref4

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref5

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref6

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref7

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref8

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref9

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref10

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref11

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref12

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref13

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref14

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref15

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref16

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref17

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref18

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref19

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref20

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref21

https://doi.org/10.1016/j.puhe.2018.11.019

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 e2 5 25
22. Gipson JD, Koenig MA, Hindin MJ. The effects of
unintended pregnancy on infant, child, and parental
health: a review of the literature. Stud Fam Plann
2008;39(1):18e38.

23. Roy TK, Sinha RK, Koenig M, Mohanty SK, Patel SK.
Consistency ans predictive ability of fertility preference
indicators: longitudinal evidence from rural India. Int Fam
Plann Perspect 2008:138e45.

24. International Institute for Population Sciences (IIPS). National
family health survey (MCH and family planning), 1992e93: India.
Mumbai: IIPS; 1995.

25. International Institute for Population Sciences (IIPS) and ORC
Macro. National family health survey (NFHS-2), 1998e99: India.
Mumbai: IIPS; 2000.

26. International Institute for Population Sciences (IIPS) & Macro
International. National family health survey NFHSe3, 2005e06:
India, vol. II. Mumbai: IIPS; 2007.

27. National family health survey-4 (2015e16). International institute
for population sciences and Ministry of health and family welfare.
2017.

28. Srinivasan K. Population concerns in India: shifting trends, policies,
and programs. SAGE Publishing India; 2017.

29. Fotso JC, Cleland J, Mberu B, Mutua M, Elungata P. Birth
spacing and child mortality: an analysis of prospective data
from the Nairobi urban health and demographic surveillance
system. J Biosoc Sci 2013;45(6):779e98.

30. Miller JE, Trussell J, Pebley AR, Vaughan B. Birth spacing and
child mortality in Bangladesh and the Philippines.
Demography 1992;29(2):305e18.

31. Goli S. Eliminating child marriage in India: progress and prospects.
2016. Available from: https://www.actionaidindia.org/
publication/eliminating-child-marriage-in-india/. [Accessed
16 February 2018].

32. Government of India. India’s Vision FP 2020. Ministry of Health
and Family Welfare (MOHFW), Government of India, New Delhi.
2014. Available from: http://ec2-54-210-230-186.compute-1.
amazonaws.com/wp-content/uploads/2015/04/Indias-Vision-
FP2020 . [Accessed 17 August 2015].

33. Integrating family planning and maternal and child health services:
history reveals a winning combination. Population Reference
Bureau; 2011. Available from: https://www.prb.org/family-
planning-maternal-child-health-integration-programs/.

34. Arokiasamy P, Shekhar C, Srinivasan K, Goli S. Family welfare
programme in India: expenditure vs performance. Econ Pol
Wkly 2011;46(43):127e34.

35. IPPF U, WHO U. Sexual and Reproductive Health & HIV/AIDS: a
framework for priority linkages. Controlling sexually transmitted
and reproductive tract infections. WHO; 2005.

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref22

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref23

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref24

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref25

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref26

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref27

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref28

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref29

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref30

Eliminating-Child-Marriage-in-India

http://ec2-54-210-230-186.compute-1.amazonaws.com/wp-content/uploads/2015/04/Indias-Vision-FP2020

https://www.prb.org/family-planning-maternal-child-health-integration-programs/

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref34

http://refhub.elsevier.com/S0033-3506(18)30389-5/sref35

https://doi.org/10.1016/j.puhe.2018.11.019

Planning of births and maternal, child health, and nutritional outcomes: recent evidence from India
Introduction
Methods
Sampling design and sample size
Outcome variables
Predictors
Statistical analyses

Results
Characteristics of the participants
Women’s health outcomes
Child health outcomes

Discussion
Conclusions

Author statements
Ethical approval
Funding
Competing interest

References

Exploring-socio-economic-inequalities-in-the-use-of-medicines–_2019_Public-

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e9
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Exploring socio-economic inequalities in the use of
medicines: is the relation mediated by health
status?
J. Maron a,b,*, E. Gomes de Matos a, D. Piontek a, L. Kraus a,c,d, O. Pogarell b

a IFT Institut für Therapieforschung, Department of Epidemiology and Diagnostics, Leopoldstr. 175, 80804 Munich,

Germany
b LMU Munich, University Hospital, Department of Psychiatry and Psychotherapy, Nußbaumstr. 7, 80336 Munich,

Germany
c Stockholm University, Department for Public Health Sciences, Sveav€agen 160, 10691 Stockholm, Sweden
d Institute of Psychology, ELTE E€otv€os Lor�and University, 1075 Budapest, Kazinczy utca 23-27, Hungary
a r t i c l e i n f o

Article history:

Received 16 April 2018

Received in revised form

3 December 2018

Accepted 19 December 2018

Available online 13 February 2019

Keywords:

Social inequalities in medicine use

Pharmacoepidemiology

Social-epidemiology

Socio-economic status

Mental and physical self-rated

health

Germany
* Corresponding author. Department of Ep
Munich, Germany; Department of Psychiat
Tel.: þ49 176 20187741; fax: þ49 89 360804 1

E-mail addresses: julian.maron@gmx.de (
de (L. Kraus), Oliver.Pogarell@med.uni-muen
https://doi.org/10.1016/j.puhe.2018.12.018
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objectives: This study evaluated mediating effects of the health status on the association

between socio-economic status (SES) and medicine use. It was hypothesized that more

privileged people show a reduced use of medicines, as compared with the underprivileged,

because of their superior health status. It was further hypothesized that people may apply

medication based on their type of health complaint (ill physical versus mental status).

Study design: Data were taken from the 2012 German Epidemiological Survey of Substance

Abuse, a nationally representative cross-sectional study of n ¼ 9084 individuals of the
German general population aged 18e64 years.

Methods: Direct and indirect effects of SES on weekly use of analgesics and sedatives/

hypnotics were examined by applying generalized structural equation modeling. Self-rated

physical and mental health statuses were considered as potential mediators. SES was

measured by using educational level as a proxy. All analyses were gender-stratified.

Results: Among men, both physical and mental health mediated the path from SES to the

use of analgesics and sedatives/hypnotics, respectively, with a stronger effect of physical

health on analgesic use and mental health on sedative/hypnotic use. These effects were

only partially found among women.

Conclusions: Social inequalities in health seem to have substantial impact on the prevalence

of medicine use. Identification and elimination of the reasons for poor health among

people of low SES may, therefore, not only help to reduce health inequalities directly. A

decline in the use of medicines would also result in less side-effects and a reduced number

of people with medicine-related misuse and addiction.

© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
idemiology and Diagnostics, IFT Institut für Therapieforschung, Leopoldstr. 175, 80804
ry and Psychotherapy, University of Munich, Nußbaumstr. 7, 80336 Munich, Germany.
9.
J. Maron), GomesdeMatos@ift.de (E. Gomes de Matos), Piontek@ift.de (D. Piontek), Kraus@ift.
chen.de (O. Pogarell).

mailto:julian.maron@gmx.de

mailto:GomesdeMatos@ift.de

mailto:Piontek@ift.de

mailto:Kraus@ift.de

mailto:Oliver.Pogarell@med.uni-muenchen.de

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2018.12.018&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e92
Introduction

The prevalence of medicine use as well as sales of medicines

have risen in the German general population and in other

European countries, a development that may have serious

consequences for public health.1,2 Evidence also points to-

ward an unequal distribution of the use of medicines across

the population. Several European studies have revealed a

strong negative association between socio-economic status

(SES) and the use of medicines, that is, the lower the SES, the

higher the use of medicines.1,3e6

Studies exploring socio-economic inequalities in the use of

medicines are rare. However, results suggest that other fac-

tors and, particularly, a person’s health status may explain the
association between SES and medicine use. First, research

showed that the correlation of SES and medicine use dis-

appeared when analyses were adjusted for health.7 Second,

several studies indicate that health status strongly predicts

the use of medicines.1,8e10 People with an ill health status

might be either more prone to use medicines prescribed by a

physician or may tend to use over-the-counter (OTC) products

to medicate themselves. At the same time, it is well known

that SES and health are positively correlated, that is, the

higher the SES, the better the health status.11,12

In accordance with these references, health status might

play a mediating role between SES and medicine use. How-

ever, studies examining this assumption are lacking so far. In

case of full mediation by individual health status, measures to

reduce inequalities in medicine use would have to focus on

the health status itself, with an emphasis on preventing ill

health among lower SES groups. On the other hand, if in-

equalities in medicine use were not fully explained by differ-

ences in health, SES would determine a person’s health status
but also directly impact on the use of medicines. In this case, a

(complete) reduction of inequalities in medicine use would

not be reached by reducing health inequalities, and further

explorations of direct reasons for SES inequalities in medicine

use would be needed.

The aim of the present study was to evaluate the mediating

role of the health status between SES and medicine use. In

particular, it considered direct and indirect pathways from

SES to the use of two different classes of medicines most

prevalently used in the German population, namely analge-

sics and sedatives/hypnotics.13 Self-rated physical and mental

health statuses were considered as potential mediators be-

tween the paths from SES to analgesic or sedative/hypnotic

use. It was hypothesized that lower rates of medicine use

among people of higher SES, compared with higher rates of

medicine use among people of lower SES, are exclusively

attributable to their superior health status. Because patterns

of medicine use should strongly depend on self-perceived

complaints (physical and mental symptoms) and on the

intended effect of the drug, it was also hypothesized that

physical health should be of greater importance in the rela-

tionship between SES and analgesic use (as compared with

sedative/hypnotic use), whereas mental health should be of

greater importance in the relationship between SES and

sedative/hypnotic use.
Methods

Study design and sample

Data were taken from the 2012 German Epidemiological Sur-

vey of Substance Abuse, a nationally representative cross-

sectional study of n ¼ 9084 individuals drawn from the 18-
year-old to 64-year-old German-speaking population living

in private households in Germany (response rate: 53.6%). A

two-stage sampling approach with oversampling younger and

undersampling older birth cohorts was applied to achieve a

representative sample of the German population (aged 18e64

years). Data collection was conducted from April to August

2012 using a mixed-mode design with paper-and-pencil

questionnaires, computer-assisted telephone interviews

(CATIs), and online questionnaires.14

Measures

Major outcome variables of the analytical model were weekly

analgesic use and weekly sedative/hypnotic use. Weekly

analgesic use was assessed by asking ‘During the last 30 days,

how often did you use analgesics (painkillers)?“, and by

providing five response categories (‘did not use it at all’, ‘less

frequent than once a week’, ‘once a week’, ‘several times a

week’, ‘daily’). A list of the most common pharmaceuticals

was provided to facilitate the allocation of a drug. Individuals

who used analgesics at least once a week in the last 30 days

were treated as weekly analgesic users; all others, i.e. non-

weekly users and non-users, were defined as the reference

group. Weekly sedative/hypnotic use (including anxiolytics)

was assessed accordingly. Weekly use was chosen as a mea-

sure of medicine use in this study because weaker indicators

(e.g. monthly use) are unlikely to vary by SES.6

SES is acting as exposure variable and was assessed by

using educational level as a proxy. The highest educational

level achieved was categorized into three groups, based on the

International Standard Classification of Education: low (less

than primary, primary, and secondary I), medium (secondary

II, post secondary/non-tertiary), and high (tertiary I and II or

higher) education.15 Individuals currently attending a school

of general education were excluded. In the statistical analysis,

SES was dummy-coded with the highest status acting as the

reference category. To simplify the outputs of a complex sta-

tistical model, tables and figures show results for the com-

parison of the extreme SES groups only (i.e., low versus high

SES is shown, medium versus high SES was also estimated but

is not shown).

Potential mediators between SES and medicine use were

self-rated physical and mental health. It was assessed by

asking ‘How would you rate your current health status?’ and

‘How would you rate your current mental well-being?’,

respectively. Five-point Likert scales were dichotomized into

good (very good, good) and ill (fair, poor, very poor) health,16

whereas the former group was defined as the reference cate-

gory. Self-rated health has been evaluated as a strong and

consistent predictor of mortality and functional health

decline.17

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e9 3
To control for potential confounder effects, age (contin-

uous, 18e64 years), marital status (married, unmarried),

regional distribution (East Germany, West Germany, Berlin;

dummy-coded), and interview mode (paper-and-pencil ques-

tionnaire, CATI, online questionnaire; dummy-coded) were

included as covariates.

Statistical analyses

A mediation analysis was applied to explore possible mech-

anisms through which an exposure and an outcome might be

associated.18 Three paths are relevant for the investigation of

mediating effects: the direct, the indirect, and the total effect.

The model of this study suggests four direct effects (from low

and medium SES to analgesic and sedative/hypnotic use,

respectively), eight indirect effects (from low and medium SES

to analgesic and sedative/hypnotic use through physical and

mental health, respectively), and two total effects (sum of

direct and indirect effects, one for analgesic use and one for

sedative/hypnotic use) (Fig. 1).

Full mediation is indicated if the indirect effect is statis-

tically significant, whereas the direct effect is non-

significant. To quantify the strength of a full mediation,

the proportion of the total effect mediated by health is

indicated in the output. If both the indirect and the direct

effect are statistically significant, there is evidence for par-

tial mediation. The requirements for a mediation analysis
Fig. 1 e Pathway model for evaluation of mediating effects of s

between SES and weekly analgesic or sedative/hypnotics use. S

P ¼ physical health (2 groups; ref.: good physical health status)
status); A ¼ weekly analgesic use (2 groups; ref.: non-weekly a
ref.: non-weekly sedative/hypnotic use); ref. ¼ reference group.
are not fulfilled if an indirect and/or a total effect, or a single

path of an indirect path, is statistically non-significant.18

To evaluate mediating effects statistically, generalized

structural equation modeling (GSEM) was applied.19 Models

were fitted by using the maximum likelihood method,

assuming logit link functions and Bernoulli distribution for

binary outcomes. For reasons of clarity in a complex GSEM

and to enable comparability of estimates between pathways

and across the models, only binary variables were included in

the model, allowing a consistent reporting of odds ratios (ORs)

(with corresponding 95% confidence intervals).

Because numerous studies showed extensive gender dif-

ferences in the use of medicines and medical service in gen-

eral,20,21 analyses were stratified by gender. Data were

statistically weighted to account for the disproportionate

sampling and differences in sociodemographic characteristics

between the sample and the German adult population in 2012.

Statistical analyses were carried out with Stata 14 (Stata Corp

LP, College Station, TX).
Results

Sample description

Characteristics of the study sample are shown in Table 1. The

mean (standard deviation) age of the total sample was 42.3
elf-rated physical and mental health status on the relation

ES ¼ socio-economic status (3 groups; ref.: high SES);
; M ¼ mental health (2 groups; ref.: good mental health
nalgesic use); S ¼ weekly sedative/hypnotic use (2 groups;

https://doi.org/10.1016/j.puhe.2018.12.018

Table 1 e Characteristics of the study sample, 18-year-old
to 64-year-old men and women (total n ¼ 9084).

Male Female

n % n %

Sample size 3929 50.8 5155 49.2

Age (years)

18e24 987 12.6 1154 12.4

25e34 639 19.2 954 18.4

35e44 559 19.4 874 20.8

45e54 855 27.7 1075 26.2

55e64 886 21.2 1098 22.3

Socio-economic status (SES)

low 464 13.2 575 14.1

medium 2267 61.0 3074 61.4

high 1042 25.8 1299 24.5

Marital status (unmarried) 2104 47.4 2455 41.4

Physical health (ill) 962 29.0 1296 28.1

Mental health (ill) 1044 30.1 1612 33.3

Weekly analgesic use 530 16.3 961 20.7

Weekly sedative/hypnotic use 105 3.3 216 5.0

n ¼ number of cases; % ¼ percent; n is unweighted; % is weighted.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e94
(13.1) years; 50.8% were male. SES groups were distributed

similarly across gender; approximately 25% of male and fe-

male respondents were ascribed to the highest SES and

approximately 15% to the lowest. Ill physical health was
Fig. 2 e Single path correlations; 18-year-old to 64-year-old me

confidence intervals (CIs, in parentheses) are shown for significa

shown; measurement models for associated errors and covaria

mode) are estimated but not shown; SES ¼ socio-economic stat
ref.: good physical health status); M ¼ mental health (2 groups; r
groups; ref.: non-weekly analgesic use); S ¼ weekly sedative/h
use); ref. ¼ reference group; n.s. ¼ non-significant.
reported by 29.0% of male respondents and 28.1% of female

respondents. A slightly higher proportion reported an ill

mental health status (male: 30.1%; female: 33.3%). Analgesics

were used at least once a week in the past 30 days by 16.3% of

men and 20.7% of women; weekly sedative/hypnotic use was

reported by 3.3% of men and 5.0% of women.

Mediating effects

Fig. 2 shows single path correlations for male respondents.

The likelihood for exhibiting ill physical health was signifi-

cantly higher among male respondents of low SES as

compared with those of high SES (OR ¼ 2.23). A slightly lower
likelihood was found for the path from low SES to mental

health (OR ¼ 1.70). Ill physical health was associated with 3.04-
fold increased odds for analgesic use, whereas the odds were

increased by a factor of 1.7 regarding mental health. The

effect of SES on analgesic use was statistically non-significant,

that is, no direct effect was indicated. Table 2 suggests that

39.8% of the total effect on analgesic use could be explained by

the indirect effect of low SES through physical health,

whereas 12.6% was due to the indirect effect of low SES

through mental health. Concerning sedatives/hypnotics,

increased odds were found for men reporting ill physical

(OR ¼ 2.35) and mental (OR ¼ 10.75) health (Fig. 2). Full
n (n ¼ 3695). Odds ratios (OR) and corresponding 95%
nt paths only; effects for medium SES are estimated but not

tes (age, marital status, regional distribution, interview

us (3 groups; ref.: high SES); P ¼ physical health (2 groups;
ef.: good mental health status); A ¼ weekly analgesic use (2
ypnotic use (2 groups; ref.: non-weekly sedative/hypnotic

https://doi.org/10.1016/j.puhe.2018.12.018

Table 2 e Direct, indirect, and total effects of SES on analgesic and sedative/hypnotic use, 18-year-old to 64-year-old to men
and women (total n ¼ 9084).

Analgesics Sedatives/hypnotics

OR 95% CI Proportion of total
effect mediated

Male

Direct effect n.s. n.s.

Total effect 9.38 (3.70e23.79) 40.19 (5.81e277.81)

Indirect effect (low SES)

Physical health 2.44 (1.57e3.80) 39.8% 1.98 (1.11e3.55) 18.6%

Mental health 1.33 (1.06e1.65) 12.6% 3.52 (1.50e8.28) 34.1%

Female

Direct effect n.s. n.s.

Total effect 2.73 (1.35e5.55)

Indirect effect (low SES)

Physical health 2.09 (1.45e3.03) 73.5% 1.74 (1.22e2.48)

Mental health n.s. n.s.

OR ¼ odds ratio; 95% CI ¼ 95% confidence interval (a¼0.05); n.s. ¼ non-significant; SES ¼ socio-economic status
Three groups; reference group: high SES; controls: age, marital status, regional distribution, interview mode; measurement models for asso-

ciated errors and covariates are estimated but not shown; indirect effects for medium SES are estimated but are not shown.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e9 5
mediation on the effect of low SES on sedative/hypnotic use

was indicated for both mediators; 18.6% of the total effect was

attributable to physical health and 34.1% to mental health

(Table 2).
Fig. 3 e Single paths correlations, 18-year-old to 64-year-old wo

confidence intervals (CIs, in parentheses) are shown for significa

shown; measurement models for associated errors and covaria

mode) are estimated but not shown; SES ¼ socio-economic stat
ref.: good physical health status); M ¼ mental health (2 groups; r
groups; ref.: non-weekly analgesic use); S ¼ weekly sedative/hy
use); ref. ¼ reference group; n.s. ¼ non-significant.
Fig. 3 shows single-path correlations for women.

Compared with high SES, women of low SES had 1.75-fold

increased odds for an ill physical health status. In turn,

women with ill physical health exhibited 3.74-fold increased
men (n ¼ 4834). Odds ratios (OR) and corresponding 95%
nt paths only; effects for medium SES are estimated but not

tes (age, marital status, regional distribution, interview

us (3 groups; ref.: high SES); P ¼ physical health (2 groups;
ef.: good mental health status); A ¼ weekly analgesic use (2
pnotic use (2 groups; ref.: non-weekly sedative/hypnotic

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e96
odds for analgesic use. No direct effect of SES on analgesic use

was indicated. Full mediation on the effect of low SES on

analgesic use through physical health was indicated; 73.5% of

the total effect could be explained (Table 2). Ill mental health

was associated with increased odds for both analgesic

(OR ¼ 1.36) and sedative/hypnotic use (OR ¼ 4.85) (Fig. 3). An
indirect effect of low SES on the use of sedatives/hypnotics

through physical health was indicated (OR ¼ 1.74) (Table 2).
However, because SES and mental health were not correlated

(Fig. 3), mediating effects could not be assumed.
Discussion

The present study evaluated mediating effects on the

pathway from SES to weekly analgesic and sedative/hypnotic

use by self-rated physical and mental health status, respec-

tively. Among men, both physical and mental health medi-

ated the path from SES to analgesic and sedative/hypnotic

use, with a stronger effect of physical health on analgesic use

and mental health on sedative/hypnotic use. Inconsistent

results were found among women.

Among men, pathways from low SES to both analgesic and

sedative/hypnotic use were fully mediated by both physical

and mental health, respectively. This suggests that socio-

economic inequalities in medicine use are rather attribut-

able to health inequalities than to differences in SES.

Although strong associations between SES and health,11,12

and in turn, between health and the use of medicines have

repeatedly been shown,1,8,9,10 evidence for mediating effects

of health was lacking so far. Moreover, no direct effect was

observed between SES and medicine use, and high pro-

portions of the total effects can be explained by mediating

effects of health in the statistical models. In total (low plus

medium SES and physical plus mental health), mediating ef-

fects explained 76.6% of analgesic use and 74.5% of sedative/

hypnotic use. Health inequalities therewith seem to be the

major driving force for differences in medicine use by SES. No

third variables which might lead to SES differences in medi-

cine use, such as willingness to use medicines at a given

health status or impaired affordability of medicines,22 seem to

be involved to a substantial degree. Identification and elimi-

nation of the reasons for poor health among people of low SES

may, therefore, help to reduce health inequalities. One well-

known reason is an elevated likelihood of engaging in un-

healthy behavior such as smoking, episodic heavy drinking,

and diminished physical activity.11,12 Accordingly, efforts to

improve health among individuals of low SES should already

start at the prevention level and make sure that preventive

measures reach all socio-economic levels. Schools, work-

places, or neighborhoods are examples of places with a socio-

economic segregation and, therefore, potential outlets to

reach individuals of low SES.23

Similar conclusions can only partially be drawn for

women. Physical health fully mediated the association be-

tween low SES and analgesic use, with no statistically sig-

nificant direct effect from SES to analgesic use and a high

proportion of medicine use being mediated (73.5%). However,

mediating effects by mental health on both analgesic and

sedative/hypnotic use could not be observed. This result was
unexpected and is not in accordance with the literature. For

instance, based on a cross-sectional sample of the UK

household population, a strong relationship between educa-

tional level and prevalence of neurotic disorders was found

for men and women24 and similar findings were reported in

other European studies.25,26 Findings of the present study are

corroborated by one study only, revealing a lacking associa-

tion between family affluence score and girls’ self-rated
mental health.27 One explanation for this inconsistency

could be found in different measurements of mental health.

The latter and the present study used a self-rated assess-

ment, whereas the other studies used clinical diagnoses.

Assessments of health status (i.e., self-rated versus clinically

diagnosed) were found to vary substantially depending on the

type of measurement.28,29 Physicians may focus on objective

symptoms and diagnoses when assessing a patient’s health
status, whereas respondents may focus on subjective symp-

toms, functional limitations, and quality of life.29 Regarding

somatoform health complaints, a physician may attribute

them to mental concerns, whereas a patient would rather

rate them as physical complaints. Considering that women

tend to somatize more than men,30 self-rating may contribute

to an inflation of the number of women with ill physical

health, which might, in turn, lead to greater SES differences

in women’s physical health status compared with mental
health status. Besides, an indirect SES-effect on sedative/

hypnotic use through physical health was found among

women but no statistically significant total effect. Mediating

effects could not be deduced here. This implies that sedative/

hypnotic use is equally distributed across SES groups and that

females of the study sample were using sedatives/hypnotics

independently of their SES. Certainly, to properly evaluate

these findings and to draw sound conclusions, further in-

vestigations are needed.

It was further hypothesized that physical health is of

greater importance in the relationship between SES and

analgesic use (as compared with sedative/hypnotic use),

whereas mental health should accordingly be of greater

importance in the relationship between SES and sedative/

hypnotic use. The present study corroborates this hypothesis

for men. Nearly 40% of the effect of low SES on analgesic use

was mediated by physical health but less than 13% by mental

health only. Regarding sedative/hypnotic use, an inverse

pattern was found. Thus, it seems that men of the study

sample used analgesics or sedatives/hypnotics as per their

type of health complaints, regardless of the individual’s SES.
For women, no conclusions can be drawn because of a lacking

mediating effect of mental health on medicine use.

Methodological considerations

Educational level was used as it is known as a reliable proxy

for SES and has been revealed to be a good predictor of self-

reported health.31,32 Education is also supposed to be fairly

stable beyond early adulthood33 and has become the most

commonly used SES measure in epidemiological studies.31

It has been refrained from adjusting the GSEM for other

potential SES indicators such as income or occupational

prestige. The study’s aim was not to explore pathways from
education itself to medicine use but from socio-economic

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e9 7
status to medicine use. In other words, education was used as

one possible proxy for SES.34 The use of a three-level measure

of SES instead of two levels (low versus high) was chosen for

the analyses to prevent loss of information and deceptions of

group differences due to large group sizes. This approach

follows the majority of social epidemiological studies and

methodological recommendations.12,15

Owing to inconsistent and unexpected findings among

women in this study and a lack of comparable studies

focusing on this issue, further investigations are needed to

draw sound conclusions. Beyond the mediator analysis, future

research might also address moderating effects of SES on

health, that is, ill health might lead to different patterns of

medicine use, depending on SES.4 It has been shown that

healthy men of lower SES were more likely to use OTC drugs

than those of high SES, whereas, among those with ill health,

high SES individuals were more likely to use prescribed

medicines.35,36

Strength

A major strength of this study is the evaluation of mediating

effects by applying GSEM. The statistical power of GSEM is

remarkably higher than that of the standard regression

method, missing data are not handled by list-wise deletion,

and a simultaneous consideration of all indicators and path-

ways is possible through GSEM.18 Another plus is the high-

quality data source, characterized by a large sample size and

a standardized data collection.14

Limitations

It was not possible to distinguish between prescribed and non-

prescribed/OTC medicine use because of the lacking infor-

mation in the data. Several international studies indicated

that socio-economic inequalities in medicine use substan-

tially vary in this respect. In Austrian samples, individuals of

higher SES were more likely to use non-prescribed medicines,

whereas those of lower SES rather used prescribed medi-

cines.36,37 A Danish study showed a declining prevalence of

prescribed medicine use by SES, whereas no association was

found for OTC drugs.38

Conclusions about causal relationships are limited

because of the cross-sectional study design. Poor health

could also lead to a lower SES, which would change the

order of SES and health. However, a stronger causal relation

from SES to health, rather than vice versa, is supported by

the literature.11

Conclusions

As per the latest estimates for Germany, about 1.5e1.9 million

people are addicted to medicines.39 This study showed that

social inequalities in health seem to have substantial impact

on the prevalence of medicine use in the population. Identi-

fication and elimination of the reasons for poor health among

people of low SES may, therefore, not only help to reduce

health inequalities directly. A decline in the use of medicines

would also result in less side-effects and a reduced number
of people with medicine-related misuse and addiction.

Inconsistent findings among women, however, showed that

further investigations are needed to draw sound conclusions

and public health implications.
Author statements

Ethical approval

All procedures were in accordance with the ethical standards

of the institutional research committee and with the 1964

Helsinki declaration and its later amendments or comparable

ethical standards. Informed consent was obtained from all

individual participants included in the study. Ethics approval

for this research was obtained from the ethics committee

of the German Psychological Society (DGPs; Reg. no:

GBLK06102008DGPS).

Funding

Funding of the Epidemiological Survey of Substance Abuse

(ESA) was provided by the German Federal Ministry of Health

(BMG) (Grant No. IIA5-2511DSM216).

Competing interests

J.M., E.G.d.M., and O.P. declare that they have no conflict of

interest. L.K. and D.P. declare having received a grant from

Lundbeck GmbH for a project on alcohol epidemiology unre-

lated to this study.

Author contribution

L.K. and D.P. conceived, designed, and managed the study. J.M.

analyzed the data and drafted the manuscript. E.G.d.M., D.P.,

L.K., and O.P. participated in the interpretation of the results,

critically revised subsequent versions of the article, and

reviewed the article for writing and intellectual content. All

authors read, approved, and contributed to the final manu-

script. All co-authors have agreed to the submission of the

final manuscript.
r e f e r e n c e s

1. Knopf H, Melchert HU. Bundes-gesundheitssurvey:
arzneimittelgebrauch. Konsumverhalten in Deutschland. [Bundes-
Gesundheitssurvey: pharmaceutical use. Patterns of use in
Germany]. Berlin, Germany: Robert Koch-Institut; 2003.
ISBN: 3-89606-147-x.

2. Kantor ED, Rehm CD, Haas JS, Chan AT. Trends in
prescription drug use among adults in the United States from
1999-2012. J Am Med Assoc 2015;314(17):1818e31. https://
doi.org/10.1001/jama.2015.13766.

3. Henkel D, von Alkohol Zum Konsum. Tabak und
psychoaktiven Medikamenten bei Arbeitslosen und
Einkommensarmen. Eine Auswertung des Nationalen
Gesundheitssurveys 1991/1992 der Bundesrepublik
Deutschland. [Consumption of alcohol, tobacco, and

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref1

https://doi.org/10.1001/jama.2015.13766

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref3

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e98
psychoactive medication by the unemployed and those on
low income. An evaluation of the 1991/1992 national survey
of health in the Federal Republic of Germany]. Abh€angigkeiten
2000;6(1):26e43.

4. Holstein BE, Hansen EH, Due P. Social class variation in
medicine use among adolescents. Eur J Public Health
2004;14(1):49e52. https://doi.org/10.1093/eurpub/14.1.49.

5. Nielsen MW, Hansen EH, Rasmussen NK. Patterns of
psychotropic medicine use and related diseases across
educational groups: national cross-sectional survey. Eur J Clin
Pharmacol 2004;60(3):199e204. https://doi.org/10.1007/s00228-
004-0741-4.

6. Maron J, Kraus L, Pogarell O, Gomes de Matos E, Piontek D.
Occupational inequalities in psychoactive substance use: a
question of conceptualization? Addict Res Theor
2015;24(3):186e98. https://doi.org/10.3109/
16066359.2015.1093122.

7. Furu K, Straume B, Thelle DS. Legal drug use in a general
population: association with gender, morbidity, health care
utilization, and lifestyle characteristics. J Clin Epidemiol
1997;50:341e9. https://doi.org/10.1016/S0895-4356(96)00362-9.

8. Rosholm JI, Christensen K. Relationship between drug use
and self-reported health in elderly Danes. Eur J Clin Pharmacol
1997;53:179e83.

9. Bath PA. Self-rated health as a risk factor for prescribed drug
use and future health and social service use in older people. J
Gerontol Biol Med Sci 1999;54(11):565e70. https://doi.org/
10.1093/gerona/54.11.M565.

10. Holstein BE, Hansen EH, Andersen A, Due P. Self-rated health
as predictor of medicine use in adolescence. Pharmacoepidemiol
Drug Saf 2008;17(2):186e92. https://doi.org/10.1002/pds.1529.

11. Mielck A. Soziale Ungleichheit und Gesundheit: einführung in die
aktuelle Diskussion. [Social inequality and health: introduction into
the current debate]. Bern, Switzerland; G€ottingen, Germany;
Toronto, Canada; Seattle, United States: Hans Huber; 2005.
345684235X.

12. Mackenbach J. Health inequalities: Europe in profile. London,
United Kingdom: Department of Health; 2006.

13. Pabst A, Kraus L, Gomes de Matos E, Piontek D.
Substanzkonsum und substanzbezogene St€orungen in
Deutschland im Jahr 2012. [substance use and substance use
disorders in Germany in 2012]. Suchttherapie
2013;59(6):321e31. https://doi.org/10.1024/0939-5911.a000275.

14. Kraus L, Piontek D, Pabst A, Gomes de Matos E. Studiendesign
und methodik des epidemiologischen suchtsurveys 2012.
[study design and methodology of the 2012 epidemiological
survey of substance abuse]. Suchttherapie 2013;59(6):309e20.
https://doi.org/10.1024/0939-5911.a000274.

15. UNESCO Institute for Statistics. ISCED 2011 operational manual.
Guidelines for classifying national education programmes and
related qualifications. Secretary-General of the OECD; 2015. 978-
92-78-41239-5.

16. Manor O, Matthews S, Power C. Dichotomous or categorical
response? analysing self-rated health and lifetime social
class. Int J Epidemiol 2000;29(1):149e57. https://doi.org/10.1093/
ije/29.1.149.

17. Latham K, Peek CW. Self-rated health and morbidity onset
among late midlife U.S. adults. J Gerontol B Psychol Sci Soc Sci
2013;68(1):107e16. https://doi.org/10.1093/geronb/gbs104.

18. Gunzler D, Chen T, Wu P, Zhang H. Introduction to mediation
analysis with structural equation modeling. Shanghai Arch
Psychiatry 2013;25(6):390e4. https://doi.org/10.3969/
j.issn.1002-0829.2013.06.009.

19. Kupek E. Beyond logistic regression: structural equations
modelling for binary variables and its application to
investigating unobserved confounders. BMC Med Res Methodol
2006;6(13). https://doi.org/10.1186/1471-2288-6-13.
20. Dunnell K, Fitzpatrick J, Bunting J. Making use of official
statistics in research on gender and health status: recent
British data. Soc Sci Med 1999;48(1):117e27. https://doi.org/
10.1016/S0277-9536(98)00294-9.

21. Roe CM, McNamara AM, Motheral BR. Gender- and age-
related prescription drug use patterns. Ann Pharmacother
2002;36(1):30e9. https://doi.org/10.1345/aph.1A113.

22. Hong S, Cagle JG, Van Dussen DJ, Carrion IV, Culler KL.
Willingness to use pain medication to treat pain. Pain Med
2016;17:74e84. https://doi.org/10.1111/pme.12854.

23. Thrane C. Explaining educational-related inequalities in
health: mediation and moderator models. Soc Sci Med
2006;62(2):467e78. https://doi.org/10.1016/
j.socscimed.2005.06.010.

24. Lewis G, Bebbington P, Brugha T, Farrell M, Gill B, Jenkins R,
Meltzer H. Socio-economic status, standard of living, and
neurotic disorder. Lancet 1998;352(9128):605e9. https://
doi.org/10.1016/S0140-6736(98)04494-8.

25. Araya R, Lewis G, Rojas G, Fritsch R. Education and income:
which is more important for mental health? J Epidemiol
Community Health 2003;57(7):501e5. https://doi.org/10.1136/
jech.57.7.501.

26. Stewart AL, Dean ML, Gregorich SE, Brawarsky P, Haas JS.
Race/ethnicity, socio-economic status and the health of
pregnant women. J Health Psychol 2007;12(2):285e300. https://
doi.org/10.1177/1359105307074259.

27. Hutton K, Nyholm M, Nygren JM, Svedberg P. Self-rated
mental health and socio-economic background: a study of
adolescents in Sweden. BMC Public Health 2014;14:394. https://
doi.org/10.1186/1471-2458-14-394.

28. Geest TA, Engberg M, Lauritzen T. Discordance between self-
evaluated health and doctor-evaluated health in relation to
general health promotion. Scand J Prim Health Care
2004;22(3):146e51. https://doi.org/10.1080/
02813430410000941.

29. Giltay EJ, Vollaard AM, Kromhout D. Self-rated health and
physician-rated health as independent predictors of
mortality in elderly men. Age Ageing 2012;41(2):165e71.
https://doi.org/10.1093/ageing/afr161.

30. Wool CA, Barsky AJ. Do women somatize more than men?
Gender differences in somatization. Psychosomatics
1994;35(5):445e52. https://doi.org/10.1016/S0033-3182(94)
71738-2.

31. Winkleby MA, Jatulis DE, Frank E, Fortmann SP. Socio-
economic status and health: how education, income, and
occupation contribute to risk factors for cardiovascular
disease. Am J Public Health 1992;82(6):816e20.

32. Arber S. Comparing inequalities in women’s and men’s
health: Britain in the 1990’s. Soc Sci Med 1997;44:773e87.

33. Krieger N, Williams DR, Moss N. Measuring social class in U.S.
public health research: concepts, methodologies, and
guidelines. Annu Rev Public Health 1997;18:341e78. https://
doi.org/10.1146/annurev.publhealth.18.1.341.

34. Krokstad S, Kunst AE, Westin S. Trends in health inequalities
by educational level in a Norwegian total population study. J
Epidemiol Community Health 2002;56(5):375e80. https://doi.org/
10.1136/jech.56.5.375.

35. Daban F, Pasarı́n MI, Rodrı́guez-Sanz M, Garcı́a-Alt�es A,
Villalbı́ JR, Zara C, Borrell C. Social determinants of prescribed
and non-prescribed medicine use. Int J Equity Health 2010;9:12.
https://doi.org/10.1186/1475-9276-9-12.

36. Mayer S, €Osterle A. Socio-economic determinants of
prescribed and non-prescribed medicine consumption in
Austria. Eur J Public Health 2015;25(4):597e603. https://doi.org/
10.1093/eurpub/cku179.

37. Vogler S, €Osterle A, Mayer S. Inequalities in medicine use in
Central Eastern Europe: an empirical investigation of socio-

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref3

https://doi.org/10.1093/eurpub/14.1.49

https://doi.org/10.1007/s00228-004-0741-4

https://doi.org/10.3109/16066359.2015.1093122

https://doi.org/10.1016/S0895-4356(96)00362-9

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref8

https://doi.org/10.1093/gerona/54.11.M565

https://doi.org/10.1002/pds.1529

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref11

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref12

https://doi.org/10.1024/0939-5911.a000275

https://doi.org/10.1024/0939-5911.a000274

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref15

https://doi.org/10.1093/ije/29.1.149

https://doi.org/10.1093/geronb/gbs104

https://doi.org/10.3969/j.issn.1002-0829.2013.06.009

https://doi.org/10.1186/1471-2288-6-13

https://doi.org/10.1016/S0277-9536(98)00294-9

https://doi.org/10.1345/aph.1A113

https://doi.org/10.1111/pme.12854

https://doi.org/10.1016/j.socscimed.2005.06.010

https://doi.org/10.1016/S0140-6736(98)04494-8

https://doi.org/10.1136/jech.57.7.501

https://doi.org/10.1177/1359105307074259

https://doi.org/10.1186/1471-2458-14-394

https://doi.org/10.1080/02813430410000941

https://doi.org/10.1093/ageing/afr161

https://doi.org/10.1016/S0033-3182(94)71738-2

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref31

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref32

https://doi.org/10.1146/annurev.publhealth.18.1.341

https://doi.org/10.1136/jech.56.5.375

https://doi.org/10.1186/1475-9276-9-12

https://doi.org/10.1093/eurpub/cku179

https://doi.org/10.1016/j.puhe.2018.12.018

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 e9 9
economic determinants in eight countries. Int J Equity Health
2015;14:124. https://doi.org/10.1186/s12939-015-0261-0.

38. Nielsen MW, Hansen EH, Rasmussen NK. Prescription and
non-prescription medicine use in Denmark: association with
socio-economic position. Eur J Clin Pharmacol 2003;59:677e84.
https://doi.org/10.1007/s00228-003-0678-z.
39. Glaeske G. Medikamente 2016 – psychotrope und andere
Arzneimittel mit Missbrauchs- und Abh€angigkeitspotenzial.
[Medicines 2016 e psychotropic and other pharmaceuticals
with potential for misuse and dependence]. In: DHS jahrbuch
sucht 2018, editor. The German centre for addiction issues.
Lengerich: Pabst Science Publishers; 2018.

https://doi.org/10.1186/s12939-015-0261-0

https://doi.org/10.1007/s00228-003-0678-z

http://refhub.elsevier.com/S0033-3506(18)30410-4/sref39

https://doi.org/10.1016/j.puhe.2018.12.018

Exploring socio-economic inequalities in the use of medicines: is the relation mediated by health status?
Introduction
Methods
Study design and sample
Measures
Statistical analyses

Results
Sample description
Mediating effects

Discussion
Methodological considerations
Strength
Limitations
Conclusions

Author statements
Ethical approval
Funding
Competing interests
Author contribution

References

Economic-sanctions-threaten-population-health–the-case-of-_2019_Public-Heal

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 e1 3
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Short Communication
Economic sanctions threaten population health:
the case of Iran
M. Aloosh a,b,*, A. Salavati c, A. Aloosh d

a Department of Health Research Methods, Evidence, and Impact, Michael G. DeGroote School of Medicine,

McMaster University, Canada
b Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Canada
c Department of Pharmacology, Montreal University, Canada
d Department of Finance, NEOMA Business School, France
a r t i c l e i n f o

Article history:

Received 31 August 2018

Received in revised form

2 December 2018

Accepted 2 January 2019

Available online 14 February 2019

Keywords:

Economic sanction

Social determinants of health

Population health

Health care

Medication
* Corresponding author. Department of Heal
Master University, David Braley Health
Tel.: 9055259140×22356.

E-mail address: alooshm@mcmaster.ca (M
https://doi.org/10.1016/j.puhe.2019.01.006
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objective: The aim of the study is to review evidence of the negative consequences of in-

ternational economic sanctions on population health in Iran and pathways via which the

sanctions affect health.

Study design: This is a narrative review.

Methods: Data from the World Bank and the Central Bank of Iran were gathered to clarify

economic consequences of sanctions. Moreover, the literature was reviewed for published

data on health consequences of economic sanctions in Iran and economic crisis in other

parts of the world. Finally, some mechanisms via which economic sanctions could affect

health were reviewed.

Results: Iran experienced 11.8% reduction in gross domestic production growth in 2012

compared with 2011, besides 40% inflation and 200% depreciation of Iranian currency.

Ultimately, it resulted in increased living costs and unemployment. One year after

termination of sanctions, Iran’s gross domestic production growth increased by 14.1% in

2016. Data revealed that mental health has been affected during sanctions. Moreover,

access to essential and lifesaving medication has been compromised, similar to other

countries during economic recession.

Conclusion: Economic sanctions have had negative consequences on population health in

Iran by impairing social determinants of health and access to medication and care. These

sanctions widen economic inequality and health gap.

© 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.
th Research Methods, Evidence, and Impact, Michael G. DeGroote School of MedicinejMc-
Sciences Centre, Suite 2006, 100 Main St W, Hamilton, ON, L8P 1H6, Canada.

. Aloosh).

mailto:alooshm@mcmaster.ca

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.006&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.006

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 e1 3 11
Introduction

Economic sanctions are political decisions to subject a country

to economic pressure for political aims. These sanctions

impose a cost on a country’s economy; thus reducing the
growth of gross domestic production (GDP). Indeed, this

reduction indicates the intensity of the sanctions.1 The in-

tensity of economic sanctions against Iran is exceptional in

magnitude and significance in the history of international

sanctions. Some of the international sanctions related to

Iran’s nuclear program have been imposed and lifted together
in a relatively short period of time, 3 years. The US sanctions

have been reimposed again since November 5, 2018. Although

GDP growth and public health are driven by many factors,

including global factors, government policies, and de-

mographic changes, we believe that the sanctions signifi-

cantly affect health at individual and population levels by

impairing social determinants of health (SDH) and access to

medication and care in Iran.

The economic sanctions were instituted against Iran after

the 1979 revolution. These sanctions were strengthened dur-

ing the 2012e2015 period because of Iran’s nuclear program. In
fact, the reduction in Iran’s GDP demonstrated that these
sanctions were some of the most intense ones ever taken

against a nation. It restricted Iran’s export of oil and petro-
chemical products, which account for around 13% of Iran’s
GDP and around 80% of its total exports, as well as foreign

investment in Iran’s oil and energy industries, among others.
Moreover, an increasing number of restrictions such as

blocking the transfer of funds to and from Iran via the inter-

national banking system (SWIFT) have made international

trade with Iran risky if not impossible. In 2015, an interna-

tional deal lifted the sanctions for 3 years, until May 2018,

when the Trump administration abandoned the deal and

reinstated economic sanctions.

Starting in 2012, the sanctions caused 11.8% reduction in

GDP growth compared with 2011, 3.1% in 2011 and 7.7% in 2012

(Fig. 1), about 40% inflation, and more than 200% depreciation of

Iranian currency. Ultimately, it resulted in increased living

costs and unemployment. Interestingly, one year after the end
Fig. 1 e Composition of Iran’s GDP growth: data from the
Central Bank of Iran.
of sanctions, Iran’s GDP growth increased by 14.1% in 2016, from
�1.6% in 2015 to 12.5% in 2016 (Fig. 1). Clearly, these changes are
important at both individual and population levels.

Moreover, international findings show that public health

expenditure and healthcare services fall during economic re-

cessions,2 possibly because of the fall in GDP. It is important to

know that more than 40% of Iran’s 82 million population are
living below the poverty line, according to the Statistical

Center of Iran, whose health and well-being is extremely

vulnerable to economical stress. Moreover, Iranian healthcare

needs are mainly provided by the government. When a sig-

nificant portion of the Iranian government income is from oil

revenue, an economic sanction targeting oil export impairs

health services, significantly.
Evidence of the adverse health consequences of
economic sanctions

There is evidence to indicate that the economic sanctions

have had adverse effects on population health in Iran. Ac-

cording to the World Health Organization (WHO), death rates

due to self-harm rose from 5.9 to 6.1 per 100,000 persons

during the 2011e2014 period. Interestingly, this rate fell back

to 5.9 in 2016, one year after the sanctions were lifted. More-

over, deaths due to interpersonal violence rose from an

average of 2.0 to 2.7 per 100,000 persons during the period of

the sanctions.3 These trends could show deterioration in

mental health among Iranians during the period of intense

economic sanctions and an improvement afterward.

The adverse mental health effects of economic crises have

been reported previously in systematic reviews. Evidence

from the 2007 global recession showed a growth in patient

admissions in the healthcare sector, especially for mental

health issues, such as depression, substance-related disor-

ders, and suicide.4 These mental health consequences were

attributed mainly to unemployment and housing issues.

Moreover, studies showed that the economic crisis affected

population health by increasing the risk of cardiovascular and

respiratory diseases. Some key areas have been acknowledged

by the WHO to protect mental health during economic crises

which can be considered in Iran during sanctions: first, active

labor market programs; second, family support programs;

third, provision of quality and equitable access to primary care

and medications for vulnerable populations; and fourth, debt

relief programs.
Economic sanctions jeopardize SDH

There is strong evidence of the importance of SDH in defining

population and individual health.5 SDH include, amongst

other factors, income, social status, employment, social en-

vironments, and personal coping skills. In fact, SDH affect

people’s health and well-being in various ways, such as by
providing or inhibiting access to resources for individuals or

populations. Many of the determinants are compiled into the

healthcare access and quality index.6 Therefore, anything that

compromises SDH could potentially endanger population

health.

https://doi.org/10.1016/j.puhe.2019.01.006

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 e1 312
The World Bank data show how Iran’s economic growth
was affected during the sanctions. In fact, GDP per capita fell

from US$ 7833 in 2012 to US$ 4862 in 2015, 3 years after

intensification of the sanctions. However, 1 year after lifting

sanctions in 2016, the GDP recovered to US$ 5415.7 Further-

more, the sanctions caused an increase in unemployment

from 10.4% in 2013 to 13.1% in 2017, 2 years after the sanctions

were lifted, according to the World Bank. This negative

consequence extended beyond the period of sanctions, indi-

cating that complete recovery is a long process. Considering

the fact that unemployment is related to worse health out-

comes,8 the negative health consequence of higher unem-

ployment caused by economic sanctions is expected to be

long-lasting at the population level. However, it may take

some time before the full negative effects of higher unem-

ployment are shown at the population level. Moreover, the

Gini coefficient of household expenditure, which measures

economic inequality, has increased from 37% to 41%, since

2012, according to the Central Bank of Iran. And, there is

strong evidence that economic inequality widens the health

gap and compromises population health.

The economic sanctions imposed on Iran’s developing
economy had serious effects on SDH, such as income,

employment, and personal coping skills. Although a strong so-

cial safety net in some European countries protected society

from the negative health effects of the 2007 recession, such a

safety net does not exist in Iran. Moreover, the enduring polit-

ical conflict between Iran and several countries, including the

US, could intensify the health consequences of economic

pressure. The WHO Commission on SDH concluded that an

unequal global and national distribution of power, income,

goods, and services causes marked health inequities. Therefore,

these sanctions has increased the health gap both among Ira-

nians and most probably between Iranians and other nations.
The most vulnerable are the most affected

The adverse health impacts of economic instability are more

prevalent among vulnerable populations. For instance, studies

have shown that women’s mental health is more susceptible to
economic crises than that of men, and minorities are more

susceptible than the general population.9 Moreover, a recent

publication showed medication shortages during the sanctions

affected patients with chronic diseases, including multiple

sclerosis, cancer, hemophilia, and asthma, in Iran.10 Chroni-

cally ill patients are extremely vulnerable to economic crises.

They may not work and most probably depend on their family

and/or government to fulfill their basic needs and/or care. In

Iran, patients pay a portion of any treatment costs out of their

own pocket: about 10% for inpatient and 30% for outpatient

care. For the abovementioned chronic diseases, out-of-pocket

payments could be extremely high over time. In unfavorable

economic situations, alongside the government’s inability to
offer sufficient social and medical support, average household

income also decreases. Thus, millions of Iranians with low in-

come or below the poverty line are extremely vulnerable to the

adverse effects of economic sanctions.

Data from the last US recession support that many

chronically ill patients, particularly if they are unemployed or
disabled, reported greater medication cost problems.11 More-

over, data from Portugal show that about one-third of a group

of elderly patients stopped using treatment or health services

during the recession. Lower perceived health status and the

presence of three or more comorbidities were associated with

lower adherence to treatment.12 These indicate the health

risks of an economic crisis for vulnerable populations.
Sanctions impair access to care

Another pathway through which economic sanctions impair

population health is by impairing access to health care and

medication. During the sanctions, international institutions,

including pharmaceutical companies and banks, were

cautious on trading with Iran. One study identified a shortage

of 73 drugs, of which 44% were classified as essential medi-

cines13. Another report showed that the availability of 13 of 26

studied lifesaving medications fell significantly in Iran be-

tween 2012 and 2015. These include interferon a-2b, the

cornerstone of multiple sclerosis treatment. Moreover, con-

sumption of cytarabine, a common anticancer chemothera-

peutic agent, fell from 1.40 mg per 1000 people per day in 2010

to 0.96 in 2013.10 Other studies confirm that the sanctions

disrupted cancer care by restricting access to expensive

medication, and treatment, such as radiotherapy.14,15

In addition, patients with hemophilia and other coagu-

lation disorders need lifelong timely administration of factor

8, a blood-clotting factor, to prevent disability and death. The

mean per capita use of factor 8 in Iran before the sanctions

was 1.6 international units (IU). It fell to 0.5 IU during the

2013e2015 period, but increased to 2.7 IU in 2017, 2 years

after the economic sanctions were lifted.16 Moreover, the

most effective group of medication in asthma attack is se-

lective b2-adrenoreceptor agonists. Use of these medications

fell during the sanctions.10 Another study by Ghiasi et al.17

showed a 20e40% reduction in asthma-related drug avail-

ability in the Iranian market in the 2010e2013 period. A sig-

nificant proportion of asthma patients are children, whose

mortality and morbidity is higher than those of the general

population. Similarly, reports from Greece have described

limited access to multiple sclerosis and rheumatoid arthritis

treatment during recession.18 Moreover, during the

2000e2010 recession, chronically ill patients in Honduras

used fewer health services and medications because of cost

concerns.19
Conclusions

In conclusion, economic sanctions have adversely affected

population health in Iran, by impairing SDH, health care de-

livery, and access to care. In fact, despite the exemption of

health necessities and humanitarian goods from these sanc-

tions, health as a fundamental human right, among others,

has been compromised and most probably will be compro-

mised by the new wave of sanctions. Moreover, the adverse

health impacts will be greater among vulnerable populations,

including those on low-income, the chronically ill, children,

women, and the elderly.

https://doi.org/10.1016/j.puhe.2019.01.006

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 0 e1 3 13
Author statements

Acknowledgments

The authors would like to thank Mr. Mark Holdsworth for his

assistance in copyediting of the manuscript.

Ethical approval

Not required. This is a retrospective descriptive analysis of

publicly available data.

Funding

None declared.

Competing interests

None declared.
r e f e r e n c e s

1. Pape RA. Why economic sanctions do not work. Int Secur
1997;22(2):90e136.

2. Mucci N, Giorgi G, Roncaioli M, Fiz Perez J, Arcangeli G. The
correlation between stress and economic crisis: a systematic
review. Neuropsychiatric Dis Treat 2016;12:983e93.

3. Institute for Health Metrics and Evaluation. Global health data
exchange. Published Online by University of Washington; 2018.
Available at: https://vizhub.healthdata.org/sdg/ (last accessed
18 September 2018).

4. Frasquilho D, Matos MG, Salonna F, Guerreiro D, Storti CC,
Gaspar T, et al. Mental health outcomes in times of economic
recession: a systematic literature review. BMC Public Health
2016;16:115.

5. World health organization health impact analysis. The
determinants of health. Available at: http://www.who.int/hia/
evidence/doh/en/ (last accessed 06 September 2018).

6. Healthcare Access and Quality Index based on mortality from
causes amenable to personal health care in 195 countries and
territories, 1990-2015: a novel analysis from the Global Burden
of Disease Study 2015. Lancet (London, England)
2017;390(10091):231e66.
7. World bank report on gross domestic production per capita of
Iran. Available at: https://data.worldbank.org/indicator/NY.
GDP.PCAP.CD?locations¼IR (last accessed 07 September 2018).

8. Bockerman P, Ilmakunnas P. Unemployment and self-
assessed health: evidence from panel data. Health Econ
2009;18(2):161e79.

9. Glonti K, Gordeev VS, Goryakin Y, Reeves A, Stuckler D,
McKee M, et al. A systematic review on health resilience to
economic crises. PLoS One 2015;10. e0123117.

10. Kheirandish M, Varahrami V, Kebriaeezade A, Cheraghali AM.
Impact of economic sanctions on access to
noncommunicable diseases medicines in the Islamic
Republic of Iran. Eastern Mediterranean health journal ¼ La revue
de sante de la Mediterranee orientale ¼ al-Majallah al-sihhiyah li-
sharq al-mutawassit 2018;24(1):42e51.

11. Piette JD, Rosland AM, Silveira MJ, Hayward R, McHorney CA.
Medication cost problems among chronically ill adults in the
US: did the financial crisis make a bad situation even worse?
Patient Prefer Adherence 2011;5:187.

12. da Costa FA, Teixeira I, Duarte-Ramos F, Proenca L, Pedro AR,
Furtado C, et al. Effects of economic recession on elderly
patients’ perceptions of access to health care and medicines
in Portugal. Int J Clin Pharm 2017;39:104e12.

13. Setayesh S, Mackey TK. Addressing the impact of economic
sanctions on Iranian drug shortages in the joint
comprehensive plan of action: promoting access to medicines
and health diplomacy. Glob Health 2016;12(1):31.

14. Ameri A, Barzegartahamtan M, Ghavamnasiri M,
Mohammadpour R, Dehghan H, Sebzari A, et al. Current and
future challenges of radiation oncology in Iran: a report from the
Iranian society of clinical oncology. Clinical oncology (Royal
College of Radiologists (Great Britain)); 2018.

15. Aloosh M. How economic sanctions compromise cancer care
in Iran. Lancet Oncol 2018;19:e334.

16. Heidari R, Akbariqomi M, Tavoosidana G. Medical legacy of
sanctions in Iran. Nature 2017;552(7684):175.

17. Ghiasi G, Rashidian A, Kebriaeezadeh A, Salamzadeh J. The
impact of the sanctions made against Iran on availability to
asthma medicines in Tehran. Iran J Pharm Res (IJPR)
2016;15(3):567e71.

18. Souliotis K, Alexopoulou E, Papageorgiou M, Politi A, Litsa P,
Contiades X. Access to care for multiple sclerosis in times of
economic crisis in Greece–the HOPE II study. Int J Health Pol
Manag 2015;5:83e9.

19. Piette JD, Mendoza-Avelares MO, Chess L, Milton EC,
Reyes AM, Rodriguez-Salda~na J. Report on Honduras: ripples
in the ponddthe financial crisis and remittances to
chronically ill patients in Honduras. Int J Health Serv
2012;42:197e212.

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref1

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref2

https://vizhub.healthdata.org/sdg/

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref4

http://www.who.int/hia/evidence/doh/en/

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref6

https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?locations=IR

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref8

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref9

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref10

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref11

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref12

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref13

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref14

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref15

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref16

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref17

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref18

http://refhub.elsevier.com/S0033-3506(19)30006-X/sref19

https://doi.org/10.1016/j.puhe.2019.01.006

Economic sanctions threaten population health: the case of Iran
Introduction
Evidence of the adverse health consequences of economic sanctions
Economic sanctions jeopardize SDH
The most vulnerable are the most affected
Sanctions impair access to care
Conclusions
Author statements
Acknowledgments

Ethical approval
Funding
Competing interests
References

Assessing-the-potential-outcomes-of-achieving-the-World-Health-Org_2019_Publ

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Themed Papere Original Research
Assessing the potential outcomes of achieving the
World Health Organization global
non-communicable diseases targets for risk factors
by 2025: is there also an economic dividend?
M. Devaux a, A. Lerouge a, B. Ventelou b, Y. Goryakin a, A. Feigl a, S. Vuik a,
M. Cecchini a,*

a Health Division, Organization for Economic Co-operation and Development, Paris, France
b Aix-Marseille School of Economics, Aix-Marseille University, Marseille, France
a r t i c l e i n f o

Article history:

Received 11 April 2018

Received in revised form

20 October 2018

Accepted 4 February 2019

Available online 12 March 2019

Keywords:

Non-communicable diseases

Projection

Obesity

Smoking

Alcohol

Healthcare expenditure
* Corresponding author. OECD, ELS /HD, 2 ru
E-mail address: michele.cecchini@oecd.o

Objectives: This study assesses the change in premature mortality and in morbidity under

the scenario of meeting the World Health Organization (WHO) global targets for non-

communicable disease (NCD) risk factors (RFs) by 2025 in France. It also estimates medi-

cal expenditure savings because of the reduction of NCD burden.

Study design: A microsimulation model is used to predict the future health and economic

outcomes in France.

Methods: A ‘RF targets’ scenario, assuming the achievement of the six targets on RFs by

2025, is compared to a counterfactual scenario with respect to disability-adjusted life years

and healthcare costs differences.

Results: The achievement of the RFs targets by 2025 would save about 25,300 (and 75,500)

life years in good health in the population aged 25e64 (respectively 65þ) years on average
every year and would help to reduce healthcare costs by about V660 million on average per

year, which represents 0.35% of the current annual healthcare spending in France. Such a

reduction in RFs (net of the natural decreasing trend in mortality) would contribute to

achieving about half of the 2030 NCD premature mortality target in France.

Conclusions: The achievement of the RF targets would lead France to save life years and life

years in good health in both working-age and retired people and would modestly reduce

healthcare expenditures. To achieve RFs targets and to curb the growing burden of NCDs,

France has to strengthen existing and implement new policy interventions.

mailto:michele.cecchini@oecd.org

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.009&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.02.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9174
Introduction

In 2011, the UN General Assembly adopted a political decla-

ration that mobilized member countries for the reduction and

control of non-communicable diseases (NCDs).1 In particular,

the resolution includes an ultimate sustainable development

goal (SDG) target to reduce by one-third premature mortality

from NCDs by 2030. To achieve this, countries agreed on nine

voluntary global targets for 2025 (with a baseline of 2010),

including a target to reduce overall mortality from the four

main NCDs (cardiovascular diseases, cancer, diabetes, or

chronic respiratory disease) by 25% (called the 25 � 25 target),
six key risk factors (RFs) targets, and two national systems

response targets (Table 1).2

The case of France is particularly interesting as there is

conflicting evidence on whether France is on track to meet the

2030 SDG and the 25 � 25 mortality targets. Some recent
modeling studies suggest that if the six RF targets are met,

Franceeas well as other high-income countriesdmay achieve

to reduce NCDs premature mortality by 2025.3,4 However,

historical data suggest that premature mortality caused by the

four main NCDs in France has declined by 20% between 2000

and 2012 and that at this pace, France would not be able to

achieve the SDG target by 2030.5 This is echoed in the Global

Burden of Disease (GBD) 2018 study, which estimated that

France would not meet the target of one-third premature

mortality reduction by 2030.6

From a public health perspective, it is important to know

how much achieving the RF targets by 2025 would contribute

to the reduction in premature mortality and the achievement

of the 2030 target. From a policy-decision perspective, it is

important to understand whether meeting these targets

would translate into health cost savings, buttressing the

economic case for sustaining a large and ambitious public

health action aiming to protect people against these risks.

Previous literature on the achievement of the 2030 SDG and

the 25 � 25 targets has mainly focused on changes in pre-
mature mortality, with only one study focusing on morbidity

and none on economic outcomes. While few studies conclude

that the 2030 premature mortality target was achievable in

high-income countries that have invested in prevention and

treatment,7e9 the GBD 2018 study estimates that the 2030

target is to be achieved for men in only 16% of countries

worldwide and for women in 19% of countries.6 Generally,
Table 1 e Nine global targets for the prevention and control of

1. A 25% relative reduction in the overall mortality from cardiovascular

2. At least 10% relative reduction in the harmful use of alcohol, as appro

3. A 10% relative reduction in prevalence of insufficient physical activity

4. A 30% relative reduction in mean population intake of salt/sodium

5. A 30% relative reduction in prevalence of current tobacco use

6. A 25% relative reduction in the prevalence of raised blood pressure or c

circumstances

7. Halt the rise in diabetes and obesity

8. At least 50% of eligible people receive drug therapy and counseling (in

9. An 80% availability of the affordable basic technologies and essential

public and private facilities.

NCD, non-communicable disease
these findings suggest that more efforts are necessary to

achieve the RFs reduction. In the United Kingdom, achieving

all RFs targets was forecast to avert 300,000 deaths and 1.3

million years lived with disabilities from NCDs for the period

2010e2025, with health gains resulting mainly from reduced

mortality and morbidity from heart disease and stroke and

reduced morbidity from diabetes, depression and dementia.9

This article assesses the health and economic outcomes

under a scenario in which France meets the six NCDs global

targets on RFs (i.e. targets 2 to 7) by 2025. More specifically, this

analysis presents the impacts on morbidity, mortality, and

healthcare expenditure. This study does not cover targets 8

and 9 on access to care, as the French health system offers

universal health coverage with adequate healthcare provision

and availability of treatment and counseling to prevent and

control NCDs.10,11 Target 1, on mortality reduction, is assessed

through the effect mediated by a reduction in the prevalence

of the RFs.
Methods

General framework

The Organisation for Economic Co-operation and Develop-

ment (OECD) Strategic Public Health Planning for NCDs

(SPHeP-NCD) model is used to predict the health and eco-

nomic outcomes of the French population between 2016 and

2030, relying on previous microsimulation and econometric

approaches.12,13 The details of the SPHeP-NCD model are

described in the Web Appendix and in a related working

paper.13

Structure of the model and data sources

The SPHeP-NCD model is a microsimulation platform simu-

lating individual lives of a group of people, representative of a

country, from birth to death, including events such as

behavioral and physiological RFs and incidence of chronic

diseases, remission, and fatality from these diseases. The

relevant outputs of the model contain disease prevalence and

incidence, death cases, life years and life years lived in good

health, and medical costs of treatment of the diseases.

The model is built on four major modules: demography,

diseases, RFs, and medical costs. The three first modules
NCDs by 2025.

diseases, cancer, diabetes, or chronic respiratory disease

priate, within the national context

ontain the prevalence of raised blood pressure, according to national

cluding glycemic control) to prevent heart attacks and strokes

medicines, including generics, required to treat major NCDs in both

https://doi.org/10.1016/j.puhe.2019.02.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9 175
reproduce the demographic, epidemiological, and RF charac-

teristics by age- and gender-specific population groups of a

given country at different points in time. The demography

module simulates births, deaths, and inward/outward

migration following the United Nation population pro-

jections14 and the Human Mortality database.15 Epidemiolog-

ical characteristics of the model include disease incidence,

prevalence, remission, and fatality using data from the Insti-

tute for Health Metrics and Evaluation (IHME) GBD 201616 for

eight major NCDs (listed in the Web Appendix). In addition,

the model accounts for all the other causes of deaths through

a residual mortality rate. The RFs module simulates key

behavioral and physiological RFs. Definitions and model as-

sumptions for these RFs are described in the Web Appendix.

The input data for population exposure to RFs come from

IHME GBD 201616 for smoking and physical inactivity, WHO

2017 and IHME GBD 2015 for alcohol consumption,17,18 and

NCD-RisC 2017 for obesity and blood pressure.19 Relative risks

(RRs) that link RFs to diseases are determined by gender and

age group. Information on RRs was collected from IHME GBD

2016,20 the dynamic modeling for health impact assessment

(DYNAMO-HIA) model,21 and the OECD alcohol model.12

Further detailed features of the model (e.g. on disability

weights used for estimating life years lived in good health) can

be found in Cecchini et al.13

Medical costs of disease treatment are derived from na-

tional health expenditure data in France (including ambulatory

care, hospital, and pharmaceutical costs) and are replicated in

the future. The estimation was carried out by Cortaredona and

Ventelou.22 Disease-related costs are expressed in constant

2014 prices (Euros) and were calculated for the following dis-

eases: chronic obstructive pulmonary disease, dementia,

depression, ischemic and hemorrhagic strokes, myocardial

infarction, cancers, chronic kidney disease, alcohol-related

injuries, diabetes, and cirrhosis. Individual healthcare access

and consumption are considered constant over time (for a

given age, gender, and diseases profile). The methodology al-

lows differentiation between average residual costs, marginal

disease-related costs (with and without comorbidities), death-

related costs, and cost of comorbidities. The average residual

cost is an average annual per capita cost for people who not do

have one of the diseases listed above.

The ‘RF targets’ and counterfactual scenarios

A ‘RF targets’ scenario is designed to reflect changes attrib-

utable to the achievement of the six RF targets by 2025 (Table

1). The salt target is combined with the high blood pressure

(HBP) target assuming that the effect of salt intake on NCDs

prevalence and mortality is entirely mediated by the HBP

exposure. Similarly, the achievement of halting the rise in

diabetes is assumed to be fully driven by the achievement of

the obesity target.

The ‘RF targets’ scenario is compared to a counterfactual

scenario to estimate the pure effect of achieving the RF targets

by 2025 on health and economic outcomes (e.g. life years in

good health and healthcare costs), independently of the nat-

ural evolution of mortality. The counterfactual scenario as-

sumes that age- and gender-specific mortality rates,

prevalence of RFs, and RRs are constant as of 2016.
The ‘RF targets’ scenario uses historical data between 2010

and 2015 (or the most recent available year) and assumes a

linear reduction in the age- and gender-specific prevalence of

the RFs until reaching targets by 2025. Age- and gender-

specific prevalence rates for the RFs are then kept constant

between 2025 and 2030. To discard the natural decreasing

trend in mortality (which contains current societal trends and

trends in RFs) and to isolate the net effect of the reduction of

RFs, age- and gender-specific mortality rates are kept constant

as of 2016 and until 2030 in both scenarios.

The comparison of the model predictions under the two

scenarios allows quantifying the effect of the reduction of RFs

on health and economic outcomesdnet of the decreasing

mortality trenddand it sheds light on how much the reduc-

tion in RFs can contribute to reducing premature mortality in

France.
Results

Mortality

The probability of dying prematurely (between ages 30 and 70

years) from one of the four main NCDs diminishes with the

reduction of RFs by 2025 in a scenario where age- and gender-

specific mortality rates are set constant as of 2016 (Fig. 1).

Specifically, the ‘RF targets’ scenario reduces premature

mortality by 18% in men and by 15% in women from 2010 to

2025 (net of the natural diminishing trend in mortality); this

would represent by 2030 about half of the one-third reduction

target. At the same time, the UN projections of life expect-

ancydwhich reflect the natural decreasing trend in mortality

in the model-result in an yearly reduction in premature

mortality of 1.48% in the period 2010e2025, which is likely to

reach a 26% reduction by 2030. In other words, the mortality

reduction produced by the achievement of the ‘RF targets’,

combined with the natural decreasing trend in mortality for

other reasons, would put France be on track to meet the 2030

SDG targets on premature mortality.

Morbidity

The achievement of the RF targets by 2025 would also lead to

an improvement in health outcomes. By meeting the six RF

targets, about 10,100 life years and 25,300 life years in good

health could be saved annually in the French adult population

aged 25e64 years between 2016 and 2030 (Fig. 2). In the pop-

ulation aged 65þ years, about 60,500 life years and 75,500 life
years in good health can be saved annually.

Expenditures

A reduction in NCD burden is associated with lowered

healthcare costs. If the RF targets were met by 2025, annual

healthcare costs would be reduced by about V220 million

(population 25e64 years) and V440 million in people aged 65

years and abovedthat is a total of V660 million on average per

year (Fig. 3), representing 0.35% of total health expenditure in

France. Additional results of health expenditure savings over

time are available in the Web Appendix.

https://doi.org/10.1016/j.puhe.2019.02.009

10%

2010 2015 2020 2025 2030

P
ro

b
ab

ili
ty

o
f p

re
m

at
u

re
m

o
rt

al
it

Female SDG Target Male SDG Target

Fig. 1 e Probability of premature mortality from the four main NCDs in France, net of the natural diminishing trend in

mortality. Note: Includes 95% confidence intervals of the repeated simulations. Source: OECD SPHeP-NCD model, 2018.

SPHeP-NCD, Strategic Public Health Planning for NCD; NCD, non-communicable disease; Organisation for Economic Co-

operation and Development (OECD).

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9176
A sensitivity analysis was carried out to test alternative

assumptions on harmful alcohol use. For instance, the alcohol

target was amended to include an additional 10% reduction in

the age- and gender-specific prevalence of binge drinking.

This sensitivity analysis showed that a further restriction on

binge drinking did not strongly affect the overall results.
Discussion

This article shows notable health and economic impacts of

achieving the WHO global targets for RFs by 2025 in France.

The sole reduction of RFs (net of the decreasing trend in
0

20000

40000

60000

80000

Age 25-64

Li
fe

y
e

ar
s

sa
ve

Life years Life ye

Fig. 2 e Life years and life years in good health saved in France, a

confidence intervals of the repeated simulations. Source: OECD S

Planning for NCD; NCD, non-communicable disease; Organisatio
mortality) would lead to sizeable health benefits improving

life years and life years in good health in both working-age

and retired people and would reduce healthcare costs by

about V660 million annually which represent about 0.35% of

the total health spending in France. Findings also suggest that

if the natural decreasing trend in mortality was combined

with the mortality reduction stemming from the reduction of

the RFs, France would be on track to meet the 2030 SDG pre-

mature mortality target.

Expected healthcare savings are limited, but this is the

consequence of prevention policies: people who survive as a

result of avoiding NCDs will continue utilizing healthcare

services and therefore will incur health expenditure.
Age 65+

ars in good health

verage per year 2016e2030, by age group. Note: Includes 95%

PHeP-NCD model, 2018. SPHeP-NCD, Strategic Public Health

n for Economic Co-operation and Development (OECD).

https://doi.org/10.1016/j.puhe.2019.02.009

Fig. 3 e Reduction in healthcare costs, average per year 2016e2030 in France, by age group in years. Source: OECD SPHeP-

NCD model, 2018. Note: Includes 95% confidence intervals of the repeated simulations. SPHeP-NCD, Strategic Public Health

Planning for NCD; NCD, non-communicable disease; Organisation for Economic Co-operation and Development (OECD).

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9 177
However, the success of prevention should not just be

measured in terms of healthcare savings. Health gains are

valuable in themselves; furthermore, social costs and indirect

costs associated with NCDs and their RFs (e.g. employment

prospect, absenteeism from work) are not taken into account

while these costs represent about 1% of gross domestic

product (GDP) in France.23

Results of this study not only present new estimates on the

concrete health and economic impact of the SDG agenda in

France but also represent a significant step forward to develop

systems modeling tools for public health policy and plan-

ning.24 SHPeP-NCD stands out against existing, similar models

in the following novel ways.

First, SHPeP-NCD models real-life counterfactuals, for

example, in the absence of exposure to smoking, the same

individual might still be afflicted with heart disease due to

other RFs. This lowers the impact of RF-specific policy sce-

narios compared with traditional approaches that, instead,

usually assume that, in the absence of the RF, individuals

would live the rest of their life in good health. However, by

relaxing this assumption, the SHPeP-NCD model represents a

much closer-to-reality scenario which, then, translates into

more realistic economic estimates. This mechanism largely

explains why the cost impact of the 25 � 25 strategy is relatively
modest: people whose morbidity and mortality is reduced

continue to consume health care because of other diseases.

Second, SHPeP-NCD, in novel fashion, distinguishes be-

tween costs at different stages of disease. For example, the

extra cost of disease in the last year of life in a cancer patient

(which is higher than when in remission) is taken into ac-

count. This approach allows to address one of the most

pertinent questions in this literature: does prolonging life lead

to overall costs savings or increased costs?25,26 The model can

then better inform on what happens to national healthcare

costs when prolonging life does not result in compression of

morbidity, but rather, the extension of years lived with dis-

ease in the population.

Lastly, SHPeP-NCD further contributes to the field of

modeling the chronic disease cost burden by incorporating the

cost of comorbidities. Only a limited amount of high quality

results exist in this area and for only a subset of conditions.22
Therefore, by modeling the cost of comorbidities and ac-

counting for the extra cost of comorbidities,22 SHPeP-NCD

provides results that not only improve on existing models

available to date but provides results with immediate policy

relevance. More specifically, our results highlight the impor-

tance of targeting no single diseases, but also overall

comorbidities.

Policy context and implications

Obesity, physical inactivity, HBP, smoking, and alcohol misuse

are public health concerns that are being addressed with

comprehensive national prevention policy programs in

France. For instance, the French national nutrition and health

programs initially implemented in 2001 and revised since

(Programme National Nutrition Sant�e 2017-21) offers dietary

guidelines and physical activity recommendations for both

general and specific population groups. The French strategy

against obesity also includes restrictions on food advertising

to children on TV and radio, and more recently, the Nutri-

Score, a voluntary front-of-pack food labeling scheme that

aims to help consumers to make healthier choices. Regarding

smoking control, a national program against tobacco addic-

tion for 2014e2019 has introduced plain tobacco packaging,

smoking bans in public places and children’s areas, and
taxation to increase the price to 10 euros per pack by 2020.

Regarding alcohol policy, France limits the age of consump-

tion, restricts locations and hours for selling alcohol, and

forbids drink-driving. While regulations of alcohol advertising

were initially introduced in 1991 by the Loi Evin, they have

been modified over recent years.27 Taxation on alcohol is

uneven across types of beverages, in particular with low taxes

on wine compared with other European countries.28 While the

2025 RF targets are not yet achieved in France, continuous

efforts for more actions and comprehensive strategies are

required to curb the growing burden of NCDs.

Limitations

At least two limitations can be discussed. First, the model

accounts for about 56% of the deaths attributable to the four

https://doi.org/10.1016/j.puhe.2019.02.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9178
major NCDs (model result not displayed here), mainly because

the disease module does not cover all types of cancers and

CVDs. This means that premature mortality due to the four

major NCDs is estimated by the model to be around 8.7% in

men and 4.2% in women, while WHO data report 14.6% and

7.3%, respectively.29 Second, model-based results rely on

input data. In some cases, different sources may suggest

slightly different trends. For instance, IHME data suggest a

slow increase in obesity rates between 2010 and 2016, whereas

other sources conclude that obesity has stagnated in France

since 2006.30 However, such assumptions on the historical

data period do not greatly affect the results, since the ‘RFs

target’ scenario predicts a halt in obesity. Further in-

vestigations using alternative assumptions on the trends of

RFs could be tested in future model simulations.
Author statements

Acknowledgments

The views expressed in this article are those of the authors

and do not necessarily reflect those of the OECD or its member

countries.

The SPHeP-NCD microsimulation model was developed at

the OECD, building on previous OECD modeling work as well

as on collaborative work carried out as part of the H2020

FRESHER project. Authors would like to acknowledge Thierry

Pellegrini’s contribution to the development of the software.

Ethical approval

Not required. The analyses carried in this article used sec-

ondary data collected at national and international levels.

Funding

The OECD program of work on public health is supported by a

number of voluntary contributions by Ministries of Health or

other national governmental institutions of OECD Member

countries or key partners, including grants from the European

Commission (DG Sant�e). The funders had no role in study

design, data collection and analysis, decision to publish, or

preparation of the manuscript. The SPHeP-NCD micro-

simulation model was developed at the OECD, building on

previous OECD modelling work as well as on collaborative

work carried out as part of the H2020 FRESHER project.

Competing interests

None declared.
r e f e r e n c e s

1. United Nations and General Assembly. Political declaration of
the high-level meeting of the general assembly on the prevention
and control of non-communicable diseases. 24 January 2012.
Available from: http://www.who.int/nmh/events/un_ncd_
summit2011/political_declaration_en . [Accessed 6 April
2018].

2. World Health Organisation. Sixty-sixth world health assembly.
Follow-up to the political declaration of the high-level meeting of the
general assembly on the prevention and control of non-
communicable diseases. 27 May 2013. Available from: http://
apps.who.int/gb/ebwha/pdf_files/WHA66/A66_R10-en ?
ua¼1. [Accessed 6 April 2018].

3. World Health Organization. In: WHO, editor. Noncommunicable
diseases progress monitor; 2017.

4. Kontis V, Mathers C, Rehm J, Stevens G, Shield KD, Bonita R,
et al. Contribution of six risk factors to achieving the 25�25
non-communicable disease mortality reduction target: a
modelling study. Lancet 2014;384:427e37.

5. Hege E, Vaill�e J, Demailly D, Brimont L. La France passera-t-elle
le test des Objectifs du d�eveloppement durable (ODD) ? Une
�evaluation des nouveaut�es et des d�efis des ODD pour la France.
French: IDDRI.; 2016.

6. NCD Countdown 2030 collaborators, Bennett J, Stevens G,
Mathers C, Bonita R, Rehm J, Kruk M, et al. NCD Countdown
2030: worldwide trends in non-communicable disease
mortality and progress towards Sustainable Development
Goal target 3.4. Lancet 2018;392:1072e88.

7. Kontis V, Mathers C, Bonita R, Stevens G, Rehm J, Shield K,
et al. Regional contributions of six preventable risk factors to
achieving the 25×25 non-communicable disease mortality
reduction target: a modelling study. Lancet Glob Health
2015;3:e746e57.

8. Li Y, Zeng X, Liu J, Liu Y, Liu S, Yin P, et al. Can China achieve a
one-third reduction in premature mortality from non-
communicable diseases by 2030? BMC Med 2017;15(1):132.

9. Cobiac L, Scarborough P. Translating the WHO 25�25 goals
into a UK context: the PROMISE modelling study. BMJ Open
2017;7:e012805.

10. World Health Organization. Chapter 6 – tackling NCDs: the
capacity of countries to respond. In: Global status report on
noncommunicable diseases. Geneva: World Health Organization;
2010. p. 72e83. Reprinted 2011.

11. World Health Organization. Assessing national capacity for the
prevention and control of noncommunicable diseases: report of the
2015 global survey. 2016.

12. Cecchini M, Devaux M, Sassi F. Assessing the impacts of alcohol
policies: a microsimulation approach. Paris: OECD Health
Working Paper No. 80; 2015.

13. Cecchini M, Cortaredona S, Devaux M, Elfakir A, Goryakin Y,
Lerouge A, et al. Scientific paper on the methodology, results and
recommendation for future research, FRESHER report Deliverable
5.3. 2017 Dec. https://www.foresight-fresher.eu/content/
uploads/2018/03/d5-3-scientific-paper-on-the-methodology-
results-and-recommendation-for-future-research .

14. United Nations. World population prospects – population division –
United Nations. Available from: https://esa.un.org/unpd/wpp/.
[Accessed January 2018].

15. Human Mortality Database. Available from: http://www.mortality.
org/. [Accessed 29 March 2016].

16. GBD 2016 Disease and Injury Incidence and Prevalence
Collaborators, Vos T, Abajobir A, Abate K, Abbafati C,
Abbas K, Abd-Allah F, et al. Global, regional, and national
incidence, prevalence, and years lived with disability for 328
diseases and injuries for 195 countries, 1990-2016: a
systematic analysis for the Global Burden of Disease Study
2016. Lancet 2017;390(10100):1211e59.

17. World Health Organization, Global Health Observatory Data
Repository. Average daily intake in grams of alcohol, by country.
Available from: http://apps.who.int/gho/data/node.main.A1037?
lang¼en. [Accessed 21 December 2017].

18. GBD 2015 Risk Factors Collaborators, Forouzanfar M,
Afshin A, Alexander L, Anderson H, Bhutta Z, Biryukov S,

http://www.who.int/nmh/events/un_ncd_summit2011/political_declaration_en

http://apps.who.int/gb/ebwha/pdf_files/WHA66/A66_R10-en ?ua=1

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref3

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref4

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref5

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref6

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref7

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref8

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref9

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref10

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref11

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref12

https://www.foresight-fresher.eu/content/uploads/2018/03/d5-3-scientific-paper-on-the-methodology-results-and-recommendation-for-future-research

https://esa.un.org/unpd/wpp/

http://www.mortality.org/

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref16

http://apps.who.int/gho/data/node.main.A1037?lang=en

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref18

https://doi.org/10.1016/j.puhe.2019.02.009

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 7 3 e1 7 9 179
et al. Global, regional, and national comparative risk
assessment of 79 behavioural, environmental and
occupational, and metabolic risks or clusters of risks,
1990e2015: a systematic analysis for the Global Burden of
Disease Study 2015. Lancet 2016;388(10053):1659e724.

19. NCD Risk Factor Collaboration (NCD-RisC). Worldwide trends
in body-mass index, underweight, overweight, and obesity
from 1975 to 2016: a pooled analysis of 2416 population-based
measurement studies in 128$9 million children, adolescents,
and adults. Lancet 2017;390(10113):2627e42.

20. Gakidou E, Afshin A, Abajobir A, Abate K, Abbafati C, Abbas K,
et al. Global, regional, and national comparative risk
assessment of 84 behavioural, environmental and
occupational, and metabolic risks or clusters of risks,
1990e2016: a systematic analysis for the Global Burden of
Disease Study 2016. Lancet 2017;390(10100):1345e422.

21. Lhachimi S, Nusselder W, Smit H, van Baal P, Baili P,
Bennett K, et al. Dynamo-HIA-a dynamic modeling tool for
generic health impact assessments. PLoS One
2012;7(5):e33317.

22. Cortaredona S, Ventelou B. The extra cost of comorbidity:
multiple illnesses and the economic burden of non-
communicable diseases. BMC Med 2017;15(1):216.

23. Minist�ere de l’�Economie et des Finances. Ob�esit�e : quelles
cons�equences pour l’�economie et comment les limiter ?. 2016. http://
www.tresor.economie.gouv.fr/tresor-eco.

24. Atkinson J, Page A, Prodan A, McDonnell G, Osgood N.
Systems modelling tools to support policy and planning.
Lancet 24 March 2018;391(10129):1158e9.
25. Nistic�o F, De Alfieri W, Troiano G, Nante N, Dei S, Piacentini P.
Age-specific patterns of health care expenditure in dying
people. Public Health 2017;152:17e9.

26. Aldridge M, Kelley A. The myth regarding the high cost of
end-of-life care. Am J Public Health 2015;105(12):2411e5.

27. OECD. Tackling harmful alcohol use: economics and public health
policy. Paris: OECD Publishing; 2015.

28. OECD/European Observatory on Health Systems and Policies.
France: country health profile 2017, state of health in the EU.
Brussels: OECD Publishing, Paris/European Observatory on
Health Systems and Policies; 2017.

29. World Health Organization. Global Health Observatory Data
Repository. Risk of premature death from the four target NCDs.
Data by country. Available from: http://apps.who.int/gho/data/
node.main.A857?lang¼en. [Accessed 27 March 2018].

30. Verdot C, Torres M, Salanave B, Deschamps V. Corpulence des
enfants et des adultes en France m�etropolitaine en 2015. R�esultats
de l’�etude Esteban et �evolution depuis 2006, 13. BEH; 2017.
p. 234e41. French.
Appendix A. Supplementary data

Supplementary data to this article can be found online at

https://doi.org/10.1016/j.puhe.2019.02.009.

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref18

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref19

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref20

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref21

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref22

http://www.tresor.economie.gouv.fr/tresor-eco

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref24

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref25

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref26

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref27

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref28

http://apps.who.int/gho/data/node.main.A857?lang=en

http://refhub.elsevier.com/S0033-3506(19)30033-2/sref30

https://doi.org/10.1016/j.puhe.2019.02.009

Assessing the potential outcomes of achieving the World Health Organization global non-communicable diseases targets for ri …
Introduction
Methods
General framework
Structure of the model and data sources
The ‘RF targets’ and counterfactual scenarios

Results
Mortality
Morbidity
Expenditures

Discussion
Policy context and implications
Limitations

Author statements
Acknowledgments

Ethical approval
Funding
Competing interests
References
Appendix A. Supplementary data

Spatial-access-to-health-care-and-elderly-ambulatory-care-sen_2019_Public-He

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 3
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Spatial access to health care and elderly
ambulatory care sensitive hospitalizations
Y. Huang a,*, P. Meyer b, L. Jin c

a Department of Computing Sciences, Texas A&M University e Corpus Christi, Corpus Christi, TX, USA
b Department of Psychology and Sociology, Texas A&M University e Corpus Christi, Corpus Christi, TX, USA
c Department of Mathematics and Statistics, Texas A&M University e Corpus Christi, Corpus Christi, TX, USA
a r t i c l e i n f o

Article history:

Received 1 December 2017

Received in revised form

17 December 2018

Accepted 2 January 2019

Available online 28 February 2019

Keywords:

Ambulatory care sensitive hospital-

izations

Spatial accessibility

Primary care

Hospitals

Inpatient hospital admissions

Emergency department visits

Older adults
* Corresponding author. Department of Com
TX, USA. Tel.: þ1 361 825 2646; fax: þ1 361 8

E-mail address: lucy.huang@tamucc.edu
https://doi.org/10.1016/j.puhe.2019.01.005
0033-3506/Published by Elsevier Ltd on beha
a b s t r a c t

Objectives: Ambulatory care sensitive condition (ACSC) admission rates have been widely

used as indicators of access to and quality of primary care as well as the efficiency of health

systems. This study examines associations of spatial access to health care with both

inpatient hospital admissions and emergency department (ED) visits for ACSCs for older

adults. This study also compares inpatient hospitalization admissions and ED visits for

elderly ACSCs by spatial access to health care.

Study design: This is a complete hospital discharge dataset study.

Methods: Hospital discharge data were obtained from all hospital systems in the Coastal

Bend area of Texas from September 1, 2009, to August 31, 2012. The enhanced two-step

floating catchment area method was adopted to measure spatial access to health care,

including primary health care and hospitals. Multivariable regression methods were used

to measure the associations between spatial access to health care and ACSC rates of both

inpatient hospitalizations and ED visits.

Results: Spatial access to primary care has a statistically significant positive relationship

with both rates of inpatient hospitalization admissions and ED visits for ACSCs for the

elderly. Spatial access to hospitals has a statistically significant negative relationship with

both rates. Spatial access to primary care has a significantly negative contribution to the

likelihood of inpatient hospitalizations compared with the likelihood of ED visits for elderly

ACSCs, whereas spatial access to hospitals has a significantly positive contribution.

Conclusions: Spatial access to health care contributes to elderly ACSC hospitalizations. A

poorer access to primary care or a better access to hospitals increases both rates of inpa-

tient hospitalizations and ED visits for elderly ACSCs. Seniors living in areas where resi-

dents had poor access to primary care or easy access to hospitals were more likely to visit

EDs instead of being inpatients for ACSC conditions. Policy action is needed to improve

spatial access to primary care for the elderly.

Published by Elsevier Ltd on behalf of The Royal Society for Public Health.
puting Sciences, Texas A
25 5848.
(Y. Huang).

lf of The Royal Society fo
&M University e Corpus Christi, Unit 5868, Corpus Christi, 78412,

r Public Health.

mailto:lucy.huang@tamucc.edu

http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.005&domain=pdf

www.sciencedirect.com/science/journal/00333506

www.elsevier.com/puhe

https://doi.org/10.1016/j.puhe.2019.01.005

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 3 77
Introduction

Certain health conditions may prevent or reduce the need for

hospitalizations through timely and effective use of outpa-

tient primary or preventive care. These ambulatory care sen-

sitive conditions (ACSCs), often referred to as potentially

preventable conditions, include, for example, diabetes,

congestive heart failure, and bacterial pneumonia. Admission

rates of ACSCs have been used extensively as indicators of

access to and quality of primary care as well as the efficiency

of health systems.1 Limited access to health care increases

admission rates of ACSCs. Barriers to access include a number

of non-spatial and spatial factors. The non-spatial factors,

such as the availability of health insurance and costs associ-

ated with health care, have been well studied.2 In this study,

we focus on spatial factors that are often measured as spatial

accessibility to health care by taking location, distribution,

and capacity of health care into account.

Despite growing research on spatial access to health care

and ACSCs in recent years, most focus is on primary care.3,4

For example, the studies consistently have shown that

higher rates of hospitalization for ACSCs are strongly associ-

ated with poorer spatial access to primary health.5e7 Exam-

ining the relationship between ACSCs and spatial access to

hospitals, another main component of health care facilities,

will provide a more complete understanding of the impact of

health care on ACSCs.

Similar to inpatient hospital admissions for ACSCs, many

patients also seek emergency care for ACSCs. The inappro-

priate use of emergency departments (EDs) for ACSCs is an

expensive burden on hospitals and payers, particularly for

older adults. The impacts of spatial and non-spatial factors in

reducing ACSC inpatient hospitalizations are well-doc-

umented,8e10 yet the factors influencing ED visits for ACSCs

have not received the same thorough attention. One study

found disproportionately higher use of EDs for ACSC care

exists for many minority populations among adults aged 16

years and older, particularly for those who are black, Hispanic,

Medicare-covered, and older than 50 years.11 Moreover, most

ACSC studies used a secondary analysis of survey data such as

the National Hospital Ambulatory Medical Care Survey for

both inpatient hospitalizations and ED visits. Hospital

discharge data might provide a better comprehensive dataset.

In this paper, we examine associations of spatial access to

health care with both inpatient hospital admissions and ED

visits for ACSCs for older adults from a complete hospital

discharge dataset from hospital systems. The adults aged 65

years and older are a well-known high-risk group for ACSCs.

They often experience more hospitalizations and ED visits

than younger adults. We also investigate differences in spatial

access to primary care and hospitals on inpatient hospital

admissions and ED visits for elderly ACSCs. Our overarching

goal is to provide a more complete understanding of the

impact of health care in reducing ACSC hospitalization rates

for the elderly.
Methods

Data sources

The hospital discharge dataset was obtained from all hospital

systems in the 15 counties in the Coastal Bend area in the state

of Texas (Fig. 1), with a total population of more than 599,000

in 2010, consisting of 56% of Hispanics, 38% of whites, and 6%

of other minorities. Majority of the areas (77.8%) are consid-

ered nonmetropolitan or rural. According to the health need

assessments of 2010, 2013, and 2016 and other reports,12 this

area has limited access to healthcare services, particularly for

older adults, and a high hospitalization rate for ACSCs for

older adults. The health insurance coverage disparities also

vary spatially in this area.13

The hospital discharge data include both inpatient and ED

discharge data for the period of September 1, 2009, to August

31, 2012. Although patients who travel out of the Coastal Bend

area for hospitalization do not appear in the area’s hospital

systems, this dataset should represent elderly patients, and

therefore, they are considered as a 100% discharge data sys-

tem in this area. The data include the patient’s home zip code,
home county, discharge data, age, principal diagnosis code,

principal diagnosis description, secondary diagnosis descrip-

tion, and so on. To protect the confidentiality of hospitals and

patients, the data do not include any information that iden-

tifies the patients, so some of the cases may be repeat

patients.

Individual-level data on healthcare facilities were obtained

from InfoUSA, which is a residential and business database.

For each facility, in addition to its address, the dataset also

includes actual employee size, sales volume, primary stan-

dard industrial classification (SIC), and so on. All hospitals

(namely, SIC code is 806) from InfoUSA are included, and

primary health care includes offices and clinics of doctors

excluding those not for senior patients, such as pediatricians.

A total of 36 hospitals and 476 primary healthcare facilities

were identified during the study period.

The full list of potentially preventable hospitalization di-

agnoses and their International classification of diseases

(ICD-9) codes proposed by the Agency for Healthcare Research

and Quality’s prevention quality indicators14 are included in
this study except those related to prenatal care and delivery

and predominantly reflect children’s morbidity. The ACSC
inpatient hospitalizations were identified mainly based on

their primary diagnosis. The ACSC ED visits were also identi-

fied based on their primary ED diagnosis.11

The hospitalization data are only available at the zip code

level because of privacy and confidentiality restrictions.

Owing to this limitation, ACSC rates for both inpatient hos-

pitalizations and ED visits were calculated by zip code, which

is the total number of ACSC admissions at a zip code divided

by the total senior population at the zip code, where senior

population data were extracted from the 2010 Census Sum-

mary File 1.

https://doi.org/10.1016/j.puhe.2019.01.005

Fig. 1 e Study area: the Coastal Bend area of Texas.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 378
Spatial accessibility measurement to health care

We examined both primary care physicians (hereafter,

referred to as primary care) and hospitals. Primary care in-

cludes family and general practitioners and general in-

ternists.15 The enhanced two-step floating catchment area

(E2SFCA) method,16 a commonly used method to measure

spatial access, was adopted to measure both spatial
accessibility of primary care and hospitals by zip code. The

original E2SFCA method6 takes both availability and accessi-

bility into account. The E2SFCA method is better than the

original method by accounting for distance decay in catch-

ments for both facilities and population areas.16,17

First, we computed the physician-to-population ratio, R,

within the facility catchment area for each facility, j, through

searching all zip codes that are within a threshold travel

https://doi.org/10.1016/j.puhe.2019.01.005

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 3 79
distance (Eq. (1)). In this step, the center of each zip code was

represented by population weight center at the zip code,

which was calculated from the block-level population data.

The number of physicians at each facility was used to repre-

sent the facility’s capacity.6,16 A 30-min driving zone was used
for the maximum catchment of primary care.6,16,18 The

maximum catchment of hospitals was within a 60-min

driving zone because the golden hour is a common stan-

dard, particularly in emergency care.19

Rj ¼
SjP

k2fdkj�dojÞ
Pk

(1)

where Pk is the senior population at zip code k, and Sj is the

capacity for facility j, d0j is the maximum catchment associ-

ated with facility j, and dkj is the travel time between facility j

and zip code k.

Next, we calculated the access value, A, for each zip code, i,

by summing up the calculated ratio in the previous step for

facilities that are within the maximum catchment area of the

zip code (Eq. (2)). In this step, similar to the previous step, a 30-

min driving zone for primary care and 60-min driving zone for

hospitals were applied to urban/metropolitan areas. In small

towns or rural areas, however, the size of maximum catch-

ments was doubled to include some isolated areas in the

analysis.19,20 The urban and rural areas were determined

based on the 2012 RuraleUrban Commuting Area Codes.21 We

considered the distance decay in both steps, that is, there was

no decay within the first 10 min.16 In the areas between 10 min

and the corresponding maximum catchment, however, a

commonly used continuous decay function was adopted.19

Travel time was calculated using StreetMap Premium road

network. Owing to the space limit, the details of the E2SFCA

method are not provided here. For more details of the E2SFCA

method, see Luo and Qi.16

Ai ¼
X

j2fdij�d0iÞ
Rj (2)

where Rj is the physician-to-population ratio for facility j,

which is calculated from Step 1. doi is the maximum catch-

ment associated with zip code i, and dij is the travel time be-

tween zip code i and facility j.

Statistical analysis

Ordinary least squares (OLS), a common global linear regres-

sion, was used to examine the relations between spatial ac-

cess to health care and elderly ACSC admissions by zip code.

More specifically, ACSC rates for both inpatient hospitaliza-

tions and ED visits were the dependent variable, depending on

whether we examined inpatient hospitalizations (model 1) or

ED visits (model 2). The two measured spatial accessibility

values, one for primary care and the other for hospitals, were

the explanatory variables. Both of them were log-transformed

using eln(x) because of the fact that a higher score represents

a better spatial accessibility whereas a lower score represents

a poorer accessibility. To further compare inpatient hospital-

izations and ED visits for elderly ACSCs by health care, a

binomial logistic regression (Eq. (3)) was used to compare the
likelihood of inpatient hospitalizations and the likelihood of

ED visits for elderly ACSCs by both spatial access to primary

care and hospitals. OLS was performed in ArcGIS 10.4.1, and

the binomial logistic regression was conducted in R package.

In
�

1 � p
�
¼ b0 þ b1x1 þ b2x2 (3)

where, p is the probability of being inpatient hospitalizations,

1�p is the probability of being ED visits, x1 is the spatial access
to primary care, and x2 is the spatial access to hospitals. The

ratio between p and 1-p is the odds of visiting a hospital,

which is an alternative way of expressing probability. It tells

how many times a patient is more likely to visit a hospital

than seeking ED services. If odds equals to 1, it is equally likely

for a patient to visit a hospital and seek ED services; if odds is

greater than 1, it is more likely for a patient to visit a hospital

than seek ED services.
Results

Table 1 summarizes overall inpatient hospitalizations and ED

visits for older adults in the Coastal Bend area between

September 1, 2009, and August 31, 2012. A total of 58,395

elderly inpatient hospital discharges and 66,878 elderly ED

visits were identified, including 14,136 (24.2%) inpatient hos-

pital discharges and 13,686 (20.5%) ED visits for ACSCs. Fig. 2

(a) and (b) shows the geographic distribution of rates of

inpatient hospitalizations and ED visits, respectively, for

ACSCs for older adults. The rates varied across the study.

Overall, southern areas have a higher rate than the northern

areas.

Fig. 3(a) and (b) present the variations of spatial accessi-

bility to primary care and hospitals, respectively, at the zip

code level. Lower values represent a poorer access whereas

higher values correspond to better access to health care. Jenks

natural breaks, a common method used to find the best

arrangement of values into different classes, is used to classify

the accessibility values into five groups. Not surprising, the

two maps show some similar patterns, that is, the areas with a

better spatial access usually are located in the places that have

a high concentration of primary care or hospitals. Lower

accessibility scores were observed in areas with scarce facil-

ities and mostly in rural areas.

Table 2 reports the results of the two OLS models, one for

the rate of inpatient hospitalizations (model 1) and the other

for ED visits (model 2). The variance inflation factor for each

variable in the two models is less than 7.5, indicating no

redundancy in the two spatial access variables. The P-value of

a robust test (Robust_Pr) was used to determine coefficient

significance because the Koenker (BP) Statistics is statistically

significant. The Moran’s I values on the residuals are 0.065 and
0.05 for model 1 and model 2, respectively, indicating no

clustering on the residuals.

Not surprising, both spatial accessibility of primary care

and hospitals is significantly associated with rates of inpatient

hospitalizations and of ED visits for elderly ACSCs. Interest-

ingly, spatial access to primary care has a statistically signif-

icant positive relationship with both rates whereas spatial

access to hospitals has a statistically significant negative

https://doi.org/10.1016/j.puhe.2019.01.005

Table 1 e Characteristics of inpatient hospitalization admissions and emergency department visits for older adults (65
years and older) (Sept. 1, 2010eAugust 31, 2012).

Characteristic Inpatient hospitalizations Emergency department visits

All admissions ACSC admissions All visits ACSC visits

No. % No. % No. % No. %

Total 58,395 14,136 (24.2%) 66,878 13,686 (20.5%)

Age in years <79 35,334 60.5% 7785 55.1% 43,550 65.1% 8838 64.6% �80 23,061 39.5% 6351 44.9% 23,328 34.9% 4848 35.4% Gender Female 31,919 54.7% 7984 56.5% 40,939 61.2% 8848 64.7% Male 26,476 45.3% 6152 43.5% 25,939 38.8% 4838 35.3% Race White 16,481 28.2% 3659 25.9% 15,173 22.7% 2966 21.7% Hispanic 13,468 23.1% 3191 22.6% 17,844 26.7% 3724 27.2% Others 1831 3.1% 423 3.0% 2227 3.3% 437 3.2% Unknown 26,615 45.6% 6863 48.5% 31,634 47.3% 6559 47.9% Occupation Retired 47,201 80.8% 11,721 82.9% 50,180 75.0% 10,363 75.7% Unemployed 6388 10.9% 1533 10.8% 9842 14.7% 2155 15.7% Others 4806 8.2% 882 6.2% 6856 10.3% 1168 8.5% ACSC, ambulatory care sensitive condition. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 380 relationship with both rates, that is, a poorer access to pri- mary care or a better access to hospitals increases both rates of inpatient hospitalizations and ED visits for elderly ACSCs. Table 3 reports the binomial logistic regression results of the comparison of inpatient hospitalizations and visits for elderly ACSCs by spatial access to health care. Interestingly, both spatial access to primary health care and hospitals are statistically significant and contribute to the comparison be- tween the likelihood of inpatient hospitalizations and ED visits for elderly ACSCs. They play different roles; however, Fig. 2 e Geographic distributions of hospitalization rates of ACS condition; ED, emergency department. spatial access to primary care has a negative coefficient, whereas spatial access to hospitals has a positive coefficient. This result shows that a senior from a zip code area with a poorer access to primary care (namely, a higher value) or a better access to hospitals (namely, a lower value) is more likely to visit EDs for ACSCs. Specifically, when the value of spatial access to hospitals is fixed, for a one unit increase (getting poorer access) in variable ‘access to primary care’, the log odds of inpatient hospitalizations decreases by �0.2369. In this case, the probability of inpatient hospitalizations Cs for older adults. ACSC, ambulatory care sensitive https://doi.org/10.1016/j.puhe.2019.01.005 https://doi.org/10.1016/j.puhe.2019.01.005 Fig. 3 e Spatial access to health care. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 3 81 decreased whereas the probability of ED visits increased. Conversely, when the value of spatial access to primary care is fixed, for a one unit increase (getting poorer access) in variable ‘access to hospitals’, the log odds of being inpatients increased by 0.22653. Discussion This study found a statistically significant positive relation- ship between spatial access to primary care and the rate of inpatient hospitalizations for ACSCs among older adults. This finding conforms with the results of previous studies showing that less access to primary care in an area increases the rate of inpatient hospitalizations for ACSCs.6,22,23 This study also confirmed a significant positive association between spatial access to primary care and the rate of ED visits for ACSCs among older adults.24 The findings suggest that timely and Table 2 e Ordinary least squares (OLS) analysis: association ef hospitalizations and elderly ED visits for ACSCs. Variables Inpatient Hospitalizations Coefficient Standard error Robust_P Model 1 Access to primary care 0.033 0.014 0.045** Access to hospitals �0.050 0.019 0.019** ED, emergency department; VIF, variance inflation factor. * An asterisk next to a number indicates a statistically significant p-value effective use of outpatient primary care facilities should decrease both rates of inpatient hospitalizations and ED visits for ACSCs among the elderly. Our study further demonstrated the significant associa- tions between spatial access to hospitals and ACSCs, including both rates of inpatient hospitalizations and ED visits, for older adults, where the relations between them are not previously documented. Interestingly, seniors living in areas where residents had easy access to hospitals had high rates of both. This finding may indicate inappropriate use of emergency services for ACSCs for the elderly. Owing to the cost, hospitals benefit more when EDs are focused on providing emergency care. Elderly patients, however, often visit the ED instead of primary care providers.25 Previous studies have shown a significant number of ED visits for ACSCs and they are related to demographic characteristics and socio-economically disadvantaged neighborhoods.11,26 Considering the US healthcare system that provides services fects of spatial accessibility on rates of elderly inpatient ED Visits r VIF Coefficient Standard error Robust_Pr VIF Model 2 3.553 0.056 0.016 0.002*** 3.553 3.553 �0.071 0.022 0.002*** 3.553 (p < 0.01): *: p < 0.1, **: p < 0.05, ***: p < 0.01, ****: p< 0.001. https://doi.org/10.1016/j.puhe.2019.01.005 https://doi.org/10.1016/j.puhe.2019.01.005 Table 3 e Comparison of rates of inpatient hospitalizations and ED visits for ACSCs for the elderly by spatial access to health care: coefficients. Characteristic Estimate Standard error Z value Pr (>jzj)

Access to

primary care

�0.2369 0.01281 �10.689 <2e-16 *** Access to hospitals 0.22653 0.01663 7.007 2.44e-12 *** ACSC, ambulatory care sensitive condition; ED, emergency department. * An asterisk next to a number indicates a statistically significant p-value (p < 0.01): *: p < 0.1, **: p < 0.05, ***: p < 0.01, ****: p< 0.001. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 382 to seniors based on need, rather than on the ability to pay, seniors who have easy access to hospitals might be more likely to visit hospitals or EDs rather than seeking primary health service. To the best of our knowledge, this study is the first to compare inpatient hospitalizations and ED visits for elderly ACSCs by spatial access to health care. We observed that se- niors living in areas where residents had poor access to pri- mary care or easy access to hospitals were more likely to visit EDs instead of being inpatients for ACSCs. Reducing ED visits for ACSCs by elderly patients has significant implications for reducing cost, improving quality, and enhancing efficiency.27 To better understand hospitalization admission rates for elderly ACSCs, both inpatient hospitalizations and ED visits should be considered. Fingar28 showed that the rate of ED visits for ACSCs by the elderly was increased although the rate of inpatient hospitalizations for the same condition was decreased. This study confirms that the elderly who are un- able to obtain timely outpatient care often seek care in EDs.29 Our study has several limitations. First, inpatient hospital admissions were probably overestimated by including pa- tients who were transferred from ED visits as we were not able to identify treat-and-release ED visits. Second, similar to other research that used hospital discharge dataset, we could not determine the repeat patients because of data confidentiality laws. Third, the study area of the Coastal Bend area in Texas in this study may have limited the study's power to detect sig- nificant associations at a national level. The findings, how- ever, are consistent with the previous research. Finally, we selected and identified ACSCs based on ICD-9 codes. Undiag- nosed ACSCs may not be included in the analysis. In conclusion, despite these limitations, our study docu- ments the contribution of primary care physicians to elderly health and efficiency of healthcare systems. Primary care is in the great position to detect the occurrence of ACSCs. Improvement in healthcare systems to achieve better spatial access to primary care is imperative, especially for the elderly. On the one hand, it improves the quality of life for the elderly by reducing their hospitalizations for ACSCs. On the other hand, it provides more efficient uses and better quality of EDs through effective primary or preventive health care. Research into strategies that increase access to primary health care for elderly living may help decrease use of EDs for primary care. Health providers and policy makers should work together to increase equitable access to primary healthcare to decrease ACSC hospitalizations and ED usage. This could provide elders with an opportunity for health and well-being. Author statements Acknowledgments The research reported in this manuscript was funded through a local contract for the Coastal Bend of Texas Health Needs Assessment Task Force TAMUCC Project #634180. Ethical approval Protocol approval for the Coastal Bend Health Needs Assess- ment was obtained through the International Review Board process in our campus. The data for this study were taken from this project. Funding This work was supported by the Coastal Bend Health Needs Assessment Task Force. Competing interests None declared. r e f e r e n c e s 1. Moy E, Chang E, Barrett M. Potentially preventable hospitalizations e United States, 2001e2009. Morb Mortal Wkly Rep (MMWR) 2013;62(03):139e43 (Suppl; November 22, 2013). 2. Guagliardo M. Spatial accessibility of primary care: concepts, methods and challenges. Int J Health Geogr 2004;3(3). Available at: https://doi.10.1186/1476-072X. 3. Wang FH, Luo W. Assessing spatial and nonspatial factors for healthcare access: towards an integrated approach to defining health professional shortage areas. Health Place 2005;11:131e46. 4. Mobley L, Root E, Luc A, Lozano-Gracia N, Koshchinsky J. Spatial analysis of elderly access to primary care services. Int J Health Geogr 2006;5:19. Available at: https://doi.org/10.1186/ 1476-072X-5-19. 5. Ricketts T, Randolph R, Howard H, Pathman D, Carey T. Hospitalization rates as indicators of access to primary care. Health Place 2001;7:27e38. 6. Luo W, Wang F. Measures of spatial accessibility to health care in a GIS environment: synthesis and a case study in the Chicago region. Environ Plan B 2003;30:865e84. 7. Lin Y, Eberth J, Probst J. Ambulatory careesensitive condition hospitalizations among medicare beneficiaries. Am J Prev Med 2016;51(4):493e501. 8. Begley C, Slater C, Engel M, Reynolds T. Avoidable hospitalizations and socio-economic status in Galveston County, Texas. J Commun Health 1994;19(5):377e87. 9. Bindman AB, Grumbach K, Osmond D, Komaromy M, Vranizan K, Lurie N, et al. Preventable hospitalizations and access to health care. J Am Med Assoc 1995;274:305e11. 10. Blustein J, Hanson K, Shea S. Preventable hospitalizations and socioeconomic status. Health Aff (Millwood) 1998;17:177e89. http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref1 https://doi.10.1186/1476-072X http://refhub.elsevier.com/S0033-3506(19)30005-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref3 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref3 https://doi.org/10.1186/1476-072X-5-19 https://doi.org/10.1186/1476-072X-5-19 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref5 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref6 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref6 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref6 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref6 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref7 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref8 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref9 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref10 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref10 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref10 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref10 https://doi.org/10.1016/j.puhe.2019.01.005 https://doi.org/10.1016/j.puhe.2019.01.005 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 7 6 e8 3 83 11. Johnson P, Ghildayal N, Ward A, Westgard B, Boland L, Hokanson J. Disparities in potentially avoidable emergency department (ED) care. Med Care 2012;50(12):1020e8. 12. Community Health Status Indicators. (CHSI 2015), Information for improving community health. 2015. http://wwwn.cdc.gov/ CommunityHealth/info/AboutProject/TX/Nueces/ (accessed 28 November 2017). 13. Huang Y, Meyer P. Mapping spatial variations of health insurance coverage in the Coastal Bend, Texas. J Maps 2012. https://doi.org/10.1080/17445647.2012.745382. 14. Agency for healthcare research and Quality (AHRQ). Ambulatory care sensitive conditions. Rockville, MD: US Department of Health and Human Services AHRQ Publication October; 2017. Available at: http://www.ahrq.gov/professionals/quality-patient-safety/ quality-resources/tools/ambulatory-care/index.html (accessed 26 November 2017). 15. Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q 2005;83(3):457e502. 16. Luo W, Qi Y. An enhanced two-step floating catchment area (E2SFCA) method for measuring spatial accessibility to primary care physicians. Health Place 2009;15:1100e7. 17. Shi X, Teaster J, Onega T, Wang D. Spatial access and local demand for major cancer care facilities in the United States. Ann Assoc Am Geogr 2012;102(5):1125e34. 18. Lee RC. Current approaches to shortage area designation. J Rural Health 1991;7:437e50. 19. McGrail M, Humphreys J. The index of rural access: an innovative integrated approach for measuring primary care access. BMC Health Serv Res 2009;9:124. 20. Wan N, Zhan F, Zou B, Wilson G. Spatial access to health care services and disparities in colorectal cancer stage at diagnosis in Texas. Prof Geogr 2012;65(3):527e41. 21. RUCA, Rural Health Research Center: rural-urban commuting area codes 2017 (RUCAS). Available at: http://depts. washington.edu/uwruca (accessed 26 November 2017). 22. Laditka J. Physician supply, physician diversity, and outcomes of primary health care for older persons in the United States. Health Place 2004;10:231e44. 23. Marques A, Montilla D, Almeida W, Andrade C. Hospitalization of older adults due to ambulatory care sensitive conditions. Publ Health Practice 2014;48(5):817e26. 24. Huang Y, Meyer P, Jin L. Neighborhood socioeconomic characteristics, healthcare spatial access, and emergency department visits for ambulatory care sensitive conditions for elderly. Prev Med Rep 2018;12:101e5. 25. Gulacti U, Lok U, Celik M, Aktas N, Polat H. The ED use and non-urgent visits of elderly patients. Turk J Emerg Med 2016;16(4):141e5. 26. McWilliams A, Tapp H, Barker J, Dulin M. Cost analysis of the use of emergency departments for primary care services in Charlotte, North Carolina. NCMJ 2011;72(4):265e71. 27. Agency for healthcare research and quality (AHRQ). National healthcare disparities report. Rockville, MD: US Department of Health and Human Services AHRQ Publication; 2012. No 12- 0006. 28. Fingar K, Barrett M, Elihauser A, Stocks C, Steiner C . Trends in potentially preventable inpatient hospital admissions and emergency department visits. Agency for Healthcare Research and Quality (AHRQ); Nov. 2015. 29. Dresden S, Feinglasss J, Kang R, Adams J. Ambulatory care sensitive hospitalizations through the emergency department by payer: comparing 2003 and 2009. Admin Emerg Med 2016;50(1):135e42. http://refhub.elsevier.com/S0033-3506(19)30005-8/sref11 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref11 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref11 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref11 http://wwwn.cdc.gov/CommunityHealth/info/AboutProject/TX/Nueces/ http://wwwn.cdc.gov/CommunityHealth/info/AboutProject/TX/Nueces/ https://doi.org/10.1080/17445647.2012.745382 http://www.ahrq.gov/professionals/quality-patient-safety/quality-resources/tools/ambulatory-care/index.html http://www.ahrq.gov/professionals/quality-patient-safety/quality-resources/tools/ambulatory-care/index.html http://refhub.elsevier.com/S0033-3506(19)30005-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref15 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref16 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref17 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref18 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref19 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref19 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref19 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref20 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref20 http://depts.washington.edu/uwruca http://depts.washington.edu/uwruca http://refhub.elsevier.com/S0033-3506(19)30005-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref22 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref23 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref24 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref24 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref24 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref24 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref24 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref25 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref25 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref25 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref25 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref26 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref26 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref26 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref26 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref27 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref28 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref28 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref28 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref28 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref29 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref29 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref29 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref29 http://refhub.elsevier.com/S0033-3506(19)30005-8/sref29 https://doi.org/10.1016/j.puhe.2019.01.005 https://doi.org/10.1016/j.puhe.2019.01.005 Spatial access to health care and elderly ambulatory care sensitive hospitalizations Introduction Methods Data sources Spatial accessibility measurement to health care Statistical analysis Results Discussion Author statements Acknowledgments Ethical approval Funding Competing interests References Effects-of-cooking-fuel-sources-on-the-respiratory-health-of-chi_2019_Public ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 8 Available online at w Public Health journal homepage: www.elsevier.com/puhe Original Research Effects of cooking fuel sources on the respiratory health of children: evidence from the Annual Health Survey, Uttar Pradesh, India Surendra Kumar Patel a,*, S. Patel a, A. Kumar b a International Institute for Population Sciences (IIPS), Mumbai, Maharashtra, India b Jawaharlal Nehru University, New Delhi, India a r t i c l e i n f o Article history: Received 29 August 2018 Received in revised form 5 December 2018 Accepted 7 January 2019 Available online 26 February 2019 Keywords: Cooking fuels Dung cake Lighting sources Acute respiratory infections * Corresponding author. E-mail addresses: surendrabhu20@gmail gmail.com (A. Kumar). https://doi.org/10.1016/j.puhe.2019.01.003 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: India is predominantly a rural country, where more than two-thirds of the popu- lation live in rural areas. The majority of the rural population use crop residue, firewood, and dung cake as cooking fuel sources. Combustion of these fuels emits life-threatening pollut- ants that contaminate the household environment, which can have serious health conse- quences, especially for young children. This study examines the use of cooking fuel sources and their association with acute respiratory infections (ARIs) in children aged 0e59 months. Study design: This study used data from the second update of the Annual Health Survey (2012e13). The prevalence rate was measured in terms of the number of children per 1000 children suffering from ARIs. Bivariate analysis was used to analyze the use of different cooking fuels (in percentage) and the prevalence of ARIs in Uttar Pradesh. Methods: District-level variations in the cooking fuels used were assessed by simple bivariate analysis for all districts of Uttar Pradesh. A logistic regression was used to examine the association of household environment and pollutants with ARIs. Results: In total, 89 per 1000 children suffered from ARIs in Uttar Pradesh. Infants (0e11 months) were significantly more likely to suffer from ARIs than older children (12e59 months). Households using dung cake for cooking and kerosene and other oils for lighting were found to have significantly higher odds for ARIs (odds ratio [OR]: 1.21; 95% confidence interval [CI]: 1.17e1.25 and OR: 1.07; 95% CI: 1.05e1.10, respectively). A considerable interdistrict difference was observed in the cooking fuel used and the prevalence of ARIs among children aged 0e59 months. Conclusions: The type of cooking fuel and lighting source used were found to be significant predictors of ARIs among children aged 0e59 months. These results highlight the need for targeted efforts for the provision of clean cooking fuels (liquid petroleum gas/biogas/ electricity) and for the improvement in knowledge and awareness of ARIs and exposure to cooking and lighting pollution. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. .com (S.K. Patel), sunitap ic Health. Published by E atel56@gmail.com, sunitapatel56@iips.net (S. Patel), ajitcpsjnu@ lsevier Ltd. All rights reserved. mailto:surendrabhu20@gmail.com mailto:sunitapatel56@gmail.com mailto:sunitapatel56@iips.net mailto:ajitcpsjnu@gmail.com mailto:ajitcpsjnu@gmail.com http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.003&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 860 Introduction Access to energy is essential to meet the basic needs of cooking and lighting. Around 2.8 billion people (38% of the global population and almost 50% of the total population in developing countries) lack access to clean cooking fuels; thus, they rely on the use of solid biomass fuels, such as firewood, crop residue, and dung cake.1 In India, the National Institution for Transforming India (NITI) Aayog estimates that around two-thirds of households use solid cooking fuels (e.g. fire- wood, crop residue, cow dung cake, and coal/lignite/charcoal). In Uttar Pradesh, an estimated 80% of households use solid cooking fuels.2 Children are the most vulnerable group within the house- hold when it comes to respiratory illnesses arising from proximity to pollution from domestic cooking fuels.3,4 Out of the total deaths in India, 85e88% are contributed to by acute respiratory infection (ARI) episodes categorized as acute upper respiratory infections and acute lower respiratory infections (ALRIs).5,6 India's under five mortality rates declined from 125 to 49 per 1000 live births from 1990 to 2013, and the Millen- nium Development Goals report suggests that India is roughly on track because of a sharp decline in child mortality.7 The most significant health impact of indoor pollution is observed among the poorest and the most vulnerable sections of society.8 Household pollution from unprocessed fuel causes respiratory problems, which is a major concern of public health in India.8 Globally, around 4.3 million deaths per year are attributed to household air pollution. Indoor air pollution emitted from burning biomass fuels in homes with inade- quate ventilation has deleterious effects on health.9 A recent study conducted in Bangladesh demonstrates the increased risk for ALRIs as a result of exposure to indoor particulate matters (PM2.5). 10,11 The conventional sources of indoor PM include combustion of biofuels (e.g. crop residue, firewood, and dung cake) used for cooking purposes.12 Several studies show that biomass fuels cause high pollu- tion.3,11 Exposure to indoor air pollution, especially PM from the combustion of biofuels has been found to be associated with respiratory illnesses.13,14 Smoke/fume emitted by biomass fuels contains hundreds of chemical compounds, such as PMs, carbon monoxide, nitrogen dioxide, sulfur di- oxide, polycyclic aromatic hydrocarbons, and other volatile organic compounds, that have a significant impact on health. These pollutants increase the respiratory problems in the form of a persistent cough among adults and ALRIs among children aged below 5 years as well as older children.14,15 Several studies show that household environment is a sig- nificant predictor of ARI prevalence among children.8,12,16,17 In the present study, we tried to examine the association of cooking fuels and household environment with ARIs among children aged 0e59 months in Uttar Pradesh, India. A few studies have investigated the prevalence of ARIs among children.18e20 However, very little has been explored with respect to the association between household environment and ARIs among children at the national level.21 This study focused on Uttar Pradesh, where most of the population re- sides in rural areas and only a small population has access to liquid petroleum gas/piped natural gas (LPG/PNG) and biogas, indicating a need for studying the causes of ARIs. Children are at high risk from pollution in the household environment.17 Therefore, we tried to examine the association between household pollution and ARIs in children aged 0e59 months in Uttar Pradesh, India. Methods For the present research article, data from the second update of the Annual Health Survey (AHS, 2012e13) have been used. The objective of the AHS is to monitor the performance and outcome of government interventions in the health sectors, including those under the National Rural Health Mission. In addition, the AHS also aims to tailor interventions through benchmarking and preparing a comprehensive district health profile on core vital and health parameters. The second update of AHS collected information from nine Empowered Action Group states, namely Bihar, Jharkhand, Uttar Pradesh, Uttar- akhand, Madhya Pradesh, Chhattisgarh, Odisha, Rajasthan, and Assam. In Uttar Pradesh, the total surveyed population was 20.9 million, comprising 4.3 million households. Uttar Pradesh accounts for about 19% of India's total population. The AHS provides information on different socio-economic and demographic characteristics, such as marriage, disability and injuries, acute illness, chronic illness, fertility, abortion, birth intervals, antenatal care, and childhood diseases. The present study focused on information about ARIs in children aged below 5 years in Uttar Pradesh, India. Variables used in the analysis Outcome variable The outcome variable used in the analysis was ARIs. The survey asked participants to select the ‘types of illness re- ported during the last 15 days?’ The responses were diarrhea, dysentery, acute respiratory illness, jaundice with fever, fever with chills/rigors (malaria), fever for short duration with rashes, other types of fever, reproductive tract infection, others, and no illness. The responses were coded in a binary form, with ‘0’ representing ‘No’ (i.e. no ARI reported) and ‘1’ standing for ‘Yes’ (i.e. ARI reported). Predictor variables The household environment and pollutant variables included the type of cooking fuels and lighting sources used in the house. Cooking fuels had five categories: LPG/PNG/electricity/ biogas, firewood, crop residue, dung cake, and other sources (e.g. coal/lignite/charcoal, kerosene, and any other). Lighting sources had three categories: electricity/solar, kerosene/other oils, and any other. The place of cooking had four categories: in-house-kitchen, in-house-no-kitchen, out-house-kitchen, and out-house-no-kitchen. These predictors (cooking fuels, lighting sources, and place of cooking) were considered the primary predictors of ARI among children aged 0e59 months. Socio-economic and demographic characteristics Characteristics included were sex (male and female), place of residence (rural and urban), religion (Hindu, Muslim, and other), and age of child (<12 months, 12e23 months, 24e35 https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Table 1 e Profile characteristics of respondents. Characteristics Sample distributiona % n Total 100 932,341 Sex of child Male 52.4 488,152 Female 47.6 444,189 Age group (in months) <12 18.8 175, 539 12e23 19.6 182,532 24e35 22.5 209,714 36e47 20.2 187,997 48e59 18.9 176,559 Caste categories Other than SC/ST 73.6 686,249 SC/ST 26.4 246,092 Religion Hindu 79.7 743,117 Muslim 19.9 185,510 Othersb 0.4 3714 Wealth index Poorest 21.3 198,720 Poor 20.0 186,220 Middle 24.0 223,267 Rich 22.1 206,193 Richest 12.1 117,941 Cooking fuel sources LPG/PNG/electricity/biogas 13.4 125,231 Firewood 40.5 377,629 Crop residue 7.7 71,979 Dung cake 37.6 350,338 Othersc 0.8 7164 Cooking place In house: kitchen 29.2 272,380 In house: No kitchen 56.9 530,628 Out house: kitchen 4.4 41,390 Out house: no kitchen 9.4 87,943 Lighting source p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 8 61 months, 36e47 months, and 48e59 months). Castes were categorized into two groups: scheduled castes/scheduled tribes (SC/ST) and other than scheduled castes/scheduled tribes (other than ST/SC). The wealth quintile was categorized into five groups: poorest, poor, middle, rich, and richest. Finally, region was classified as follows: eastern region, southern upper Ganga plain, northern upper Ganga plain, southern region, and central region. Statistical analysis The prevalence rate was estimated as number of children per 1000 children suffering from ARIs. Bivariate analysis was used to analyze the use of different cooking fuels (in percentage) and the prevalence of ARI in Uttar Pradesh. District-level variations in the cooking fuels used were assessed by simple bivariate analysis for all the districts of Uttar Pradesh. A lo- gistic regression was fitted to examine the association of household environment and pollutant with ARI. Using logistic regression analysis, the following three models were identified: Model 1: including only cooking fuels; Model 2: including household environment, pollutant (cooking fuel and lighting sources), and place of cooking; and Model 3: adjusted for household environment, pollutant, and place of cooking and socio-economic and demographic characteristics. The odds ratio (OR) was estimated at 95% confidence in- tervals (CIs). All the data were analyzed using STATA (Stata- Corp LLC, USA) (version 13), and Arc-GIS (ESRI, USA) has been used to show district-level variations in the use of dung cake and prevalence of ARI. Electricity and solar 32.8 305,881 Kerosene and other oils 66.1 616,147 Other sourcesd 1.1 10,313 Place of residence Rural 84.5 787,746 Urban 15.5 144,595 Regional categories Eastern region 37.2 346,763 Southern Upper Ganga plane 33.2 309,868 Northern Upper Ganga plane 16.1 150,111 Southern region 8.4 77,948 Central region 5.1 47,651 LPG, liquid petroleum gas; PNG, piped natural gas; SC/ST, sched- uled castes/scheduled tribes. a Percentage of sample based on weighted analysis and sample size (n) are un-weighted. b Includes Christian, Sikh, Buddhist, Jain, and other religion group. c Includes coal, lignite, charcoal, kerosene, and any other source. d Includes other oils and any other sources for lighting sources. Results AHS data indicate that around 18.8% of children were <12 months old, and 84.5% of children were from the rural areas. In total, 73.6% of children belonged to ‘other than SC/ ST’ communities and 79.7% belonged to the Hindu religion. Household environment, pollutant (cooking fuels and lighting sources), and place of cooking in the household were taken into consideration. Data showed that 40.5% of households were using firewood, and 37.6% of households were using dung cake as cooking fuels. Also, 56.9% of households cooked food inside the house without a kitchen, whereas only 29.2% had a kitchen inside the house. Results indicate that 66.1% of households used kerosene and other oils as lighting sources in the house, whereas only 32.8% used electricity/solar energy as lighting sources (Table 1). Fig. 1 shows the pattern of cooking fuels used in rural Uttar Pradesh. It was found that 44% of households were using firewood, 43% dung cake, and only 5% were using LPG/PNG/ electricity/biogas as cooking fuels. In contrast, in the urban areas, 57% of the households were using LPG/PNG/electricity/ biogas, whereas 22% and 16% were using firewood and dung cake, respectively, for cooking purposes (Fig. 2). The distribution of the source of cooking fuels used with the type of kitchen has been presented for rural areas in Fig. 3 and for urban areas in Fig. 4. In rural areas, those having an in- house kitchen mostly used firewood (41.3%) and dung cake (37.8%); only 15.0% of households used LPG/PNG/electricity/ https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Fig. 1 e Cooking fuel sources used in rural areas in Uttar Pradesh, 2013. LPG, liquid petroleum gas; PNG, piped natural gas. Fig. 2 e Cooking fuel sources used in urban areas in Uttar Pradesh, 2013. LPG, liquid petroleum gas; PNG, piped natural gas. Fig. 3 e Cooking fuel sources used in kitchen in rural areas in Uttar Pradesh, 2013. LPG, liquid petroleum gas; PNG, piped natural gas. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 862 https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Fig. 4 e Cooking fuel sources used in kitchen in urban areas in Uttar Pradesh, 2013. LPG, liquid petroleum gas; PNG, piped natural gas. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 8 63 biogas. A similar scenario was observed in the case of the in- houseeno-kitchen, out-house-kitchen, and out-house-no- kitchen. In the urban areas, 75.2% of households having an in-house-kitchen were using LPG/PNG/electricity/biogas, and 11.1% and 10.0% were using firewood and dung cake, respec- tively. However, households having no kitchen inside the house were using firewood (38.0%) more than any other cooking fuel. Households having an out-house kitchen facility were also using firewood (41.0%) as well as dung cake (39.4%) for cooking (Fig. 4). Fig. 5 shows the distribution of the use of dung cake as a cooking fuel in Uttar Pradesh. More than 45% of households in the southern and northern upper Ganga plain regions were using dung cake as the dominant cooking fuel (except Baghpat [5%] and Saharanpur [24%]). In the districts of the central re- gion, dung cake was used as a cooking fuel in less than 15% of households (except Kanpur [59%] and Jalaun [48%]). Districts in the southern region also used less dung cake than in the upper Ganga plain regions. In the eastern region, the use of dung cake was very diverse, with less than 15% of households using it in the Sonbhadra and Maharajganj districts and more than 60% of households using it in Chandauli and Ghazipur. The distribution of the prevalence of ARI is similar to the use of dung cake in Uttar Pradesh. There was a higher prevalence of ARI in the districts of the southern upper Ganga plain region and the eastern region (Fig. 6). Fig. 6 shows that a high prev- alence of ARI was found in Azamgarh (17.7%), followed by Srawasti (16.6%), and Ambedkar Nagar (15.5%). Districts in central and southern regions had less than 1% prevalence of ARI. The prevalence of ARI among children aged 0e59 months was found to be about 90 per 1000 children (Table 2). A slightly higher prevalence of ARI was observed among male compared with female children. The prevalence of ARI was found to decrease with increasing age. For example, the prevalence of ARI was 112 per 1000 children aged <12 months compared with 77 per 1000 children aged 48e59 months. The prevalence of ARI was higher among the ‘other than SC/ST’ group compared with the ‘SC/ST’ group and among Muslims compared with Hindus. The ARI prevalence was higher among middle (93 per 1000) and poor households (90 per 1000) compared with the poorest (87 per 1000) and richest households (88 per 1000). In terms of the type of cooking fuel used, ARI prevalence was highest among households that used dung cake (107 per 1000 children). The corresponding figures were 98 per 1000 children for those who used crop residue and 72 per 1000 children for those who used firewood. The source of lighting is also an important component of the household environment. The prevalence of ARI was highest among households that used kerosene and other oils (93 per 1000 children) compared with those who used elec- tricity and solar energy sources (80 per 1000 children). A huge difference in the prevalence of ARI was found across the different regions of Uttar Pradesh, with prevalence being highest in the eastern region (119 per 1000) followed by the northern upper Ganga plain region (89 per 1000) and lowest in the southern region (only 14 per 1000 children). Table 3 presents the OR of the predictors of ARI in Uttar Pradesh. Model 1 (unadjusted for socio-economic predictors) shows that crop residue, dung cake, and other cooking fuels had significantly higher odds than LPG/PNG/electricity/biogas. Households using dung cake were significantly more likely to suffer from ARIs (OR: 1.33; 95% CI: 1.29e1.36) compared with those using LPG/PNG/electricity/biogas. After adjustment with lightening source and cooking place, the crop residue became insignificant, and dung cake had slightly lower odds compared with the unadjusted result of cooking fuels. However, with socio-economic and demographic adjustment, the crop res- idue and firewood have significantly lower odds, and dung cake has higher odds compared with LPS/PNG/electricity/ biogas. There was a linear association between age and the prevalence of ARI; children aged 48e59 months were 35% less likely to experience ARIs compared with those aged <12 months old. The ORs for children aged 48e59 months and 36e47 months were 0.64 and 0.72, respectively. Compared with Hindus, Muslims were significantly more likely (OR: 1.09; 95% CI: 1.06e1.11) to have an ARI. Children from urban areas https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Fig. 6 e Prevalence of acute respiratory infections among ages 0e59 months, Uttar Pradesh, 2013. Fig. 5 e Percentage distribution of dung cake used for cooking in households, Uttar Pradesh, 2013. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 864 were significantly more likely to have an ARI (OR: 1.28; 95% CI: 1.24e1.32) than those from rural areas. Compared with chil- dren from the eastern region, the odds of experiencing an ARI were significantly lower among children from the southern (OR: 0.11; 95% CI: 0.11, 0.12), central (OR: 0.20; 95% CI: 0.19, 0.22), and southern upper Ganga plain regions (OR: 0.55; 95% CI: 0.54, 0.56). Discussion The study found that approximately two-thirds of households in rural areas and about two-fifths of households in urban areas were using firewood and dung cake as cooking fuels. This finding is similar to the NITI Aayog report that shows a https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Table 2 e Prevalence rate of acute respiratory infections (ARIs) among children aged 0e59 months in Uttar Pradesh, 2013. Characteristics Prevalence rate of ARI (per 1000) Total 88.8 Sex of child Male 89.6 Female 88.0 Age (in months) <12 112.1 12e23 84.3 24e35 86.7 36e47 84.8 48e59 77.2 Caste categories Other than SC/ST 91.3 SC/ST 82.0 Religion Hindu 87.4 Muslim 94.5 Othersa 78.2 Wealth index Poorest 87.2 Poor 90.0 Middle 93.4 Rich 84.4 Richest 88.4 Cooking fuel sources LPG/PNG/electricity/biogas 82.9 Firewood 71.8 Crop residue 98.4 Dung cake 107.1 Othersb 95.2 Cooking place In house: kitchen 75.7 In house: No kitchen 91.8 Out house: kitchen 100.2 Out house: no kitchen 106.5 Lighting source Electricity and solar 79.8 Kerosene and other oils 93.3 Other sourcesc 80.1 Place of residence Rural 89.4 Urban 86.4 Regional categories Eastern region 119.3 Southern upper Ganga plane 72.3 Northern upper Ganga plane 89.4 Southern region 14.4 Central region 26.6 LPG, liquid petroleum gas; PNG, piped natural gas; SC/ST, sched- uled castes/scheduled tribes. a Includes Christian, Sikh, Buddhist, Jain, and other religion group. b Includes coal, lignite, charcoal, kerosene, and any other source. c Includes other oils and any other sources for lighting sources. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 8 65 higher percentage of households were using dung cake as cooking fuel in Uttar Pradesh than in any other states.2 Results presented here show that those who cooked in an in-house kitchen, an out-house kitchen, or in the absence of a kitchen were more likely to be using firewood and dung cake as cooking fuels in rural areas. The availability of clean fuel is a major challenge for poor rural households due to affordability factors.22 A study by Dasgupta et al.23 in Bangladesh suggested that ventilation and the structure of the kitchen were also important factors when considering ARI prevalence. However, no information on ventilation was collected in the AHS survey. This study found that the use of crop residue, dung cake, and other fuels for cooking is a significant predictor of ARIs in children, and this is consistent with the findings of other studies.8,16,17 The results show that place of cooking is also a predictor of ARI among children. Houses with no kitchen or having an out-house kitchen or no out-house kitchen reported a significantly higher ARI prevalence. In agreement with pre- vious studies,14,16 household pollution and availability of a kitchen facility are significantly associated with the preva- lence of ARI. Cooking fuels in households that lack a kitchen emit more indoor pollutants, resulting in a more severe impact on the child's health.21 Kerosene and other oils used for lighting were also found to be one of the predictors of ARI because of the emissions released by them, which directly impact on the health of the child.8 The study finding shows ARI prevalence significantly decreases with increasing age of the child, which could be a result of older children having more developed immune systems and playing outside the home more often. The findings of this study show that the type of kitchen facility in dwellings, together with the type of fuels used, may be better predictors of ARI prevalence. A recent study by Babatola et al.24 dealt with the use of briquettes (production from sawdust) as an alternative source of household fuel. The results show that the use of crop residue and dung cake played a significant role in causing ARIs. This study included the lighting sources that have been associated with ARIs. A strength of this study is that it considers comprehensive district-level data collected by the AHS to examine the asso- ciation of household environment and pollutant with preva- lence of ARI among children. The field staff for the survey were trained on the National Family Health Survey and the District Level Household Survey for collecting information on specific symptoms of ARI. However, there are several limita- tions; first, we could not measure the role of ventilation in exposure to pollutants because of the lack of information on this aspect in the survey. This is significant because wall and roof permeability are known to reduce pollution.23 Second, tobacco smoking was not taken into consideration, even though it is a well-known risk factor contributing to ARIs.25 Also, other factors, such as the availability of household members,21 hand washing, and breastfeeding,20 were not examined in this study. Nevertheless, another strong point of this study is that it includes information on lighting sources in the household environment, which is an important contrib- utor to ARI in Uttar Pradesh. In conclusion, the Empowered Action Group state of Uttar Pradesh has a large number of households that do not have access to clean fuels, such as LPG/NPG/electricity/biogas, for cooking or lighting purposes. These households are forced to use solid fuels, such as firewood, crop residue, dung cake, and other fuels, that are known to cause higher emissions than clean fuels. These emissions have long-term health conse- quences, such as a long-lasting cough, respiratory infections, and lung diseases for those who are directly exposed to the emissions. According to World Health Organization report on https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Table 3 e Odds ratios (ORs) for acute respiratory infections (ARIs) with characteristics of children aged 0e59 months in Uttar Pradesh, 2013. Characteristics OR (95% CI) Model 1a Model 2b Model 3c Cooking fuels LPG/PNG/electricity/biogas Firewood 0.86***(0.83e0.88) 0.71***(0.69e0.73) 0.78***(0.75e0.80) Crop residue 1.21***(1.17e1.25) 0.97 (0.93e1.01) 0.94***(0.90e0.98) Dung cake 1.33***(1.29e1.36) 1.09***(1.06e1.12) 1.21***(1.17e1.25) Othersd 1.16***(1.06e1.28) 1.08 (0.98e1.18) 1.06 (0.96e1.16) Lighting source in house Electricity and solar Kerosene and other oils 1.15***(1.13e1.17) 1.07***(1.05e1.10) Otherse 0.97 (0.90e1.05) 0.91**(0.84e0.99) Cooking Place In house: kitchen In house: No kitchen 1.22***(1.19e1.24) 1.26***(1.23e1.29) Out house: kitchen 1.35***(1.30e1.40) 1.19***(1.15e1.24) Out house: no kitchen 1.40***(1.36e1.44) 1.40***(1.35e1.44) Sex Male Female 0.97***(0.96e0.99) Age group (months) <12 12e23 0.72***(0.70e0.74) 24e35 0.75***(0.73e0.77) 36e47 0.72***(0.70e0.74) 48e59 0.64***(0.62e0.66) Caste categories Other than SC/ST SC/ST 0.94***(0.92e0.96) Religion Hindu Muslim 1.09***(1.06e1.11) Othersf 1.02 (0.89e1.16) Wealth index Poorest Poor 1.01 (0.98e1.03) Middle 1 (0.97e1.02) Rich 0.96***(0.93e0.98) Richest 1.03*(1.00e1.07) Place of residence Rural Urban 1.28***(1.24e1.32) Regional categories Eastern region Southern upper Ganga plane 0.55***(0.54e0.56) Northern upper Ganga plane 0.66***(0.64e0.67) Southern region 0.11***(0.11e0.12) Central region 0.20***(0.19e0.22) Constant 0.09***(0.09e0.09) 0.08***(0.08e0.08) 0.14***(0.14e0.15) @ Reference Category ***P < 0.01, **P < 0.05, *P < 0.1. CI, confidence interval; LPG, liquid petroleum gas; PNG, piped natural gas; SC/ST, scheduled castes/scheduled tribes. a Model 1: including only cooking fuels. b Model 2: including household environment, pollutant (cooking fuel and lighting sources) and place of cooking. c Model 3: adjusted for household environment, pollutant, and place of cooking and socio-economic and demographic characteristics. d Includes coal, lignite, charcoal, kerosene, and any other source. e Includes other oils and any other sources for lighting sources. f Includes Christian, Sikh, Buddhist, Jain, and other religion group. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 866 public health and environment, smoke inhaled by women and children from unclean fuels is equivalent to burning 400 cig- arettes in an hour. The Rajiv Gandhi Gramin LPG Vitaran Yojana (2009) and the Pradhan Mantri Ujjwala Yojana (PMUJ- 2016) projects were started to provide LPG gas connections to the socially and economically disadvantaged or marginalized populations. The aim of these initiatives is to enhance the health of women and children by providing them with access https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 8 67 to clean cooking fuels, so they would not have to work in smoky kitchens and wander in unsafe areas to collect fire- wood.22 These steps are significant in achieving the goal of access and use of clean fuels. This study helps to understand the prevalence rate of ARI and its determinants. The knowl- edge and information related to the risk for ARI should be spread across individual, household, as well as community levels, to reduce the prevalence of ARI in Uttar Pradesh. Author statements Ethical approval The study used data sets that are available in the public domain; thus, there was no requirement to seek ethical consent. Funding This research work did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Competing interests None. Author contributions Surendra Kumar Patel conceptualized the study and con- ducted literature review. Surendra Kumar Patel and Sunita Patel conceived and designed the experiments and conducted the statistical analysis. Surendra Kumar Patel performed the experiments, analyzed the data, and wrote the first draft of the manuscript. Sunita Patel, Surendra Kumar Patel contrib- uted to writing of final paper. All authors contributed to and have approved the final manuscript. r e f e r e n c e s 1. IEA, energy access outlook: 2017 from poverty to prosperity. 2017. Paris. Available from: https://webstore.iea.org/download/ summary/274?fileName¼English-Energy-Access-Outlook- 2017-ES . 2. Aayog Niti. Distribution of Households by type of fuel used for cooking. In: National Institution for Transforming India (NITI). Government of India; 2017. Available from: http://niti.gov.in/ content/distribution-households-type-fuel-used-cooking. 3. Munroe RL, Gauvain M. Exposure to open-fire cooking and cognitive performance in children. Int J Environ Health Res 2012;22(2):156e64. Available from: https://www.ncbi.nlm.nih. gov/pubmed/?term¼Exposureþtoþopen-fireþcookingþandþ cognitiveþperformanceþinþchildren. 4. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair- Rohani H, et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012;380(9859):2224e60. 5. Krishnan SS, Amarchand R, Gupta V, Lafond KE, Sulaiankatchi RA, Saha S. Epidemiology of acute respiratory infections in children - preliminary results of a cohort in a rural north Indian community. BMC Infect Dis 2015;15:462. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/ PMC4624162/. 6. Tambe MP, Shivaram C, Chandrashekhar Y. Acute respiratory infection in children: a survey in the rural community. Indian J Med Sci 1999;53(6):249e53. Available from: https://www.ncbi. nlm.nih.gov/pubmed/10776505. 7. MoSPI. Millennium development goals India country report 2015. 2015 [New Delhi]. 8. Fullerton DG, Bruce N, Gordon SB. Indoor air pollution from biomass fuel smoke is a major health concern in the developing world. Trans R Soc Trop Med Hyg 2008;102(9):843e51. 9. WHO. Summary of results: burden of disease from household air pollution for 2012. 2014. Geneva, Switzerland. Available from: http://www.who.int/phe/health_topics/outdoorair/databases/ FINAL_HAP_AAP_BoD_24March2014 . 10. Gurley ES, Homaira N, Salje H, Ram PK, Haque R, Petri W, et al. Indoor exposure to particulate matter and the incidence of acute lower respiratory infections among children: a birth cohort study in urban Bangladesh. Indoor Air 2013;23(5):379e86. Available from: https://www.ncbi.nlm.nih. gov/pmc/articles/PMC3773273/. 11. Gurley ES, Salje H, Homaira N, Ram PK, Haque R, Petri W, et al. Indoor exposure to particulate matter and age at first acute lower respiratory infection in a low-income urban community in Bangladesh. Am J Epidemiol 2014;179(8):967e73. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/ PMC3966716/. 12. Shibata T, Wilson JL, Watson LM, LeDuc A, Meng C, Ansariadi, et al. Childhood acute respiratory infections and household environment in an Eastern Indonesian urban setting. Int J Environ Res Publ Health 2014;11(12):12190e203. Available from: https://www.ncbi.nlm.nih.gov/pubmed/25429685. 13. Chen BH, Hong CJ, Pandey MR, Smith KR. Indoor air pollution in developing countries. Wid hlth statist. quart; 1990. p. 43. Available from: https://pdfs.semanticscholar. org/283e/f1d09b88b59fae2d126611ee8de41e89c6c0 . 14. Ezzati M, Kammen D. Indoor air pollution from biomass combustion and acute respiratory infections in Kenya: an exposure-response study. Lancet 2001;358(9282):619e24. Available from: https://www.ncbi.nlm.nih.gov/pubmed/ 11530148. 15. Bruce N, Perez-Padilla R, Albalak R. Indoor air pollution in developing countries: a major environmental and public health challenge. Bull World Health Organ 2000;78(9):1078e92. Available from: http://www.who.int/bulletin/archives/78(9) 1078 . 16. Buchner H, Rehfuess EA. Cooking and season as risk factors for acute lower respiratory infections in African children: a cross-sectional multi-country analysis. PLoS One 2015;10(6):e0128933. 17. Naz S, Page A, Agho KE. Household air pollution from use of cooking fuel and under-five mortality: the role of breastfeeding status and kitchen location in Pakistan. PLoS One 2017;12(3):e0173256. Available from: http://journals.plos. org/plosone/article?id¼10.1371/journal.pone.0173256. 18. Kumar SG, Majumdar A, Kumar V, Naik BN, Selvaraj K, Balajee K. Prevalence of acute respiratory infection among under-five children in urban and rural areas of puducherry, India. J Nat Sci Biol Med 2015;6(1):3e6. 19. Prajapati B, Talsania N, Sonaliya KN. A study on prevalence of Acute Respiratory Tract Infections (ARI) in under five children in urban and rural communities of Ahmedabad District, Gujarat. Nat J Commun Med 2011;2(2):255e9. https://webstore.iea.org/download/summary/274?fileName=English-Energy-Access-Outlook-2017-ES https://webstore.iea.org/download/summary/274?fileName=English-Energy-Access-Outlook-2017-ES https://webstore.iea.org/download/summary/274?fileName=English-Energy-Access-Outlook-2017-ES https://webstore.iea.org/download/summary/274?fileName=English-Energy-Access-Outlook-2017-ES http://niti.gov.in/content/distribution-households-type-fuel-used-cooking http://niti.gov.in/content/distribution-households-type-fuel-used-cooking https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children https://www.ncbi.nlm.nih.gov/pubmed/?term=Exposure+to+open-fire+cooking+and+cognitive+performance+in+children http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref4 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4624162/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4624162/ https://www.ncbi.nlm.nih.gov/pubmed/10776505 https://www.ncbi.nlm.nih.gov/pubmed/10776505 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref7 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref7 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref8 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref8 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref8 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref8 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref8 http://www.who.int/phe/health_topics/outdoorair/databases/FINAL_HAP_AAP_BoD_24March2014 http://www.who.int/phe/health_topics/outdoorair/databases/FINAL_HAP_AAP_BoD_24March2014 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3773273/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3773273/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3966716/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3966716/ https://www.ncbi.nlm.nih.gov/pubmed/25429685 https://pdfs.semanticscholar.org/283e/f1d09b88b59fae2d126611ee8de41e89c6c0 https://pdfs.semanticscholar.org/283e/f1d09b88b59fae2d126611ee8de41e89c6c0 https://www.ncbi.nlm.nih.gov/pubmed/11530148 https://www.ncbi.nlm.nih.gov/pubmed/11530148 http://www.who.int/bulletin/archives/78(9)1078 http://www.who.int/bulletin/archives/78(9)1078 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref16 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref16 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref16 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref16 http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0173256 http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0173256 http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0173256 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref18 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref18 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref18 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref18 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref18 https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 5 9 e6 868 Available from: https://pdfs.semanticscholar.org/53ed/ 11bf093bb92e01bf247dfa8f6d797b0fdc7c . 20. Selvaraj K, Chinnakali P, Majumdar A, Krishnan IS. Acute respiratory infections among under-5 children in India: a situational analysis. J Nat Sci Biol Med 2014;5(1):15e20. 21. Upadhyay AK, Singh A, Kumar K, Singh A. Impact of indoor air pollution from the use of solid fuels on the incidence of life threatening respiratory illnesses in children in India. BMC Public Health 2015;15:300. Available from: https:// bmcpublichealth.biomedcentral.com/track/pdf/10.1186/ s12889-015-1631-7. 22. Ahmad N, Sharma S, Singh AK. Pradhan Mantri Ujjwala Yojana (PMUY) step towards social inclusion in India. International Journal of Trend in Research and Development 2018;5(1). Available from: http://www.ijtrd.com/papers/ IJTRD14650 . 23. Dashupta S, Huq M, Khaliquzzaman M, Pandey K, Wheeler D. Indoor air quality for poor families: new evidence from Bangladesh. Indoor Air 2006;16(6):426e44. Available from: https://www.ncbi.nlm.nih.gov/pubmed/17100664. 24. Babatola JO, Lasisi KH, Ajibade FO. Production of Sawdust Briquettes as Alternative household fuel using water and cow dung as Binders. Afr J Renew Altern Energy 2017;2(3):51e8. Available from: https://www.ncbi.nlm.nih.gov/pubmed/ 22721493. 25. DiFranza JR, Masaquel A, Barrett AM, Colosia AD, Mahadevia PJ. Systematic literature review assessing tobacco smoke exposure as a risk factor for serious respiratory syncytial virus disease among infants and young children. BMC Pediatr 2012;12:81. Available from: https://www.ncbi. nlm.nih.gov/pubmed/22721493. https://pdfs.semanticscholar.org/53ed/11bf093bb92e01bf247dfa8f6d797b0fdc7c https://pdfs.semanticscholar.org/53ed/11bf093bb92e01bf247dfa8f6d797b0fdc7c http://refhub.elsevier.com/S0033-3506(19)30003-4/sref20 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref20 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref20 http://refhub.elsevier.com/S0033-3506(19)30003-4/sref20 https://bmcpublichealth.biomedcentral.com/track/pdf/10.1186/s12889-015-1631-7 https://bmcpublichealth.biomedcentral.com/track/pdf/10.1186/s12889-015-1631-7 https://bmcpublichealth.biomedcentral.com/track/pdf/10.1186/s12889-015-1631-7 http://www.ijtrd.com/papers/IJTRD14650 http://www.ijtrd.com/papers/IJTRD14650 https://www.ncbi.nlm.nih.gov/pubmed/17100664 https://www.ncbi.nlm.nih.gov/pubmed/22721493 https://www.ncbi.nlm.nih.gov/pubmed/22721493 https://www.ncbi.nlm.nih.gov/pubmed/22721493 https://www.ncbi.nlm.nih.gov/pubmed/22721493 https://doi.org/10.1016/j.puhe.2019.01.003 https://doi.org/10.1016/j.puhe.2019.01.003 Effects of cooking fuel sources on the respiratory health of children: evidence from the Annual Health Survey, Uttar Prades ... Introduction Methods Variables used in the analysis Outcome variable Predictor variables Socio-economic and demographic characteristics Statistical analysis Results Discussion Author statements Ethical approval Funding Competing interests Author contributions References The-cost-effectiveness-of-public-health-interventions-examined-b_2019_Public ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2 Available online at w Public Health journal homepage: www.elsevier.com/puhe Themed Papere Original Research The cost-effectiveness of public health interventions examined by the National Institute for Health and Care Excellence from 2005 to 2018*,** L. Owen a,*, A. Fischer b a Centre for Guidelines, National Institute for Health and Care Excellence, London, WC1V 6NA, UK b Office of Health Economics, Southside, 105 Victoria Street, London, SW1E 6QT, UK a r t i c l e i n f o Article history: Received 27 April 2018 Received in revised form 24 October 2018 Accepted 4 February 2019 Available online 16 March 2019 Keywords: Public Health Cost Effectiveness Health Economics Cost utility analysis * The work was undertaken in the autho available on the NICE website. ** The views expressed in this article are th for. * Corresponding author. National Institute fo þ44 020 7045 2145. E-mail address: Lesley.owen@nice.org.uk https://doi.org/10.1016/j.puhe.2019.02.011 0033-3506/Crown Copyright © 2019 Publishe a b s t r a c t Background: Reviews of economic evaluations of public health (PH) interventions assessed by the National Institute for Health and Care Excellence (NICE) in the periods 2005e2010 and 2011e2016 have been undertaken. This study combines these analyses, adds six further guidelines published since then, and thus provides a summary of cost- effectiveness of NICE's PH interventions to the present. Methods: As in previous studies, economic evaluations carried out between 2005 and 2018 were categorised by the type of economic analysis used to extract and summarise base-case ICERs. A number of ‘sensitivity analyses’ were carried out to test the validity of the approach. Results: Of 71 guidelines examined, 27 used cost utility analysis (CUA) for specific in- terventions, yielding 380 individual base-case ICER estimates (or 221 taking into account clustering of interventions). The median cost per quality-adjusted life-year (QALY) ICER for the 380 estimates was £1,986. Of these, 21% were cost saving, and 54% ranged from £1 to £20,000, 3% were between £20,001 and £30,000, 16% were above £30,000 and 5% were dominated. Taking clustering into account made relatively little difference to these results. Reducing the threshold from £20,000/QALY to £15,000/QALY would result in 2% of ICERs moving across the threshold. Conclusions: Seventy-five percent of PH interventions assessed were cost-effective at a threshold of £20,000 per QALY when disregarding clustering, and 68% were cost-effective when clusters were represented by a single ICER. Other analyses gave similar results for the distribution of ICERs. Limitations of the analysis are discussed. Crown Copyright © 2019 Published by Elsevier Ltd on behalf of The Royal Society for Public Health. All rights reserved. rs' own time using data extracted from economic analyses for public health guidelines ose of the authors and do not necessarily reflect the views of the organisations they work r Health and Care Excellence, 10 Spring Gardens, St. James's, London, SW1A 2BU, UK. Tel.: (L. Owen). d by Elsevier Ltd on behalf of The Royal Society for Public Health. All rights reserved. mailto:Lesley.owen@nice.org.uk http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.02.011&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.02.011 https://doi.org/10.1016/j.puhe.2019.02.011 https://doi.org/10.1016/j.puhe.2019.02.011 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2152 Introduction In 2011, Owen et al.1 analysed 200 base-case cost-effective- ness estimates of public health (PH) interventions considered in 21 PH guidelines developed by the National Institute for Health and Care Excellence (NICE). Most PH interventions assessed by NICE was estimated to be cost-effective. In 2017, Owen et al.2 updated this information for 2012 to 2016. They considered 138 further base-case cost-effectiveness esti- mates. A somewhat smaller majority was estimated to be cost-effective, but a slightly larger proportion was cost saving.2 The funding for PH interventions changed in 2013 because of the Health and Social Care Act 2012. Local authorities, which are now responsible for improving PH and reducing health inequalities,3 still require economic analyses and evi- dence of cost-effectiveness to secure funding.4 The process for selecting topics for PH guideline develop- ment by NICE, together with the methodology used by NICE in assessing the cost-effectiveness of interventions, has been outlined in the study by Owen et al.2 Recent changes impinging on the way that the past PH interventions will be viewed in future are (a) the Department of Health (DH) has introduced a threshold ICER of £15,000 per QALY for its impact assessments, lower than the £20,000 used currently by NICE;5,6 (b) the DH has reduced the rate of discounting future health costs and benefits to 1.5% from 3.5% in the estimation of ICERs;7 and (c) the role of the precautionary principle in appraising the prevention of ill health has been strengthened.8 This study gives a consolidated summary of all the esti- mates of cost-effectiveness for PH interventions published by NICE until March 2018 and explores a number of ‘sensitivity analyses’ of the distribution of ICERs. Methods We follow the methods outlined in the study by Owen et al.,2 augmented as follows. We include a sixth category for analysing ICERs to examine the impact of using a lower threshold of £15,000/QALY for cost-effectiveness. We examine the effect of providing a single estimate of cost-effectiveness for an intervention. Clusters of ICERs for an intervention occur because either the results have been re- ported for subgroups of the population (e.g. by age and gender) or the intervention has been assessed in multiple guidelines. Clustered ICERs are not independent of each other; treating them separately, as in the previous analyses and carried through to the combined period analysis, will overstate the importance of the intervention by counting it a number of times. In addition, we excluded two of the eight ICERs for transition support services for looked after children (Guideline PH28). The source study for these indicated the intervention was not effective, besides which the two arms of the study were not randomised. We also study the impact of the intervention's comparator. Suppose that currently there is no intervention being applied to a PH problem. If intervention A is tested, the effectiveness and cost-effectiveness for A should use ‘no intervention’ (also sometimes called ‘placebo’) as its comparator. If A is found to be both effective and cost effective with respect to placebo, and intervention B is now tested, B's comparator could be either A or placebo. The cost-effectiveness of B will depend critically on which one is used. If B is to replace A, then the comparator should be A, but if A does not work for a patient, then in that case, B's comparator should be placebo. In smoking cessation, if a person is unable to quit using one intervention, then either another intervention or placebo could be the appropriate comparator. In the analyses by Owen et al.,1,2 no account was taken of which comparator was chosen, and in some cases, several comparators for a particular intervention were used. We have dealt with this in two ways. The first is to divide the population of ICERs into three categories: (i) the absence of an intervention; (ii) ‘usual treatment’ which might be the lack of any intervention or alternatively an intervention, the na- ture of which has not been stated; and (iii) a known inter- vention. The second way uses a case study to illustrate the impact of using different comparators. Results There were 71 PH guidelines were published by NICE between March 2006 and March 2018. Twenty-seven of these used CUA for specific interventions, yielding 380 base-case ICER esti- mates; 12 used CUA for a threshold or ‘what if’ analysis, one used a threshold and cost effectiveness analysis (CEA), one used CEA, three used cost consequences analysis (CCA), two used cost benefit analysis (CBA) and CUA and one used CEA and CUA, and seven did not require economic modelling. CUA base-case ICERs The median cost per QALY ICER for the 380 estimates was £1,986. Of these, 21% were cost saving, 54% ranged from £1 to £20,000, 3% were between £20,001 and £30,000, 16% were above £30,000 and 5% were dominated (Table 1). Lowering the threshold to £15,000/QALY would result in nine fewer ICERs being considered cost-effective (2% of the total sample). In total 221 ICERs remained after clustering were taken into account. The reduction was mostly due to a decrease in the number of smoking cessation ICERs, which accounted for 49% before and 36% after declustering. The median for the 221 ICERs increased somewhat to £3,629/QALY, but the percentage that would be considered not cost-effective at the lower threshold of £15,000/QALY remained the same (Table 2). https://doi.org/10.1016/j.puhe.2019.02.011 https://doi.org/10.1016/j.puhe.2019.02.011 Table 1 e Number (%) and median values of base-case estimated incremental cost per QALY for public health interventions assessed and published by NICE between March 2006 and March 2018. ICERs Cost saving £1 e£15,000 £15,001 e£20,000 £20,001 e£30,000 >£30,000 Intervention was
dominated

Overall

Number (%) 81 (21) 197 (52) 9 (2) 12 (3) 62 (16) 19 (5) 380 (100)

Median NA £1,648 £15,962 £25,175 £179,008 NA £1,986

Interquartile

range

NA £437e£4,833 £15,192e£18,000 £23,023e£26,542 £63,446e£350,817 NA £87e£18,175

NA, not available; NICE, National Institute for Health and Care Excellence; QALY, quality-adjusted life year.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2 153
A summary of the interventions and associated ICERs is

shown in Table 3.

The effect of clustering is illustrated in Fig. 1. It shows a

slight increase in the proportion of ICERs that are cost saving,

a small decrease in the proportion that fall between £1/QALY

and £15,000/QALY and a relatively large increase in the pro-

portion that exceed £30,000 per QALY offset in part by a

reduction in the proportion dominated.

Analysis of comparators
Table 4 shows that the group comparing interventions with

other interventions had more interventions that were cost

saving (37%) than the other two groups (22% and 19%, usual care

and on intervention, respectively); it also had more interventions

that were dominated (27%) than those compared against usual

care (3%) or no intervention (3%) (c2 ¼ 45.6, P < 0.01). Case study of comparators: smoking cessation interventions Table 5 shows the impact of a comparator in assessing the cost-effectiveness of an intervention. In this example, the ICER for combination patch and nasal spray is substantially lower than £20,000/QALY, irrespective of the type of compar- ator used to assess the intervention. The same applies to the assessment of the combination of bupropion and lozenge. In contrast, the ICER for bupropion is dominant when compared with cognitive behavioural therapy (CBT) but dominated when compared with nicotine replacement therapy (NRT). We identified other problems that had previously been noted in comparing the cost-effectiveness of PH in- terventions; these include updates of the evidence, assump- tions around duration of effect, other model assumptions and the length of the time horizon. For example, the ICERs for brief advice for smoking cessation were between £0 and £20,000/ QALY in PH1 but cost saving in subsequent guidelines. The change reflects differences in the effect sizes, costs and the assumption about the background quit rate (the comparator). Perhaps only the most recent estimates, which better reflect the current context, assuming all other things are equal, should be used in assessment of cost-effectiveness. Table 2 e Number (%) and median values of base-case estimate assessed and published by NICE between March 2006 and Mar ICERs Cost saving £1 e£15,000 £15,001 e£20,000 £ e Number (%) 52 (24) 93 (42) 5 (2) Median NA £2,465 £15,962 Interquartile range NA £765e£5,278 £15,192e£16,859 £23,5 Discussion Main finding of this study This article summarises the cost-effectiveness estimates that have come in evidence before the NICE PH guidelines com- mittee from its first guideline, published in 2006, until the most recent, published in March 2018. It examines some weaknesses in the previous work, but the changes this has made have little material impact on the results. The median ICER has increased somewhat, but crucially, most ICERs (68%) are still below £20,000 per QALY. Clustering of ICERs Many ICERs reported in the evidence on cost-effectiveness for NICE PH were carried out on subgroups of the population, such as the cost-effectiveness for an intervention for smoking cessation, with a separate ICER for men and women, and for each of a number ofage ranges. The reason for this isto seeif the intervention is cost-effective for each subgroup or only for some. If the differences between the subgroups are small enough, each subgroup could be considered to be part of a meta- analysis and lumped together into a single, more powerful analysis with many more degrees of freedom. Assuming the cost per person was also similar across the subgroups, the combined ICER with a much greater weight than that of each of its component subgroups should be used. Working backwards from this point would imply that the analysis we have under- taken should weight all the ICERs we have used according to the uncertainty with which the ICER is associated to determine the median ICER for the whole population of NICE PH ICERs. How- ever, this uncertainty also depends on the variability of costs and the unknown correlation between individual costs and health benefits, so weighting could not easily be carried out. The subgroups for which there is a separate ICER may have different intervention effects (assuming similar cost distri- butions for each of the subgroups), so that a meta-analysis with fixed effects is not possible, or if the heterogeneity is d incremental cost per QALY for public health interventions ch 2018 excluding clusters. 20,001 £30,000 >£30,000 Intervention
was dominated

Overall

7 (3) 60 (27) 4 (2) 221 (100)

£25,199 £179,008 NA £3,620

23e£27,568 £65,112e£336,272 NA £87 e £42,145

https://doi.org/10.1016/j.puhe.2019.02.011

Table 3 e The value of incremental cost-effectiveness estimates for public health interventions assessed and published by NICE between 2006 and 2018.

Guideline topic and ID Intervention Comparator ICER

NG30: Oral health promotion: general dental practice One-to-one counselling to parents of children aged

5 years for high-risk caries in socio-economically

deprived areas in Northwest England

Usual care Dominant

NG30: Oral health promotion: general dental practice Dental hygienists OH prog for children aged

12 years at high risk

Usual care Dominant

NG30: Oral health promotion: general dental practice Dental hygienists OH prog for children aged

12 years at average risk

Usual care £ 14,408

NG30: Oral health promotion: general dental practice One-to-one counselling to parents of children aged

5 years for average risk caries in socio-economically

deprived areas in Northwest England

Usual care £ 99,826

NG32: Older people: independence and mental wellbeing Friendship programme No intervention (waiting list) Dominant

NG32: Older people: independence and mental wellbeing Internet and computer training No intervention (waiting list) £ 15,962

NG34: Sunlight exposure: risks and benefits Mass media No intervention Dominant

NG34: Sunlight exposure: risks and benefits Tailored message No intervention £ 16,859

NG34: Sunlight exposure: risks and benefits Text messages No intervention £ 65,945

NG34: Sunlight exposure: risks and benefits Living with the sun No intervention £ 312,744

NG34: Sunlight exposure: risks and benefits Photoageing No intervention £ 316,968

NG55: Harmful sexual behaviour Multisystemic therapy for problem sexual

behaviours

Cognitive behavioural therapy dominant

NG55: Harmful sexual behaviour Cognitive behavioural therapy Play therapy £ 2,685

NG6: Cold homes Energy efficiency COPD No intervention £ 28,324

NG6: Cold homes Energy efficiency þ fuel subsidy COPD No intervention £ 33,771
NG6: Cold homes Fuel subsidy COPD No intervention £ 39,437

NG6: Cold homes Energy efficiency HD No intervention £ 157,137

NG6: Cold homes Energy efficiency age 65þ years No intervention £ 157,661
NG6: Cold homes Energy efficiency þ fuel subsidy HD No intervention £ 174,467
NG6: Cold homes Energy efficiency þ fuel subsidy age 65þ years No intervention £ 180,456
NG6: Cold homes Fuel subsidy HD No intervention £ 188,301

NG6: Cold homes Fuel subsidy age 65þ years No intervention £ 204,076
NG6: Cold homes Energy efficiency low income No intervention £ 275,896

NG6: Cold homes Energy efficiency þ fuel subsidy low income No intervention £ 317,927
NG6: Cold homes Fuel subsidy low income No intervention £ 358,089

NG6: Cold homes Energy efficiency CMD No intervention £ 394,556

NG6: Cold homes Energy efficiency þ fuel subsidy CMD No intervention £ 452,154
NG6: Cold homes Fuel subsidy CMD No intervention £ 509,205

NG64: Drug misuse – targeted interventions Family intervention called STRIVE Standard care £ 117,000

NG64: Drug misuse – targeted interventions Motivational interviewing intervention to reduce

club drug use and HIV risk behaviours among men

who have sex with men

Educational control £ 131,000

NG64: Drug misuse – targeted interventions Familias Unidas No intervention £ 241,000

NG64: Drug misuse – targeted interventions Motivational interviewing to reduce drug use in

young gay and bisexual men

Content matched education £ 301,000

NG64: Drug misuse – targeted interventions Brief, web-based personalised feedback No intervention £ 329,000

NG64: Drug misuse – targeted interventions Motivational interviewing to reduce ecstasy use Assessment only 3-month delay

treatment

£ 485,000

NG64: Drug misuse – targeted interventions Focus on families No intervention £ 99,000,000

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
4

https://doi.org/10.1016/j.puhe.2019.02.011

NG70: Outdoor air Street washing and sweeping No intervention £ 441

NG70: Outdoor air Speed restrictions No intervention £ 1,293

NG70: Outdoor air Vehicle idling No intervention £ 1,572

NG70: Outdoor air Low-emission zones No intervention £ 2,465

NG70: Outdoor air Off-road cycle paths No intervention (on road cycle) £ 5,075

NG70: Outdoor air Bypass construction No intervention £ 6,971

NG70: Outdoor air Motorway barriers No intervention £ 25,199

NG90: Physical activity Active Living by Design No intervention (before, after and no

control)

£ 1,397

NG90: Physical activity Cycling demonstration towns Placebo (matched town) £ 2,496

NG90: Physical activity Smarter Choices, Smarter Places Placebo (matched control) £ 4,423

NG90: Physical activity Connswater Community Greenway Placebo (control group) £ 7,652

NG90: Physical activity Park renovations Placebo (before, after and control) £ 215,989

NG92: Smoking cessation interventions and services Sequence (varenicline, bupropion, SSRI) ¼ 40.30% No intervention ¼ 2% Dominant
NG92: Smoking cessation interventions and services Bupropion and lozenge ¼ 25.60% No intervention ¼ 2% Dominant
NG92: Smoking cessation interventions and services Lozenge ¼ 14.38% No intervention ¼ 2% Dominant
NG92: Smoking cessation interventions and services Patch only ¼ 11.00% No intervention ¼ 2.00% Dominant
NG92: Smoking cessation interventions and services Varenicline þ brief advice ¼ 25.00% Brief advice ¼ 6.60% Dominant
NG92: Smoking cessation interventions and services Bupropion (PP) ¼ 47.33% CBT (PP) ¼ 38.20% Dominant
NG92: Smoking cessation interventions and services Bupropion and lozenge ¼ 25.60% Lozenge ¼ 14.38% Dominant
NG92: Smoking cessation interventions and services Bupropion (PP) ¼ 47.33% (pt prev) Minimal intervention (PP) ¼ 33.66% Dominant
NG92: Smoking cessation interventions and services NRT (PP) ¼ 41.30% Minimal intervention (PP) ¼ 33.66% Dominant
NG92: Smoking cessation interventions and services Varenicline þ counselling ¼ 27.90% Placebo þ counselling ¼ 15.90% Dominant
NG92: Smoking cessation interventions and services Varenicline þ counselling ¼ 30.50% Placebo þ counselling ¼ 17.30% Dominant
NG92: Smoking cessation interventions and services Varenicline þ counselling ¼ 16.61% Placebo þ counselling ¼ 5.91% Dominant
NG92: Smoking cessation interventions and services Self-determination intervention ¼ 10.10% Standard care ¼ 3.51% Dominant
NG92: Smoking cessation interventions and services Patch and nasal spray ¼ 27.00% No intervention ¼ 2.00% £13
NG92: Smoking cessation interventions and services Patch and nasal spray ¼ 27.00% Patch only ¼ 11.00% £948
NG92: Smoking cessation interventions and services CBT (PP) ¼ 38.20% Minimal intervention (PP) ¼ 33.66% £3,620
NG92: Smoking cessation interventions and services NRT OTC ¼ 8.65% No intervention ¼ 13.19% Dominated
PH2: Physical Activity Intensive interviews Brief advice from researcher at the

baseline assessment

£ 75.06

PH2: Physical Activity Intensive interviews with exercise voucher Brief advice from researcher at the

baseline assessment

£ 432.13

PH3: Sexually transmitted diseases Tailored skill session Usual care, didactic messages £3,200

PH3: Sexually transmitted diseases Information and behaviour skills (women) Usual care (information only delivered by

counsellors in didactic style)

£ 10,286.00

PH3: Sexually transmitted diseases Brief counselling Usual care (didactic messages,

information intervention to approximate

treatment as usual)

£ 12,194.00

PH3: Sexually transmitted diseases Accelerated partner therapy, doxycycline Patient referral £ 12,525.00

PH3: Sexually transmitted diseases Information, motivation and behaviour skills Usual care (information only delivered by

counsellors in didactic style)

£ 14,143.00

PH3: Sexually transmitted diseases Accelerated partner therapy, azithromycin Patient referral £ 19,425.00

PH3: Sexually transmitted diseases Intensive counselling Usual care £ 24,000.00

PH3: Sexually transmitted diseases Enhanced counselling Usual care (didactic messages,

informational intervention designed to

approximate treatment as usual)

£ 45,606.50

(continued on next page)

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
5

https://doi.org/10.1016/j.puhe.2019.02.011

Table 3 e (continued)

Guideline topic and ID Intervention Comparator ICER

PH3: Sexually transmitted diseases Behavioural skills counselling Standard 15-min risk reduction

counselling

£ 96,000.00

PH4: Substance misuse Life skills training Normal education £ 3,492.00

PH4: Substance misuse ‘Say Yes First’ Normal education £ 90,786.00

PH4: Substance misuse Teacher training Usual care (no training, standard

education curriculum)

£ 157,384.00

PH4: Substance misuse Teacher training Usual care (no training, standard

education curriculum)

£ 195,225.00

PH6: Behaviour change Mass media to promote healthy eating No intervention £ 87.00

PH8: Physical activity and the environment Trail No intervention £ 10,445.00

PH10: Smoking cessation Brief advice No intervention Dominant

PH10: Smoking cessation Nicotine patch, pharmacy consultation No intervention Dominant

PH10: Smoking cessation Nicotine patch, pharmacy

consultation þ behavioural programme
No intervention Dominant

PH10: Smoking cessation Brief advice plus self-help material No intervention Dominant

PH10: Smoking cessation Less intensive counselling and bupropion, workplace No intervention Dominant

PH10: Smoking cessation More intensive counselling and bupropion,

workplace

No intervention Dominant

PH10: Smoking cessation Nicotine patch, group counselling No intervention Dominant

PH10: Smoking cessation Nicotine patch, individual counselling No intervention Dominant

PH10: Smoking cessation Brief advice plus self-help material plus NRT No intervention £ 984.00

PH12: Social and emotional wellbeing in primary schools Universal intervention, emotion þ cognition No intervention £5,278.00
PH12: Social and emotional wellbeing in primary schools Universal, emotion only No intervention £10,594.00

PH12: Social and emotional wellbeing in primary schools Focused intervention, emotion þ cognition No intervention £ 177,560.00
PH12: Social and emotional wellbeing in primary schools Focused intervention, emotion only No intervention £ 988,404.00

PH13: Physical activity in the workplace Walking programme No intervention £ 686.34

PH13: Physical activity in the workplace Counselling Usual care (control group no details in

abstract)

£ 864.50

PH14: Preventing the uptake of smoking by children and young people Mass media No intervention £ 49.00

PH14: Preventing the uptake of smoking by children and young people Point of sale No intervention £ 1,690.00

PH15: Risk of dying prematurely – smoking cessation general population Drop-in/rolling community-based sessions No intervention (background quit rate) £ 91.00

PH15: Risk of dying prematurely – smoking cessation general population Client-centred approaches No intervention £ 93.00

PH15: Risk of dying prematurely – smoking cessation general population Proactive telephone counselling (Ockene et al.

review)

Usual care or intervention but no

telephone counselling

£ 195.00

PH15: Risk of dying prematurely – smoking cessation general population Dentist-based interventions Usual care £ 331.00

PH15: Risk of dying prematurely – smoking cessation general population Quit and win DK, most in Bains review £ 342.00

PH15: Risk of dying prematurely – smoking cessation general population Pharmacist-based interventions Usual care £ 546.50

PH15: Risk of dying prematurely – smoking cessation general population Proactive telephone counselling Usual care or intervention but no

telephone counselling

£ 568.00

PH15: Risk of dying prematurely – smoking cessation general population Identify smokers through other means No intervention £ 644.00

PH15: Risk of dying prematurely – smoking cessation general population Incentive NRT No intervention £ 671.00

PH15: Risk of dying prematurely – smoking cessation general population Workplace intervention (WI) No intervention £ 1,399.00

PH15: Risk of dying prematurely – smoking cessation general population Client-centred social marketing No intervention £ 1,564.00

PH15: Risk of dying prematurely – smoking cessation general population Incentive, NRT prescription No intervention £ 1,627.00

PH15: Risk of dying prematurely – smoking cessation general population Incentive workplace Usual care (WI with no incentive) £ 2,089.00

PH15: Risk of dying prematurely – smoking cessation general population Identifying and reaching Usual care £ 2,535.00

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
6

https://doi.org/10.1016/j.puhe.2019.02.011

PH15: Risk of dying prematurely – smoking cessation general population Pharmacist-based interventions No intervention £ 3,151.00

PH15: Risk of dying prematurely – smoking cessation general population Brief intervention, pregnant women Usual care £ 3,792.50

PH15: Risk of dying prematurely – smoking cessation general population Invitation for screening No intervention £ 4,260.00

PH15: Risk of dying prematurely – smoking cessation general population Pharmacist-based interventions No intervention £ 4,892.00

PH16: Mental wellbeing of older people Tri-weekly walking programme after 6 months. Information and education £ 7,400.00

PH16: Mental wellbeing of older people Advice about physical activity Usual care £ 35,900.00

PH16: Mental wellbeing of older people Advice about physical activity Nutrition Dominated

PH17: Promoting physical activity for children and young people Walking buses No intervention £ 4,007.63

PH17: Promoting physical activity for children and young people Dance classes No intervention £ 27,570.06

PH17: Promoting physical activity for children and young people Free swimming No intervention £ 40,461.56

PH17: Promoting physical activity for children and young people Community sports No intervention £ 71,456.21

PH19: Management of long-term sickness and incapacity for work WI Usual care for musculoskeletal disorders Dominant

PH19: Management of long-term sickness and incapacity for work Physical activity and education and workplace visit

(PW)

Usual care for musculoskeletal disorders Dominant

PH19: Management of long-term sickness and incapacity for work Physical activity and education (PA) Usual care for musculoskeletal disorders £ 2,758.00

PH20: Social and emotional wellbeing in secondary education Intervention to reduce bullying No intervention £9,600.00

PH22: Promoting mental wellbeing at work Individual stress management, health coach No intervention £3,470.00

PH22: Promoting mental wellbeing at work Individual stress management, six group sessions No intervention £4,998.00

PH22: Promoting mental wellbeing at work Individual stress management, seven group sessions No intervention £15,031.00

PH23: School based interventions to prevent the uptake of smoking Delay/delay No intervention (or usual education) £7,282.50

PH24: Alcohol use disorders: preventing harmful drinking Screening and brief intervention by practice nurse at

GP registration

No intervention Dominant

PH24: Alcohol use disorders: preventing harmful drinking Screening and brief intervention by GP during

appointment

No intervention Dominant

PH24: Alcohol use disorders: preventing harmful drinking Screening and brief intervention at A&E No intervention £0.00

PH26: Quitting smoking in pregnancy Rewards No intervention (aggregate of controls) Dominant

PH26: Quitting smoking in pregnancy Other No intervention (aggregate of controls) Dominant

PH26: Quitting smoking in pregnancy Feedback No intervention (aggregate of controls) £1,992.00

PH26: Quitting smoking in pregnancy Pharmacotherapies No intervention (aggregate of controls) £2,253.00

PH26: Quitting smoking in pregnancy Stages of change No intervention (aggregate of controls) £3,033.00

PH26: Quitting smoking in pregnancy Cognitive behaviour strategies No intervention (aggregate of controls) £4,005.00

PH27: Weight management in pregnancy Weight management interventions Conventional postnatal care £ 9,096

PH28: Looked after children, Transition support services Georgiades (2005) men Usual care/no intervention Dominant

PH28: Looked after children, Transition support services Georgiades (2005) women Usual care/no intervention Dominant

PH28: Looked after children, Transition support services Lindsey & Ahmed (1999) men Usual care/no intervention Dominant

PH28: Looked after children, Transition support services Lindsey & Ahmed (1999) women Usual care/no intervention Dominant

PH28: Looked after children, Transition support services Scannapieco (1996) men Usual care/no intervention Dominant

PH28: Looked after children, Transition support services Scannapieco (1996) women Usual care/no intervention Dominant

PH30: Unintentional injuries in the home Free smoke alarm programme No intervention £ 23,046

PH31: Unintentional injuries on the road Advisory 20 mph zones No intervention £ 22,952

PH31: Unintentional injuries on the road Mandatory 20 mph zones high casualties No intervention £ 89,700

PH31: Unintentional injuries on the road Mixed priority routes No intervention £ 304,823

PH31: Unintentional injuries on the road Mandatory 20 mph zones low casualties No intervention £ 457,762

PH32: Skin cancer prevention Verbal advice and print to parentsechildren at home

(Turissi)

No intervention (current practice) £ 6,700

PH32: Skin cancer prevention Verbal advice group session, uni students No intervention (current practice) £ 42,000

PH32: Skin cancer prevention Multicomponent community No intervention £ 207,339

(continued on next page)

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
7

https://doi.org/10.1016/j.puhe.2019.02.011

Table 3 e (continued)

Guideline topic and ID Intervention Comparator ICER

PH32: Skin cancer prevention Verbal advice, in school and at home activities

children at school and newsletter (School) Buller

No intervention (current practice) £ 260,000

PH32: Skin cancer prevention Multicomponent community No intervention £ 1,069,469

PH32: Skin cancer prevention Multicomponent work setting 21- to 65-year-olds Delayed intervention £ 1,298,476

PH32: Skin cancer prevention Construction of shade sail No built shade £ 2,394,901

PH32: Skin cancer prevention Multicomponent beach and pool No intervention £ 10,621,954

PH32: Skin cancer prevention Multicomponent education 2e7 years (Bauer) 3-h education £ 32,498,835

PH32: Skin cancer prevention Multicomponent education 13- to 15-year-olds 3-h education £ 50,940,170

PH32: Skin cancer prevention Multicomponent healthcare, 13- to 15-year-olds PA and diet £ 82,264,556

PH33: HIV testing, increasing uptake in black Africans Mass media No intervention £ 27,566

PH33: HIV testing, increasing uptake in black Africans Sport No intervention £ 30,509

PH33: HIV testing, increasing uptake in black Africans Choice of rapid or standard testing No intervention £ 31,333

PH33: HIV testing, increasing uptake in black Africans Music No intervention £ 32,357

PH34: HIV testing, increasing uptake in MSM Opt-out intervention No intervention (no testing) £ 42,145

PH34: HIV testing, increasing uptake in MSM Choice or rapid, oral or standard testing No intervention (no testing) £ 42,632

PH34: HIV testing, increasing uptake in MSM Peer referral intervention No intervention (no testing) £ 50,358

PH34: HIV testing, increasing uptake in MSM Retreat intervention No intervention (no testing) £ 56,285

PH34: HIV testing, increasing uptake in MSM Multicomponent mass media No intervention (no testing) £ 62,613

PH35: Type 2 diabetes: pop and comm Large-scale region-wide multicomponent No intervention Dominant

PH35: Type 2 diabetes: pop and comm Multicomponent small scale No intervention £ 562

PH35: Type 2 diabetes: pop and comm Broad dietary education/cooking skills No intervention £ 878

PH35: Type 2 diabetes: pop and comm New food retail outlet No intervention Dominated

PH38: Type 2 diabetes, S Asians 25-39 LPDS � 5.25, HbA1c � 6.0% (þintensive intervention) Vascular checks (without intervention) £ 11,273
PH38: Type 2 diabetes, high risk LPDS � 4.75, HbA1c � 5.85% (þintensive intervention) Vascular checks (with intervention) £ 15,192
PH40: Social emotional wellbeing early years Sure start, years 1, 3 and 5 No intervention Dominant

PH40: Social emotional wellbeing early years Weekly home visits No intervention £ 85,097

PH41: Physical activity: walking and cycling Multicomponent sustainable travel towns No intervention £ 997

PH41: Physical activity: walking and cycling TravelSmart No intervention £ 1,400

PH41: Physical activity: walking and cycling Pedometer No intervention £ 1,763

PH41: Physical activity: walking and cycling Pedometer sustained No intervention £ 4,774

PH41: Physical activity: walking and cycling Multicomponent cycling demonstration No intervention £ 4,830

PH41: Physical activity: walking and cycling Pedometer 4 week No intervention £ 12,351

PH43: Hep B&C testing GP education and paid targeted testing of ex-IDUs No intervention £ 13,877

PH43: Hep C testing Dried blood spot testing in addiction services No intervention (control not offering DBS,

i.e., do nothing)

£ 14,632

PH43: Hep B&C testing DBS in prison No intervention (control not offering DBS,

i.e., do nothing)

£ 59,418

PH44: Physical activity: brief advice for adults in primary care Brief advice for 1 year Usual care £ 1,730

PH45: Smoking: harm reduction CDTQ þ generic support No intervention Dominant
PH45: Smoking: harm reduction CDTQ No intervention Dominant

PH45: Smoking: harm reduction Reduce No intervention Dominant

PH45: Smoking: harm reduction CDTQ þ specialist support No intervention £ 437
PH45: Smoking: harm reduction CDTQ þ NCP No intervention £ 544
PH45: Smoking: harm reduction CDTQ þ NCP þ generic support No intervention £ 668
PH45: Smoking: harm reduction Temporary abstinence þ generic support No intervention £ 706
PH45: Smoking: harm reduction Reduce þ generic support No intervention £ 706

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
8

https://doi.org/10.1016/j.puhe.2019.02.011

PH45: Smoking: harm reduction Temporary abstinence þ NCP þ generic support No intervention £ 765
PH45: Smoking: harm reduction Reduce þ NCP þ generic support No intervention £ 765
PH45: Smoking: harm reduction CDTQ þ NCP þ specialist support No intervention £ 2,294
PH45: Smoking: harm reduction Temporary abstinence þ NCP þ specialist support No intervention £ 2,458
PH45: Smoking: harm reduction Reduce þ NCP þ specialist support No intervention £ 2,458
PH45: Smoking: harm reduction Abrupt þ NCP substitute þ generic support No intervention £ 3,558
PH45: Smoking: harm reduction Abrupt þ NCP substitute (nb. Source says includes

brief advice)

No intervention £ 7,388

PH45: Smoking: harm reduction Temporary abstinence þ NCP No intervention £ 7,843
PH45: Smoking: harm reduction Reduce þ NCP No intervention £ 7,843
PH45: Smoking: harm reduction Temporary abstinence þ specialist support No intervention £ 8,464
PH45: Smoking: harm reduction Reduce þ specialist support No intervention £ 8,464
PH45: Smoking: harm reduction Temporary abstinence No intervention Dominated

PH48: Smoking cessation secondary care High-intensity behavioural therapy Brief advice Dominant

PH48: Smoking cessation secondary care High intensity behavioural

therapy þ pharmacological therapy
Brief advice/low intensity Dominant

PH48: Smoking cessation secondary care Total smoke-free policy, indoor and outdoor

Gadomski

Indoor smoke-free policy Dominant

PH48: Smoking cessation secondary care Pharmacological for general inpatients Low-intensity behavioural therapy Dominant

PH48: Smoking cessation secondary care Behavioural therapyþ pharmacological therapy for
patients with PTSD

Usual care Dominant

PH48: Smoking cessation secondary care Pharmacological for COPD Borglykke Usual care Dominant

PH48: Smoking cessation secondary care High-intensity behavioural intervention for pregnant

women

Usual care £ 634

PH48: Smoking cessation secondary care Conditional incentives for pregnant women Unconditional incentives £ 3,306

PH50: Domestic violence and abuse: multiagency working Harm reduction, cognitive trauma therapy, battered

women

No intervention Dominant

PH50: Domestic violence and abuse: multiagency working Incidence reduction, independent domestic violence

advisors

No intervention (assuming a percent will

access services without IDVA)

Dominant

PH54: Physical activity exercise referral schemes ERS Usual care £ 88,742

OH prog, oral health programme; COPD, chronic obstructive pulmonary disease; HD, heart disease; CMD, common mental disorder; STRIVE, support to reunite, involve and value each other; SSRI,

selective serotonin reuptake inhibitors; PP, point prevalence; NRT, nicotine replacement therapy; OTC, over the counter; PW, physical activity and workplace; PA, physical activity; GP, general

practitioner; A&E, accident and emergency; MSM, men who have sex with men; LPDS, leicester practice database score; HbA1c, hemoglobin A1c; IDUs, injecting drug users; DBS, dried blood spot;

CDTQ, cut down to quit; NCP, nicotine containing product; IDVA, independent domestic violence advisors; ERS, exercise referral scheme; NICE, national institute for health and care excellence; ICER,

incremental cost-effectiveness ratio.

p
u
b
l
i
c

h
e
a
l
t
h

1
6
9

(
2
0
1
9
)
1
5
1
e
1
6
2

1
5
9

https://doi.org/10.1016/j.puhe.2019.02.011

Table 4 e Number (%) of ICERs identified by cost-
effectiveness and the type of comparator.

Comparator

No
intervention

Usual
care

Other
intervention

Cost saving 42 (19%) 17 (22%) 22 (37%)

�£20,000/
QALY

124 (54%) 46 (60%) 22 (37%)

>£20,000/
QALY

56 (25%) 12 (16%) 6 (10%)

Dominated 4 (2%) 2 (3%) 10 (17%)

Total N 226 77 60

c
2 ¼ 45.6, P < 0.01; in 17 of 380 cases, the comparator could not be categorised, given the information in the reports. ICERs, incremental cost-effectiveness ratios. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2160 great enough, not to enable a meta-analysis to be considered at all. In this case, the ICERs should presumably be kept apart because the effect of the intervention differs either by the population subgroup or by setting. For the exercise of considering the median of all ICERs considered by NICE PH, the assumption of similar cost distri- butions for each subgroup (used in the paragraph previously) may not hold. For the so-called hard-to-reach groups or by setting, such as prison or workplace, the average per-person cost may change. This introduces a further complication, which made analysis even more difficult when individual- level costs and effects are correlated.9 Thus, a more in-depth analysis is well-nigh impossible to undertake. The best we felt that we could do was to collapse ICERs in places where it looks as if many individual ICERs belong to a similar group but without weighting the corre- sponding ‘meta-ICER’ for the increase in degrees of freedom generated by consolidation. This acts as a kind of sensitivity analysis, flawed though it may be. More fundamentally, what does the distribution of ICERs across all NICE PH evidence really tell us? Its usefulness must be cautionary: of the evidence considered, a moderate number of interventions were cost saving and a high percentage very cost-effective. However, a considerable tail was either not cost-effective (ICERs over £20,000) or a few were not even effective (intervention is dominated by its comparator). Type of intervention On the composition of the interventions, some of the least costly interventions, and thus almost certainly, those with low ICERs, have probably been underrepresented. Foremost among these are laws, regulations and taxes to promote good health and avoid illness, such as the ban on smoking in pubs, clubs, restaurants, shops and workplaces that had a one-off cost of legislating and a relatively low enforcement cost, TV and social media advertising giving out messages (on healthy 0% 10% 20% 30% 40% 50% 60% cost saving £1-£15K £15001-£20k 388 ICERs Fig. 1 e Percentage of ICERs from cost saving to dominated for p from March 2006 to March 2018, including (n ¼ 380) and excludi ICERs, incremental cost-effectiveness ratios; NICE, National Ins eating and lifestyle, safe sex and so on) and taxing sugary drinks. Even then, not all such laws and regulations would be cost-effective. For example, some building regulations to reduce accidents may cost very large sums for relatively small benefits. However, the absence of the appraisal of areas such as those mentioned previously was not part of NICE's remit after 2010, so the full value of PH interventions will not have been captured in our analysis. Comparing the cost-effectiveness of PH interventions with technology appraisals It would be tempting to compare the median ICER estimated for PH interventions with that found in technology appraisals at NICE. However, we think this is ill advised. The underlying conditions that drive the ICERs in the two areas are very different. Many technology appraisal (TA) treatments are concerned with new drugs that are under an active patent at the time of appraisal. Patents are designed to promote inno- vation and technological advancement. They do this by £20001-£30k >£30k Dominated

221 ICERs

ublic health interventions assessed and published by NICE

ng (n ¼ 221) clusters of the same intervention or subgroup.
titute for Health and Care Excellence.

https://doi.org/10.1016/j.puhe.2019.02.011

Table 5 e Effect of comparator on incremental cost-effectiveness ratios (ICERs)dillustrative examples from assessment of
smoking cessation interventions.

Intervention quit rate Comparator quit rate Incremental costs Incremental QALYs ICER

Patch and nasal spray 27% (NG92) No intervention 2% £3 0.26 £13

Patch and nasal spray 27% (NG92) Patch only 11% £158 0.17 £948

Abrupt quit and NRT 6% (PH45) No intervention 2% £197 0.035 £5,699

Abrupt quit and NRT 6% (PH45) Cut down to quit and NRT 7.8% £292 �0.048 Dominated
Abrupt quit and NRT 6% (PH45) Abrupt quit and NRT and specialist support 15% £114 �0.078 Dominated
Bupropion (PA) 29.00% (NG92) CBT (PA) 20.90% �£330 0.08 Dominant
Bupropion (PA) 29.00% (NG92) NRT (PA) 29.60% £171 �0.01 Dominated
PA is prolonged abstinence, which is a more conservative measure of quitting than point prevalence.

QALYs, quality adjusted life years.

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2 161
keeping generics at bay for a limited time, allowing the

innovator to charge a price sufficiently high to recoup

research and development costs and additionally provide a

reward for doing so. As a result, new technologies can be ex-

pected to have high incremental costs.10 This situation rarely

occurs in PH guidelines. The appraisal of any new drug for

secondary prevention (and primary prevention) will be con-

ducted by the TA directorate of NICE and inserted into an

appropriate PH guideline (if one exists).

Appropriateness of the QALY and thresholds for local
authorities
The needs of local councils with respect to PH are different

from those of the NHS, so perhaps, for PH, there should be a

wider generic measure of benefit (or at least an augmented

measure starting with QALYs but not ending there) that takes

other needs of citizens into account.11,12 For example, what

concerns many old people under the care of local authorities

is their level of independence and loneliness. For teenagers, in

particular, problems, such as boredom and the changing

pattern of leisure time towards virtual reality and the conse-

quent lack of social interactions, are not captured by the way

health is currently measured. For that reason, it is under-

standable that many local councils may place less reliance on

NICE guidelines on PH than the NHS does on clinical

guidelines.

In addition, a lack of sufficient funds for LAs could point to

a widening ICER threshold gap between PH interventions

carried out by LAs and treatment interventions carried out by

the NHS, whose funds have not been cut to the same extent. In

a previous paragraph, we pointed out that new drugs had a

higher threshold ICER than other interventions in the NHS

because that was the way that new drug discoveries were

funded by way of patents. However, the erosion of funds for

PH13e15 means that not everything can be funded at the

existing ICER threshold. If cuts in PH services were carried out

efficiently, the interventions that would no longer be funded

would be those with the highest ICERs, so that, de facto, a new

and lower ICER would be established. To the extent that some

extremely cost effective interventions have been cut,16 and

others without evidence of cost effectiveness have been

retained it follows the new and smaller PH budget is not

producing as much health gain as it should. This is most

obvious for local authorities who, due to commissioning cy-

cles that operate over the short term, are less interested in

measuring benefits that occur after 1 to 3 years into the future.
Many PH programmes that will be highly cost-effective taken

across a lifetime will not be countenanced if the time horizon

is truncated so severely. Such neglect of preventive in-

terventions that give few immediate health benefits is likely to

lead to huge future costs caused by having to treat what

should more cheaply have been prevented.

Owen et al. reported growing evidence that these signifi-

cant financial pressures are leading local authorities to

disinvest in highly cost-effective non-statutory PH services.2

Unless the trends to lower funding are turned around, the

warning given in that article will lead to lower life expectancy

in future.

Limitations of this study

In this study, we attempted to address the inclusion of mul-

tiple estimates for the same intervention. We did this by

combining estimates based on either subgroup analyses or

different model assumptions where we thought this to be

appropriate. We also limited ICERs to those based on the most

recent evidence for a given intervention. These modifications

required a judgement of some kind to be made. For example,

we combined the ICERs for four interventions of walking trails

reported in the physical activity guideline PH8. The ICERs for

these trails ranged from £87/QALY to £25,150/QALY. The main

difference between them is the wide range in costs of the

construction material used for the trails (e.g. concrete or

woodchip). Similarly, we combined the four ICERs for an

intervention to prevent the uptake of smoking (PH23). The

ICERs reflect different modelling assumptions about the long-

term effects of school-based prevention programmes. For

example, one represented a decrease in smoking prevalence

persisting beyond adolescence and another, a delay in

smoking uptake without any change in prevalence beyond

adolescence. Assuming a decrease in uptake produced an

ICER of £2,030/QALY, while assuming a delay in uptake yielded

an ICER of £11,300/QALY. Finally, regarding the use of the most

recent evidence, some models used estimates of effectiveness

from individual studies that were considered to offer the best

quality evidence; some used multiple estimates to reflect the

range of evidence from different studies, and others used

meta-analyses where these were available. Others may take a

different view on what is appropriate to combine. Neverthe-

less, we think it unlikely that reanalyses will change the

conclusion that most of the PH interventions considered in

NICE guidelines to date are cost-effective.

https://doi.org/10.1016/j.puhe.2019.02.011

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 5 1 e1 6 2162
A further limitation concerns our attempt to explore the

impact of the comparator. There were considerable chal-

lenges in classifying comparators into three types: ‘no inter-

vention’, ‘usual care’ and ‘another intervention’. For example,

‘usual care’ comparators might be no intervention, waiting list

control or another intervention. Comparators comprising ‘no

interventions’ require the estimation of baselines against

which the interventions can be compared. Thus, further

subcategories might be warranted to take account of different

approaches to establish baselines. Our initial expectation that

interventions compared against ‘no intervention’ are most

likely to be identified as cost-effective or cost saving is not

what we found. It is likely there are more meaningful ways,

than aggregating data across interventions, to assess the

impact of comparators. Nevertheless, the analysis, and our

case study of smoking cessation interventions, reinforces the

importance of the comparator in determining the cost-

effectiveness of an intervention. As concluded by Owen

et al.,2 it is imperative that decision makers ensure that they

consider economic analysis that is similar to their decision

problem in terms of population, intervention and comparator.

Conclusion

PH grants to local authorities have significantly reduced over

recent years, with further cuts set to continue to 2020/2021. It

has been said recently that ‘Rising rates of obesity, an un-

healthy relationship with alcohol and fast food and obstinate

pockets of tobacco use are combining with rising life expec-

tancy to exert huge pressures on existing health services.’17

Further cuts to PH spending and activity will be the ‘falsest

of false economies, not least for the NHS’.14

This analysis shows that most PH interventions considered

by NICE represent good value for money but some do not. It

also shows that the results can be sensitive to the approach

adopted. It is crucial, therefore, that decision makers and

commissioners consider these factors, along with the relevant

costs and benefits, when deciding which interventions to

fund.
Author statements

Ethical approval

Not required.

Funding

Neither author received funding for the writing of this article.

Competing interests

Authors declared no competing interests.
r e f e r e n c e s

1. Owen L, Morgan A, Fischer A, Ellis S, Hoy A, Kelly M. The cost-
effectiveness of public health interventions. J Public Health
2011:1e9.

2. Owen L, Pennington B, Fischer A, Jeong K. (2017) the cost-
effectiveness of public health interventions examined by
NICE from 2011 to 2016. J Public Health 2017 Sep;18:1e10.
https://doi.org/10.1093/pubmed/fdx119 [Epub ahead of print].

3. Health and Social Care Act 2012. Chapter 7. 2012. Available from:
http://www.legislation.gov.uk/ukpga/2012/7/pdfs/ukpga_
20120007_en . [Accessed 16 August 2016].

4. Wilmott M, Womack J, Hollingworth W, Campbell R. Making
the case for investment in public health: experiences of
Directors of Public Health in English local government. J Public
Health 2016;38(2):237e42.

5. CEMIPP report (paras 69-70): https://www.gov.uk/
government/uploads/system/uploads/attachment_data/file/
683872/CEMIPP_report_2016__2_ .

6. https://assets.publishing.service.gov.uk/government/
uploads/system/uploads/attachment_data/file/663094/
Accelerated_Access_Collaborative_-_impact_asssessment.
pdf.

7. CEMIPP report (paras 28-29): https://www.gov.uk/
government/uploads/system/uploads/attachment_data/file/
683872/CEMIPP_report_2016__2_ .

8. Fischer A, Ghelardi G. The precautionary principle, evidence-
based medicine, and decision theory in public health
evaluation. Front Public Health 2016;4:107.

9. Briggs A, Sculpher M, Claxton K. Decision modelling for health
economic evaluation (page 95). OUP; 2006.

10. Grootendorst P, Hollis A, Levine DPhD, Pogge T, Edwards AM.
New approaches to rewarding pharmaceutical innovation.
CMAJ 2011 Apr 5;183(6):681e5.

11. Coast J, Smith R, Lorgelly P. Should the capability approach be
applied in health economics? Health Econ 2008
Jun;17(6):667e70.

12. Brazier J, Tsuchiya A. Improving cross-sector comparisons:
going beyond the health-related QALY. Appl Health Econ Health
Policy 2015;13(6):557e65.

13. Department of Health. 2018-19 And 2019-20 ring-fenced
public health grants to local authorities:written statement –
HCWS387. https://www.parliament.uk/business/
publications/written-questions-answers-statements/written-
statement/Commons/2017-12-21/HCWS387.

14. Bunn J. Funding public health by business rates a ‘double-edged
sword’. Local Government Chronicle; 2017. article, https://
www.lgcplus.com/services/health-and-care/funding-public-
health-by-business-rates-a-double-edged-sword/7022569.

15. Buck D. Local government spending on public health: death by a
thousand cuts. The Kings Fund; 2018. https://www.kingsfund.
org.uk/blog/2018/01/local-government-spending-public-
health-cuts.

16. Williams C. Public health cuts hit smoking cessation services.
http://www.publicfinance.co.uk/news/2018/01/public-health-
cuts-hit-smoking-cessation-services.

17. Allen L. Why cutting spending on public health is a false
economy. The Conversation. Global Health Policy, University
of Oxford. www.ox.ac.uk/research/why-cutting-spending-
public-health-false-economy.

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref1

https://doi.org/10.1093/pubmed/fdx119

http://www.legislation.gov.uk/ukpga/2012/7/pdfs/ukpga_20120007_en

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref4

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/683872/CEMIPP_report_2016__2_

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/663094/Accelerated_Access_Collaborative_-_impact_asssessment

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/683872/CEMIPP_report_2016__2_

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref8

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref9

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref10

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref11

http://refhub.elsevier.com/S0033-3506(19)30035-6/sref12

https://www.parliament.uk/business/publications/written-questions-answers-statements/written-statement/Commons/2017-12-21/HCWS387

https://www.lgcplus.com/services/health-and-care/funding-public-health-by-business-rates-a-double-edged-sword/7022569

https://www.kingsfund.org.uk/blog/2018/01/local-government-spending-public-health-cuts

http://www.publicfinance.co.uk/news/2018/01/public-health-cuts-hit-smoking-cessation-services

http://www.ox.ac.uk/research/why-cutting-spending-public-health-false-economy

https://doi.org/10.1016/j.puhe.2019.02.011

The cost-effectiveness of public health interventions examined by the National Institute for Health and Care Excellence fro …
Introduction
Methods
Results
CUA base-case ICERs
Analysis of comparators
Case study of comparators: smoking cessation interventions

Discussion
Main finding of this study
Clustering of ICERs
Type of intervention
Comparing the cost-effectiveness of PH interventions with technology appraisals
Appropriateness of the QALY and thresholds for local authorities

Limitations of this study
Conclusion

Author statements
Ethical approval
Funding
Competing interests
References

Changes-in-cold-related-mortalities-between-1995-and-2016-in-_2019_Public-He

ww.sciencedirect.com

p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 3 6 e4 0
Available online at w
Public Health

journal homepage: www.elsevier.com/puhe
Original Research
Changes in cold-related mortalities between 1995
and 2016 in South East England
G.C. Donaldson a,*, C. Witt b, S. N€ayh€a c

a National Heart and Lung Institute, Guy Scadding Building, Imperial College London, Dovehouse Street, London,

SW3 6LY, United Kingdom
b Division of Ambulatory Pneumology, Charit�e e Universit€atsmedizin Berlin, Berlin, Germany
c Centre for Environmental and Respiratory Health Research, University of Oulu, Finland
a r t i c l e i n f o

Article history:

Received 13 August 2018

Received in revised form

12 December 2018

Accepted 8 January 2019

Available online 18 February 2019

Keywords:

Cold

Mortality

Ischaemic heart disease

Cerebrovascular disease

Respiratory disease
* Corresponding author. National Heart and
Street, London, SW3 6LY, United Kingdom.

E-mail address: gavin.donaldson@imperi
https://doi.org/10.1016/j.puhe.2019.01.008
0033-3506/© 2019 The Royal Society for Publ
a b s t r a c t

Objective: The aim of the study was to examine trends in cold-related mortalities between

1995 and 2016.

Study design: This is a longitudinal mortality study.

Methods: For men and women aged 65e74 years or those older than 85 years in South East

England, the relationship between daily mortality (deaths per million population) and

outdoor temperatures below 18 �C, with allowance for influenza epidemics, was assessed

by linear regression on an annual basis. The regression coefficients were expressed as a

percentage of the mortality at 18 �C to adjust for changes in mortality through health care.

Trends in ‘specific’ cold-related mortalities were then examined over two periods, 1977

e1994 and 1995e2016.

Results: In contrast to the early period, annual trends in cold-related specific mortalities

showed no decline between 1995 and 2016. ‘Specific’ cold-related mortality of women, but

not men, in the age group older than 85 years showed a significant increase over the 1995

e2016 period, which was different from the trend over the earlier period (P < 0.01). Conclusion: Despite state-funded benefits to help alleviate fuel poverty and public health advice, very elderly women appear to be at increasing risk of cold-related mortal- itydgreater help may be necessary. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. Introduction Cold exposure, either indoors or outdoors, is responsible for increased morbidity and mortality from cardiovascular and respiratory causes.1 The mechanisms probably involve Lung Institute, Guy Scad al.ac.uk (G.C. Donaldson) ic Health. Published by E vasoconstriction of the surface blood vessels that forces plasma out of the circulatory system, thereby concentrating clotting factors in the blood and consequently increasing the risk of myocardial infarction and cerebrovascular stroke.2 Cold weather will also promote survival of respiratory ding Building, Room B141, Imperial College London, Dovehouse . lsevier Ltd. All rights reserved. mailto:gavin.donaldson@imperial.ac.uk http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.008&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 3 6 e4 0 37 viruses in nasal secretions originating from infected in- dividuals that contaminate other people.3 Excess winter mortality is seen across Europe and places a familiar winter burden on healthcare systems.1 Excess winter mortality (after allowance for influenza) fell between 1977 and 1994.4 Similar findings have been reported in the Netherlands5,6 and in London,7 with a downwards but non-significant trend in Stockholm.8 This fall was attributed to increased provision of central heating and car ownership and to changes in medical care and general health.4 We now report, with an identical methodology, changes in winter mortality in the same area (South East England) and same population (men and women, aged 65e74 years) over the subsequent 21 years, 1995e2016, and examine an older age group (older than 85 years). Methods Data on daily deaths of men and women aged 65e74 years and older than 85 years in South East England were extracted using 9th and 10th International Classification of Disease (ICD) codes of 410.0e414.9 and I20eI25 for ischaemic heart disease (IHD), 430.0e438.9 and I60eI69 for cerebrovascular disease (CVD), 460.0e519.9 and J00eJ99 for respiratory disease (RES) and 0e999.9 and A00eY00 for all causes, respectively. For all causes, over the 1976e2016 period, there were 849,455 and 588,792 deaths of people aged 65e74 years and 641,930 and 1,386,278 deaths of those older than 85 years, for men and women, respectively. A fifth-order polynomial was fitted to mid-year population data obtained from the Office for National Statistics (ONS) for each sex and age group to provide daily population estimates. Daily deaths divided by these population estimates provided mortality (daily deaths per million individuals). Adjustments were made for differences in coding instructions using ONS bridge-coding data between 1984 and 1994 and for differences between ICD-9 and ICD-10 after 2001. Mean daily temperatures were calculated from hourly measurements at Heathrow Airport and obtained from the British Atmospheric Data Centre. Regression coefficients of mortalities per 1 �C fall in tem- perature for days with temperatures between 0 and 18 �C to avoid overlap with the effects of heat were calculated for each year by linear regression. The equation describing this relationship is as follows: M ¼ (A * T) þ C, where M ¼ mortality, T ¼ temperature, A ¼ regression coefficient and C ¼ constant. The mortality at 18 �C would be calculated as M ¼ (A * 18) þ C. Owing to the delay between temperature and peak mor- tality, mortalities from IHD were lagged on temperature for the regressions by 2 days, CVD by 5 days, RES by 12 days and all causes by 3 days;9 the delays may reflect the pathological processes, with myocardial infarction and ischaemic stroke causing rapid deaths, whereas respiratory infection takes time to develop and overcome defences. Allowance was made for influenza epidemics using mean influenza deaths in men and women older than 55 years averaged from 10 days before to 10 days after each day. To allow for changes unrelated to temperature, such as improvements in healthcare, specific mortality was calculated as the increase in mortality per 1 �C fall in temperature divided by the estimated mortality at 18 �C. Changes in annual values with time were analysed by ordinary linear regression, with the F test for significance, for 1977e1994 and 1995e2016 periods, and with combined data, an interaction term was used to test for the difference in the slopes between the two periods. Results Fig. 1 shows that the climate in South East England has become milder over the recent decades and that deaths from influenza have generally decreased since the large epidemic that occurred in 1976. The percentage of women exceeds that of men in both age groups, with the percentage of the 65- to 74-year-old group declining and that of the group older than 85 years growing. Fig. 2 shows that the annual increase in all-cause mortality per 1 �C fall in temperature continued to decline over the latest 1995e2016 period. Similar results were seen with major causes of mortality, IHD and CVD, although slightly less for RES. Estimates of mortality at 18 �C, an approximate temper- ature at which both the effects of heat and cold are minimal, also continued to decline. However, significant declines in IHD and all-cause annual specific mortality in the 1977e1994 period were not seen in the 1995e2016 period. Fig. 3 shows changes in men and women, and separately for each sex, in the age category above 85 years. In the earlier 1977e1994 period, the declines in all-cause mortality, cold- related mortality, 18 �C mortality and specific mortality were similar to those seen in the younger age groups. Worryingly, in the more recent 1995e2016 period, mortality per 1 �C fall in temperature, mortality at 18 �C and specific cold-related mortality have risen significantly in women (P < 0.01). Discussion Mortality per 1 �C fall in temperature has continued to fall in the age group of 65e74 years over the 21 years since the last report.4 This can be attributed to better medical care and healthier lifestyles as mortality at 18 �C has also fallen. Consequently, as a percentage, there was no significant change in specific mortality (the percentage of cold-related mortality). RES now contributes to most cold-related deaths, despite influenza immunisation, but other clini- cally important respiratory viruses such as coronaviruses, respiratory syncytial viruses and rhinoviruses still circulate in wintertime. The absence of any decline in specific mor- tality particularly after the introduction of winter fuel pay- ments (a state benefit paid to all pensioners to help with heating costs) in 1997 could be due to a trend towards wearing more casual but less warm clothing and elderly people spending more time outdoors where they are exposed to the cold. The limitations of this study should be mentioned. We did not use distributed lag non-linear models10 that would have https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 Fig. 1 e (A) Average daily temperature on days below 18 �C between 1976 and 2016. (B) Average daily mortality per million people from influenza between 1976 and 2016. (C) Men and women in 65- to 74-year-old group and age group older than 85 years as a percentage of total population in 1976e2016. B women 65e74, C men 65e74, ▫ women 85þ, - men 85þ. Fig. 2 e Mortality related to cold in people aged 65e74 years in South East England in 1977e2016 from all-cause mortality and ischaemic heart, cerebrovascular and respiratory diseases. * P < 0.05, **P < 0.01 and ***P < 0.001 for trends over the 1977e1994 and 1995e2016 periods separately. yP < 0.05, yyP < 0.01 and yyyP < 0.001 for differences in the slope between the two periods. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 3 6 e4 038 better captured the temperatureemortality relationship because the multiple coefficients with the various lags would have complicated the assessment of trends over time. Our linear regression approach might underestimate or over- estimate the mortality/temperature if the lag structure changed over time or if confounders disturbed the linearity of the relationship. This study used temperatures recorded at a single weather station to be consistent with the earlier study,4 but it is possible that temperatures derived from multiple stations might have an improved estimation of the mortalityetemperature relationship. In contrast, cold-related specific mortality in women older than 85 years appears to be increasing. This group is growing in population size, and their age may make their admission to hospitals more likely than other groups, thus in part explaining the worsening winter pressures on the National Health Service. The seasonality of mortality has always been greater in older age groups partly because of the reduced thermal perception and body’s heat- generating capacity. As exposure to the cold can occur both indoors and outdoors,1,11 it may be worthwhile raising the age threshold for the winter fuel allowance and directing aid to neglected areas such as public transport waiting areas and community centres. Cold- related mortality particularly in those older than 85 years needs to be monitored. https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 Fig. 3 e All-cause cold-related mortality in people, and men and women separately, older than 85 years in South East England in 1977e2016. *P < 0.05, **P < 0.01 and ***P < 0.001 for trends over the 1977e1994 and 1995e2016 periods separately. yP < 0.05, yyP < 0.01 and yyyP < 0.001 for differences in the slope between the two periods. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 3 6 e4 0 39 Author statements Acknowledgements The authors are grateful to the Office for National Statistics for mortality and population data, and the Meteorological Office and the British Atmospheric Data Centre for temperature data. Ethical approval Data were anonymous and aggregated, and thus, no partici- pant approval was required. The study plan was approved by the Office for National Statistics before the data were ob- tained. Mortality and population data were obtained from the Office for National Statistics, and climate data, from the Meteorological Office. Funding No funding organisation was involved. Competing interests The authors have no competing interests. Patient involvement None. Author contributions G.C.D. conceived the study and performed the analysis. G.C.D., C.W. and S.N. wrote and critically reviewed the manuscript. G.C.D. acted as the guarantor. r e f e r e n c e s 1. Keatinge WR, Donaldson GC, Bucher K, Cordioli E, Dardanoni L, Jendritzky G, et al. Cold exposure and winter mortality from ischaemic heart, cerebrovascular, and respiratory disease, and all causes, in warm and cold regions of Europe. Lancet 1997;349:1341e6. http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref1 https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 3 6 e4 040 2. Keatinge WR, Donaldson GC. Cardiovascular mortality in winter. Arctic Med Res 1995;54(Suppl. 2):16e8. 3. Donaldson GC, Wedzicha JA. The causes and consequences of seasonal variation in COPD Exacerbations. Int J COPD 2014;9:1101e10. 4. Donaldson GC, Keatinge WR. Mortality related to cold weather in elderly people in southeast England, 1979e94. BMJ 1997;315:1055e6. 5. Kunst AE, Looman CWN, Mackenbach JP. The decline in winter excess mortality in The Netherlands. Int J Epidemiol 1991;20(4):971e7. 6. Ekamper P, Van Poppel F, Van Duin C, Garssen J. 150 Years of temperature-related excess mortality in The Netherlands. Demogr Res 2009;21:385e426. 7. Carson C, Hajat S, Armstrong B, Wilkinson P. Declining vulnerability to temperature-related mortality in London over the 20th century. Am J Epidemiol 2006;164(1):77e84. 8. �Astr€om DO, Forsberg B, Edvinsson S, Rockl€ov J. Acute fatal effects of short-lasting extreme temperatures in Stockholm, Sweden: evidence across a century of change. Epidemiology 2013 Nov 1;24(6):820e9. 9. Donaldson GC, Keatinge WR. Early increases in ischaemic heart disease mortality dissociated from, and later changes associated with, respiratory disease, after cold weather in south-east England. J Epidemiol Community Health 1997;51:643e8. 10. Gasparrini A. Modelling lagged associations in environmental time series data: a simulation study. Epidemiology 2016;27(6):835e42. 11. Keatinge WR, Coleshaw SRK, Holmes J. Changes in seasonal mortalities with improvement in home heating in England and Wales from 1964 and 1984. Int J Biometeorol 1989;33:71e6. http://refhub.elsevier.com/S0033-3506(19)30008-3/sref2 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref2 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref2 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref3 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref3 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref3 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref3 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref4 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref4 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref4 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref4 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref4 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref5 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref5 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref5 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref5 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref6 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref6 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref6 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref6 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref7 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref7 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref7 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref7 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref8 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref9 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref10 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref10 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref10 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref10 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref11 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref11 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref11 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref11 http://refhub.elsevier.com/S0033-3506(19)30008-3/sref11 https://doi.org/10.1016/j.puhe.2019.01.008 https://doi.org/10.1016/j.puhe.2019.01.008 Changes in cold-related mortalities between 1995 and 2016 in South East England Introduction Methods Results Discussion Author statements Acknowledgements Ethical approval Funding Competing interests Patient involvement Author contributions References The-relationship-between-self-reported-sensory-impairments-and-psy_2019_Publ ww.sciencedirect.com p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8 Available online at w Public Health journal homepage: www.elsevier.com/puhe Original Research The relationship between self-reported sensory impairments and psychosocial health in older adults: a 4-year follow-up study using the English Longitudinal Study of Ageing A. Yu a, A.E.M. Liljas b,* a Research Department of Epidemiology and Public Health, Institute of Epidemiology and Health Care, University College London, London, WC1E 6BT, United Kingdom b Research Department of Primary Care and Population Health, Institute of Epidemiology and Health Care, University College London, London, NW3 2PF, United Kingdom a r t i c l e i n f o Article history: Received 19 July 2018 Received in revised form 15 January 2019 Accepted 31 January 2019 Available online 21 March 2019 Keywords: Ageing Hearing impairment Vision impairment Self-rated health Quality of life Depression E-mail address: ann.liljas.13@ucl.ac.uk (A https://doi.org/10.1016/j.puhe.2019.01.018 0033-3506/© 2019 The Royal Society for Publ a b s t r a c t Objectives: To explore cross-sectional and longitudinal relationships between self-reported hearing and vision impairments and self-rated health, quality of life (QoL) and depressive symptoms at 4-year follow-up. Study design: The study involved cross-sectional and longitudinal analyses with 4-year follow-up using data from the English Longitudinal Study of Ageing. Methods: Community-dwelling adults (n ¼ 3931) aged �50 years from the English Longitu- dinal Study of Ageing participated in this study. Self-reported hearing and vision were defined as good or poor. Self-rated health was treated as a dichotomous variable (good and poor health). QoL was based on the 19-item Critical Appraisal Skills Programme and treated as a continuous variable (score 0e57). Depressive symptoms were assessed using the eight- item Center for Epidemiologic Studies Depression Scale (CES-D8) and defined as CES-D�3. Relationships between sensory impairments and self-rated health and depressive symp- toms were analysed using logistic regression. Linear regression was used to assess the relationships between sensory impairments and QoL. Results: In cross-sectional analyses, both self-reported hearing and vision impairment were positively associated with all outcomes assessed. In longitudinal analyses, self-reported poor hearing and vision were associated with increased risks of poor self-rated health (hearing: odds ratio [OR] 1.65, 95% confidence interval [CI] 1.32, 2.05; vision: OR 1.57, 95% CI 1.16, 2.12) and depressive symptoms (hearing: OR 1.35, 95% CI 1.07, 1.71; vision: OR 1.44, 95% CI 1.09, 1.90) after adjustment for sociodemographic and lifestyle factors, chronic illness, mobility limitations and cognition. Poor hearing and poor vision were not associ- ated with reduced QoL after adjustment for covariates. Conclusions: The findings stress the importance of identifying and addressing sensory im- pairments in older adults to improve their health and well-being. © 2019 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved. 87 * Corresponding author. Tel.: þ20 7794 0500x36721; fax: þ20 7472 6 .E.M. Liljas). ic Health. Published by E 1. lsevier Ltd. All rights reserved. mailto:ann.liljas.13@ucl.ac.uk http://crossmark.crossref.org/dialog/?doi=10.1016/j.puhe.2019.01.018&domain=pdf www.sciencedirect.com/science/journal/00333506 www.elsevier.com/puhe https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8 141 Introduction The population of England is ageing due to increased life ex- pectancy.1 Loss of hearing and vision is common in later life affecting 20% and 11%, respectively, of British adults aged �60 years.2,3 Poor hearing and poor vision have been associated with adverse health outcomes, including disability,4,5 frailty6,7 and dementia.8,9 Hearing and vision impairments have also been associated with poor self-rated health,10e12 a strong subjective measure for overall health that encompasses both physical health and psychological well-being, predicting major adverse health outcomes such as chronic illness,13 functional decline14,15 and depression.16 There is also evi- dence from cross-sectional studies of a relationship between impairments in hearing and vision with psychosocial health factors including quality of life (QoL) and depression;17e20 however, little is known about such relationships longitudi- nally. QoL is a broad term that in addition to health also measures perceived sense of control, autonomy, self- realisation and pleasure in a participant's life.21 Reporting poor QoL has been shown to be associated with prevalent chronic illness22 and poor physical functioning.17 Depression is a common condition with prevalence of depression in older adults in England estimated at 10%;23 however, this percentage may be underestimated because 22% of men and 28% of women aged �65 years reported depression assessed using the 10-item Geriatric Depression Scale in the cross-sectional Health Survey for England in 2005.24 Older adults are believed to be particularly susceptible to depression because of higher levels of comorbidities and physical and cognitive decline.25 Besides physical and cogni- tive decline, depression in older age has also been associated with social exclusion, dementia and increased risks of sui- cide.26 Given the increasing number of older adults, poor psychosocial health in older age has become a growing public health concern due to the associated burden on the individual, their families, the health and social care system and society.27 Investigating longitudinal relationships between sensory im- pairments and psychosocial health in later life has the po- tential to provide more insight into the influence of sensory impairments on older adults' health. Such relationships are of importance to examine as sensory impairments often are preventable or treatable.8,28 Therefore, this study aims to longitudinally explore associations between impairments in hearing and vision and subsequent low self-rated health, changes in QoL and development of depressive symptoms in older English adults. Methods Design and study population This study used data from the English Longitudinal Study of Ageing (ELSA), a nationally representative cohort of English adults aged �50 years, drawn from the Health Survey for En- gland in 1998, 1999 and 2001.29 Participants have been fol- lowed up every 2 years including additional nurse visits every 4 years. This study used data collected in 2004 (wave 2) as the baseline data and data from 2008 (wave 4) as follow-up data. This study sample was restricted to 3931 individuals with data on all three outcomes at both time points and data on sensory impairments and covariates at the baseline, allowing for complete case analysis. All participants provided written informed consent. Ethical approval for ELSA was obtained from the Multicentre Research and Ethics Committee, the system of approval for multicentre studies in England. Exposures of interest: hearing impairment and vision impairment Data on self-reported hearing were collected by asking ‘Is your hearing (using a hearing aid if you use one)...’ with the answer options ‘excellent’, ‘very good’, ‘good’, ‘fair’ and ‘poor’. For vision, participants were asked ‘Is your eyesight (using glasses or corrective lenses if you use them)...’ with the same answer options as for hearing. Similar to previous studies,6,7,30 reporting ‘fair’ or ‘poor’ hearing and vision, respectively, was classified as having poor hearing/poor vision. ‘Good’, ‘very good’ or ‘excellent’ were combined into good hearing and good vision, respectively. Outcomes of interest Self-rated health Data on self-rated health were collected by asking ‘Would you say your health is...’ with the answer options ‘excellent’, ‘very good’, ‘good’, ‘fair’ and ‘poor’. For the analyses, ‘fair’ or ‘poor’ were classified as poor self-rated health and ‘good’, ‘very good’ or ‘excellent’ were classified as good self-rated health. Quality of life QoL was assessed using the validated 19-item Critical Appraisal Skills Programme (CASP-19) instrument comprising four domains that measure degrees of the par- ticipant's perceived sense of control (four questions), au- tonomy (five questions), self-realisation (five questions) and pleasure (five questions).31 Each of the 19 items were recoded so that the values ranged from 0 to 3 with 0 being ‘often’, 1 ‘sometimes’, 2 ‘not often’ and 3 ‘never’. All items were then summed arithmetically to form a score ranging from 0 to 57. QoL was treated as a continuous measure. A higher score referred to greater QoL, whereas a lower score referred to poorer QoL. Depressive symptoms Data on depressive symptoms were collected using the Center for Epidemiologic Studies Depression Scale (CES-D8), a vali- dated eight-item questionnaire on feelings of sadness, lone- liness and restless sleep.32,33 Although CES-D8 can be treated as a continuous measure, clinically, it may be more relevant to categorise participants into depressed and non-depressed.34 In this study, depressive symptoms were defined as having three or more of the eight items, a cut-off used to indicate a clinical diagnosis of depression.35e37 CES-D8 has provided 63% sensitivity and 90% specificity using the Euro-D scale as the reference.38 https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8142 Covariates Covariates used in the analyses include age, sex, cohabitation, education, wealth, physical activity, body mass index (BMI), smoking, chronic disease, cognitive function and mobility. Individuals were grouped into those with no educational qualifications versus those with secondary education or higher. Wealth referred to the total net non-pension wealth (financial, housing and physical wealth) of the household presented as quintiles. Cohabitation referred to the current living arrangements and included spouse, partner, relatives and caretakers living with the respondent. Physical activity, known for having a positive impact on overall health and depression,39 was based on frequency and intensity of exer- cise and allowed for four groups: sedentary, mild, moderate and vigorous physical activity carried out on at least a weekly basis. The BMI referred to someone's weight in relation to their height and was considered as a covariate owing to the higher prevalence of overweight and obesity among those who have sensory impairments compared with non-sensory impaired individuals40 and its influence on overall health41 and depression.42 Chronic illness was based on self-reported doctor-diagnosed heart attack, stroke and diabetes, where reporting one or more was classified as having chronic illness. Cognitive function and mobility limitations were considered because they have previously been associated with both hearing impairment17,43,44 and vision impairment45e47 and may be linking sensory impairments with the outcomes explored including self-reported health,48 poor QoL49 and depression.50 Cognitive function was based on a validated 24- item cognitive score,51 which was calculated by adding par- ticipants' scores from the three tests: immediate recall (scored 0e10), delay recall (0e10) and orientation (0e4) to create a final cognition score. For the recall tests, the participants were presented 10 words of which they were asked to recall as many words as possible immediately and again 5 min later. For questions on orientation, the participants were asked to correctly report the day, date, month and year. Having a mobility limitation was defined as reporting difficulty climb- ing a flight of stairs or walking 100 yards. Statistical analyses Sample characteristics (Table 1) in participants with good and poor self-reported hearing and good and poor self-reported vision were calculated using percentages for variables with categorical data; means and standard deviations (SDs) were used for variables with continuous data. Linear regression was used to determine cross-sectional and longitudinal relation- ships between sensory impairments and continuous mea- sures of QoL. For the longitudinal analyses, the models were adjusted for the baseline QoL. Logistic regression was used to assess cross-sectional and longitudinal relationships between sensory impairments and self-rated health and depressive symptoms (CES-D�3), using good hearing and good vision, respectively, as reference groups. The longitudinal analyses of sensory impairments and self-rated health and depressive symptoms were adjusted for the baseline self-rated health and baseline depressive symptoms, respectively. The baseline weighting was used for all analyses to reduce the risk of bias caused by non-responses. All models were adjusted for age and sex followed by further adjustment for additional cova- riates (cohabitation, education, wealth, physical activity, BMI, smoking and chronic illness). Fully adjusted models also included adjustment for cognitive function and mobility lim- itations. The analyses were conducted using R and its default statistical packages. Results Of 3931 participating individuals (55% women) aged �50 years (mean age 64 years [SD 8.4]), 19% self-reported poor hearing and 10% self-reported poor vision (Table 1). Compared with participants who self-reported good hearing, participants who self-reported poor hearing were more likely to be older, male and less wealthy with no educational qualifications, lead a sedentary lifestyle and have a chronic illness, poorer cognitive function and mobility limitations. Compared with partici- pants who self-reported good vision, those with poor self- reported vision were more likely to be older, female and less wealthy with fewer educational qualifications, were less likely to be living with someone and a current smoker, lead a sedentary lifestyle and have chronic illness, poorer cognitive function and mobility limitations. Cross-sectional associations Cross-sectional findings are presented in Table 2 and show that compared with participants who self-reported good hearing, participants with poor self-reported hearing were nearly three times as likely to report poor self-rated health (age- and sex-adjusted odds ratio [OR] 2.89, 95% confidence interval [CI] 2.39, 3.49). Poor hearing was also associated with poorer QoL (age- and sex-adjusted b �3.57, 95% CI �4.26, 2.87) and depressive symptoms (age- and sex-adjusted OR 2.07, 95% CI 1.69, 2.53). All associations remained after further adjust- ment for cohabitation, education, wealth, physical activity, BMI, smoking, chronic illness, cognition and mobility limitation. Compared with good self-reported vision, self-reported poor vision was associated with poorer health outcomes including more than fourfold increased odds of poor self-rated health (age- and sex-adjusted OR 4.42, 95% CI 3.53, 5.54). Self- reported poor vision was also associated with poorer QoL (age- and sex-adjusted b �4.62, 95% CI �5.57, 3.67) and with depressive symptoms (age- and sex-adjusted OR 2.04, 95% CI 1.59, 2.61). The associations remained after further adjust- ment for covariates. Longitudinal associations Findings from the longitudinal analyses are shown in Table 3. Compared with participants who self-reported good hearing, those who self-reported poor hearing were significantly more likely to self-report poor health at 4-year follow-up after adjustment for socio-economic and lifestyle factors, chronic illness, mobility limitation and cognition (OR 1.65, 95% CI 1.32, 2.05). Self-reported poor hearing was furthermore associated https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 Table 1 e Characteristics of the study sample at the baseline (2004). Hearing Vision Overall Good Poor P-value Good Poor P-value Totals, n (%) 3931 (100) 3199 (81) 732 (19) 3551 (90) 380 (10) Age in years, mean (SD) 64.4 (8.4) 63.8 (8.2) 66.9 (8.8) <0.001 64.1 (8.3) 66.8 (9.5) <0.001 Female, n (%) 2145 (55) 1851 (58) 294 (40) <0.001 1914 (54) 231 (61) 0.012 Cohabiting, n (%) 2915 (74) 2366 (74) 549 (75) 0.594 2672 (75) 243 (64) <0.001 Smoker, n (%) 509 (13) 415 (13) 94 (13) 0.973 441 (12) 68 (18) 0.003 No educational qualifications, n (%) 609 (15) 521 (16) 88 (12) 0.005 576 (16) 33 (9) 0.001 Wealth quintiles, n (%) 1 (least wealthy) 465 (12) 354 (11) 111 (15) <0.001 370 (10) 95 (25) <0.001 2 645 (16) 492 (15) 153 (21) 561 (16) 84 (22) 3 816 (21) 659 (21) 157 (21) 754 (21) 62 (16) 4 935 (24) 787 (25) 148 (20) 859 (24) 76 (20) 5 (most wealthy) 1070 (27) 907 (28) 163 (22) 1007 (28) 63 (17) Body mass index in kg/m2, mean (SD) 27.9 (4.8) 27.87 (4.8) 27.93 (4.6) 0.725 27.9 (4.7) 28.2 (5.2) 0.262 Physical activity (intensity on a weekly basis), n (%) 0 (sedentary) 148 (4) 95 (3) 53 (7) <0.001 109 (3) 39 (10) <0.001 1 (mild) 452 (11) 349 (11) 103 (14) 386 (11) 66 (17) 2 (moderate) 1987 (51) 1612 (50) 375 (51) 1795 (51) 192 (51) 3 (vigorous) 1344 (34) 1143 (36) 201 (27) 1261 (36) 83 (22) Chronic illness (heart attack, stroke and/ or diabetes), n (%) 507 (13) 378 (12) 129 (18) <0.001 293 (8) 87 (23) <0.001 Cognitive function (0-24), mean (SD) 14.7 (1.2) 14.9 (3.2) 13.7 (3.2) <0.001 14.8 (3.2) 13.8 (3.4) <0.001 Mobility (difficulty taking stairs and/or walking 400 yards), n (%) 474 (12) 325 (10) 149 (20) <0.001 377 (11) 97 (26) <0.001 Prevalence of outcome measures by sensory function Poor self-reported health, n (%) 762 (19) 504 (16) 258 (35) <0.001 582 (16) 180 (47) <0.001 Quality of life (0-57), mean (SD) 43.8 (8.1) 44.5 (7.9) 40.84 (8.4) <0.001 44.3 (7.9) 39.5 (8.8) <0.001 Depressive symptoms (�3), n (%) 708 (18) 520 (16) 188 (26) <0.001 596 (23) 112 (29) <0.001 SD, standard deviation p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8 143 with odds of depressive symptoms (adjusted for age, sex and the baseline depressive symptoms OR 1.52, 95% CI 1.21, 1.91), and the association remained after further adjustments. Self- reported poor hearing was not associated with poorer QoL at 4-year follow-up. Compared with participants who self-reported good vision, participants who self-reported poor vision were nearly twice as likely to self-report poor health at 4-year follow-up (adjusted for age, sex and the baseline self-reported health Table 2 e Cross-sectional associations between sensory impair in a sample of 3931 adults aged ≥50 years in 2004 (baseline). Models Self-rated health Good hearing Poor hearing P-value Poor h [OR] [OR (95% CI)] [b (95 Adjusted M1a 1.00 2.89 (2.39, 3.49) <0.001 �3.57 (�4 Adjusted M2b 1.00 2.65 (2.16, 3.26) <0.001 �2.89 (�3 Adjusted M3c 1.00 2.39 (1.92, 2.98) <0.001 �2.47 (�3 Good vision Poor vision P-value Poor [OR] [OR (95% CI)] [b (95 Adjusted M1a 1.00 4.42 (3.53, 5.54) <0.001 �4.62 (�5 Adjusted M2b 1.00 3.45 (2.70, 4.41) <0.001 �3.16 (�4 Adjusted M3c 1.00 3.45 (2.65, 4.48) <0.001 �2.75 (�3 BMI, body mass index; OR, odds ratio; CI, confidence interval; QoL, qualit a M1 (model 1) ¼ adjusted for age and sex. b M2 (model 2) ¼ M1 þ cohabitation, education, wealth, physical activity c M3 (model 3) ¼ M2 þ mobility limitations and cognition. OR 1.76, 95% CI 1.33, 2.33), and the association remained after further adjustment (OR 1.75, 95% CI 1.16, 2.12). Compared with good vision, poor vision was associated with poorer QoL at 4- year follow-up (adjusted for age, sex and the baseline QoL b �0.93, 95% CI �1.64, 0.22). However, the association dimin- ished after further adjustment for covariates. Participants who self-reported poor vision were more likely to report depressive symptoms at 4-year follow-up (adjusted for age, sex and the baseline depressive symptoms OR 1.82, 95% CI ments and self-rated health, QoL and depressive symptoms QoL Depressive symptoms earing P-value Good hearing Poor hearing P-value % CI)] [OR] [OR (95% CI)] .26, �2.87) <0.001 1.00 2.07 (1.69, 2.53) <0.001 .54, �2.24) <0.001 1.00 1.93 (1.56, 2.38) <0.001 .11, �1.83) <0.001 1.00 1.77 (1.42, 2.19) <0.001 vision P-value Good vision Poor vision P-value % CI)] [OR] [OR (95% CI)] .57, �3.67) <0.001 1.00 2.04 (1.59, 2.61) <0.001 .08, �2.23) <0.001 1.00 1.60 (1.23, 2.09) <0.001 .66, �1.84) <0.001 1.00 1.49 (1.13, 1.96) 0.004 y of life. , smoking, BMI and chronic illness. https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 Table 3 e Longitudinal associations between sensory impairments in 2004 (baseline) and self-rated health, QoL and depressive symptoms at 4-year follow-up (2008) in a sample of 3931 adults aged ≥50 years in 2004. Models Self-rated health QoL Depressive symptoms Good hearing Poor hearing P-value Poor hearing Good hearing Poor hearing P-value [OR] [OR (95% CI)] [b (95% CI)] P-value [OR] [OR (95% CI)] Adjusted M1a 1.00 1.71 (1.39, 2.12) <0.001 �0.41 (�0.93, 0.11) 0.124 1.00 1.52 (1.21, 1.91) <0.001 Adjusted M2b 1.00 1.70 (1.37, 2.11) <0.001 �0.33 (�0.85, 0.18) 0.207 1.00 1.40 (1.11, 1.77) 0.004 Adjusted M3c 1.00 1.65 (1.32, 2.05) <0.001 �0.26 (�0.78, 0.26) 0.322 1.00 1.35 (1.07, 1.71) 0.011 Good vision Poor vision P-value Poor vision P-value Good vision Poor vision P-value [OR] [OR (95% CI)] [b (95% CI)] [OR] [OR (95% CI)] Adjusted M1a 1.00 1.76 (1.33, 2.33) <0.001 �0.93 (�1.64, �0.22) 0.010 1.00 1.82 (1.39, 2.38) <0.001 Adjusted M2b 1.00 1.57 (1.17, 2.11) 0.002 �0.68 (�1.39, 0.03) 0.061 1.00 1.48 (1.12, 1.95) 0.005 Adjusted M3c 1.00 1.57 (1.16, 2.12) 0.003 �0.63 (�1.34, 0.08) 0.083 1.00 1.44 (1.09, 1.90) 0.009 BMI, body mass index; OR, odds ratio; CI, confidence interval; QoL, quality of life. a M1 (model 1) ¼ adjusted for age, sex and outcome of interest at the baseline: self-rated health/QoL/depression, respectively. b M2 (model 2) ¼ M1 þ cohabitation, education, wealth, physical activity, smoking, BMI and chronic illness. c M3 (model 3) ¼ M2 þ mobility limitations and cognition. p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8144 1.39, 2.38) compared with their counterparts who self- reported good vision, and the association remained after further adjustments. Discussion The cross-sectional analyses showed that self-reported poor hearing and poor vision were associated with poor self-rated health, poorer QoL and depressive symptoms. In the longitu- dinal analysis, compared with individuals who self-reported good hearing, self-reported poor hearing was associated with increased risk of poor self-rated health and depressive symptoms. Self-reported poor hearing was also associated with subsequent poorer QoL; however, the association was attenuated after adjustment for covariates. Self-reported poor vision was associated with subsequent poor self-rated health and depressive symptoms and the associations remained after adjustment for covariates. Self-reported poor vision was also associated with poorer QoL; however, the association disappeared after adjustment for covariates. Analyses of the influence of self-reported sensory impair- ments on self-rated health is of interest because self-rated health is a subjective measure of both physical health and well-being10 and known for being a strong predictor of mor- tality.14 Our longitudinal finding supports previous cross- sectional studies reporting an association between both sub- jectively and objectively assessed hearing impairment and self-rated health.10,12,52,53 Similarly, the observed association between self-reported poor hearing and reporting depressive symptoms 4 years later supports recent research showing an association between objectively assessed hearing impairment and clinically diagnosed depression in older adults.54 In our study, self-reported poor hearing was not associated with subsequent poorer QoL at 4-year follow-up. The lack of asso- ciation observed is inconsistent with previous studies reporting an association between hearing impairment and poorer QoL;55e57 however, these previous studies were of cross-sectional design. Indeed, our cross-sectional analysis showed an association between self-reported poor hearing and QoL. Addressing hearing problems has been shown to positively influence QoL,58 and it is possible that participants who reported poor hearing at the baseline had addressed their hearing problems 4 years later. However, most older adults with hearing problems do not seek help,28 and therefore, the lack of association between self-reported poor hearing and poor QoL may not be due to action taken on their hearing loss. Instead the lack of association may be explained by the socio- economic status and physical health, factors of importance previously shown to influence QoL.49,59,60 Less wealth and poor physical health might outweigh the impact of poor hearing on QoL in older age.58,61e67 Similar to self-reported poor hearing, cross-sectional ana- lyses showed associations between self-reported poor vision and poor self-rated health, poorer QoL and depressive symp- toms. Self-reported poor vision was also associated with subsequent poor self-rated health and depressive symptoms at 4-year follow-up. Our longitudinal findings support previ- ous studies reporting a relationship of age-related vision impairment (self-reported and objectively assessed) with poor self-rated health10e12 and with depression.68,69 Contrary to existing literature,70,71 self-reported poor vision was not associated with poorer QoL. In our study, QoL was measured using the validated CASP-19. However, using a vision-related measure such as the Activities of Daily Vision Scale, con- cerned exclusively with measuring specific activities related to eyesight, might have captured poor QoL related to vision more successfully.72 However, the Activities of Daily Vision Scale has not been used in ELSA. Furthermore, the severity of uncorrected vision impairment has been shown to negatively influence QoL with greater vision impairment having worse negative impact on QoL.73 Using more than two categories of vision function (good and poor vision) might, therefore, have been useful. However, because of the small number of par- ticipants reporting ‘poor’ vision (n ¼ 62 [2% of the study sample]), this answer option was combined with ‘fair’ vision (n ¼ 318). It is possible that the lack of association is due to low statistical power.74,75 The associations observed in this study may be explained by unmeasured factors such as hypertension and anxiety. https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8 145 Hypertension is a modifiable risk factor for cardiovascular disease (CVD) and highly prevalent in the older population.76 Hypertension was, however, not included as a possible confounder because previous research has reported incon- sistent findings, with some studies showing no association between sensory impairments and hypertension.77e79 In ELSA, data on anxiety refer to a single-item self-reported question likely to highly correlate with the measures of depression and therefore not necessarily appropriate to be considered as a covariate. Furthermore, social isolation due to communication problems may link sensory impairments to poor self-rated health, poorer QoL and depression.80,81 The associations observed may also be explained by poorly assessed conditions including self-reported doctor-diagnosed CVD and the limited aspects of cognition included. Strengths and limitations A major strength of this study is that it uses prospective data collected from a large nationally representative sample of English adults aged �50 years. Validated standard measures of self-rated health, QoL and depression were used to assess the outcomes.31,33 Participants were followed up for 4 years, and the relationships explored were adjusted for a multiple covariates. Limitations include that ELSA participants without com- plete data on the variables investigated were removed from our study, causing potential selection bias. Non-respondents were older (<0.001), less wealthy (<0.001) and had worse health outcomes including more chronic illness (<0.001), worse cognitive function (<0.001) and poorer mobility (0.001), suggesting the associations observed may be even stronger among non-respondents. Hearing and vision function were self-reported. However, the questions on hearing and vision used have been validated against objective measures,82,83 and this study's prevalence rates are similar to national es- timates.2,3 Self-reported data on hearing and vision were collected at the baseline with no information on changes overtime. Furthermore, the outcome measures' self-rated health, QoL and depressive symptoms were only assessed at the baseline and again at 4-year follow-up with no record of changes in any direction during the follow-up period. Furthermore, the cognitive score adjusted for referred only to orientation and immediate and delayed recall. Other as- pects of cognition were not included proposing the role of cognition has not been fully accounted for. Another limita- tion is that the data used for the analyses are from 2004 to 2008 and may be considered dated. Finally, the ELSA cohort comprised predominantly white English adults, making it difficult to generalise the findings to other ethnic groups. No effort was made to oversample groups with small numbers such as the oldest-old and those from ethnic minority backgrounds. Conclusions and implications This study shows that self-reported hearing and vision prob- lems are associated with greater odds of poor self-rated health and depressive symptoms at 4-year follow-up. The findings stress the importance of identifying and addressing sensory impairments in older adults to potentially reduce their risk of experiencing worse general health and depressive symptoms overtime. It is possible that poor hearing in older adults to some extent can be corrected by provision and use of hearing aids as population-based data have shown that only 15% of adults aged 48e92 years with objectively assessed hearing impairment use a hearing aid.84 Similarly, it is estimated that 20e50% of older adults have undetected reduced vision and that the majority of these cases are due to refractive errors or cataract and thus correctable.85 Provision and use of specta- cles or lenses of the appropriate prescription, or cataract extraction, respectively, might have the potential to address most of the vision problems experienced in older adults.8 Finally, research replicating the study using objectively assessed hearing and vision is needed to evaluate whether the observed relationships of self-reported sensory impairments with adverse health outcomes are consistent with objectively measured sensory function. Author statements Acknowledgements The English Longitudinal Study of Ageing was developed by a team of researchers based at the University College London, National Centre for Social Research and the Institute for Fiscal Studies. The data were collected by the National Centre for Social Research. The funding is provided by the National Institute of Aging in the United States and a Consortium of UK Government Departments Coordinated by the Economic and Social Research Council (ESRC). The developers and funders of the English Longitudinal Study of Ageing and the UK Data Archive do not bear any responsibility for the analyses or in- terpretations presented here. Ethical approval Ethical approval for ELSA was obtained from the Multicentre Research and Ethics Committee. Funding A.E.M.L. is funded by the National Institute for Health Research (NIHR), School for Primary Care Research (grant number 538469). The views expressed are those of the authors and not necessarily those of the NIHR, National Health Service or Department of Health. Competing interests The authors have no competing interests to declare. r e f e r e n c e s 1. National population projections: 2016-based statistical bulletin, ONS (Office for National Statistics). 2017. Available at: https:// www.ons.gov.uk/peoplepopulationandcommunity/ populationandmigration/populationprojections/bulletins/ https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8146 nationalpopulationprojections/ 2016basedstatisticalbulletin#a-growing-number-of-older- people. [Accessed 29 January 2018]. 2. Akeroyd MA, Foreman K, Holman JA. Estimates of the number of adults in England, Wales, and Scotland with a hearing loss. Int J Audiol 2014;53:60e1. 3. Sightloss UK 2013. The latest evidence., RNIB (Royal National Institute of Blind People). 2013. Available at: https://www.rnib. org.uk/sites/default/files/Sight_loss_UK_2013 . [Accessed 29 January 2018]. 4. Liljas AE, Wannamethee SG, Whincup PH, Papacosta O, Walters K, Iliffe S, et al. Hearing impairment and incident disability and all-cause mortality in older. British community- dwelling men Age and ageing 2016;45:662e7. 5. West SK, Munoz B, Rubin GS, Schein OD, Bandeen-Roche K, Zeger S, et al. Function and visual impairment in a population-based study of older adults. The SEE project. Salisbury Eye Evaluation. Invest Ophthalmol Vis Sci 1997;38:72e82. 6. Liljas AE, Carvalho LA, Papachristou E, De Oliveira C, Wannamethee SG, Ramsay SE, et al. Self-reported vision impairment and incident prefrailty and frailty in English community-dwelling older adults: findings from a 4-year follow-up study. J Epidemiol Community Health 2017;71:1053e8. 7. Liljas AE, Carvalho LA, Papachristou E, Oliveira C, Wannamethee SG, Ramsay SE, et al. Self-reported hearing impairment and incident frailty in English community- dwelling older adults: a 4-year follow-up study. J Am Geriatr Soc 2017;65:958e65. 8. Bowen M, Edgar DF, Hancock B, Haque S, Shah R, Buchanan S, et al. The Prevalence of Visual Impairment in People with Dementia (the PrOVIDe study): a cross-sectional study of people aged 60e89 years with dementia and qualitative exploration of individual, carer and professional perspectives. Health Serv Deliv Res 2016;4. 9. Davies HR, Cadar D, Herbert A, Orrell M, Steptoe A. Hearing impairment and incident dementia: findings from the English longitudinal study of ageing. J Am Geriatr Soc 2017;65:2074e81. 10. Jylha M, Guralnik JM, Ferrucci L, Jokela J, Heikkinen E. Is self- rated health comparable across cultures and genders? J Gerontol B Psychol Sci Soc Sci 1998;53:S144e52. 11. Wang JJ, Foran S, Mitchell P. Age-specific prevalence and causes of bilateral and unilateral visual impairment in older Australians: the Blue Mountains Eye Study. Clin Exp Ophthalmol 2000;28:268e73. 12. Liu PL, Cohen HJ, Fillenbaum GG, Burchett BM, Whitson HE. Association of Co-existing impairments in cognition and self- rated vision and hearing with health outcomes in older adults. Geront & Geriatric Med 2016;2. 13. Molarius A, Janson S. Self-rated health, chronic diseases, and symptoms among middle-aged and elderly men and women. J Clin Epidemiol 2002;55:364e70. 14. Idler EL, Kasl SV. Self-ratings of health: do they also predict change in functional ability? J Gerontol B Psychol Sci Soc Sci 1995;50:S344e53. 15. Leinonen R, Heikkinen E, Jylha M. Changes in health, functional performance and activity predict changes in self- rated health: a 10-year follow-up study in older people. Arch Gerontol Geriatr 2002;35:79e92. 16. Mulsant BH, Ganguli M, Seaberg EC. The relationship between self-rated health and depressive symptoms in an epidemiological sample of community-dwelling older adults. J Am Geriatr Soc 1997;45:954e8. 17. Arlinger S. Negative consequences of uncorrected hearing loss–a review. Int J Audiol 2003;42(Suppl 2). 2s17-20. 18. Boi R, Racca L, Cavallero A, Carpaneto V, Racca M, Dall' Acqua F, et al. Hearing loss and depressive symptoms in elderly patients. Geriatr Gerontol Int 2012;12:440e5. 19. Evans JR, Fletcher AE, Wormald RP. Depression and anxiety in visually impaired older people. Ophthalmology 2007;114:283e8. 20. Dalton DS, Cruickshanks KJ, Klein BE, Klein R, Wiley TL, Nondahl DM. The impact of hearing loss on quality of life in older adults. Gerontol 2003;43:661e8. 21. Nussbaum M, Sen A. The quality of life. Oxford: Clarendon Press; 1993. 22. Wikman A, Wardle J, Steptoe A. Quality of life and affective well-being in middle-aged and older people with chronic medical illnesses: a cross-sectional population based study. PLoS One 2011;6:e18952. 23. McDougall FA, Matthews FE, Kvaal K, Dewey ME, Brayne C. Prevalence and symptomatology of depression in older people living in institutions in England and Wales. Age Ageing 2007;36:562e8. 24. Health Survey for England 2005, health of older people, HSE (health Survey for England). 2007. Available at: https://digital.nhs.uk/ catalogue/PUB01184. [Accessed 27 February 2018]. 25. Mirowsky J, Ross CE. Age and depression. J Health Soc Behav 1992;33:187e205. discussion 206-112. 26. Fiske A, Wetherell JL, Gatz M. Depression in older adults. Annu Rev Clin Psychol 2009;5:363e89. 27. Wang PS, Simon G, Kessler RC. The economic burden of depression and the cost-effectiveness of treatment. Int J Methods Psychiatr Res 2003;12:22e33. 28. Davis A, Smith P, Ferguson M, Stephens D, Gianopoulos I. Acceptability, benefit and costs of early screening for hearing disability: a study of potential screening tests and models. Health Technol Assess 2007;11:1e294. 29. Marmot M, Oldfield Z, Clemens S, Blake M, Phelps A, Nazroo J, et al. English longitudinal study of ageing: waves 0-6, 1998-2013. 23rd ed. London: UK Data Service; 2015. https://doi.org/10. 5255/UKDA-SN-5050-10. [Accessed 20 February 2018]. 30. Maharani A, Dawes P, Nazroo J, Tampubolon G, Pendleton N. Visual and hearing impairments are associated with cognitive decline in older people. Age Ageing 2018;47:575e81. 31. Hyde M, Wiggins RD, Higgs P, Blane DB. A measure of quality of life in early old age: the theory, development and properties of a needs satisfaction model (CASP-19). Aging Ment Health 2003;7:186e94. 32. Wallace RB, Herzog AR, Ofstedal MB, Steffick DE, Fonda S, Langa KM. Documentation of affective functioning measures in the health and retirement study. 2000. Available at: http://hrsonline. isr.umich.edu/sitedocs/userg/dr-005 . [Accessed 20 February 2018]. 33. Karim J, Weisz R, Bibi Z, Rehman uS. Validation of the eight- item center for epidemiologic studies depression scale (CES- D) among older adults. Curr Psychol 2015;34:681e92. 34. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385e401. 35. Schane RE, Walter LC, Dinno A, Covinsky KE, Woodruff PG. Prevalence and risk factors for depressive symptoms in persons with chronic obstructive pulmonary disease. J Gen Intern Med 2008;23:1757e62. 36. Blazer D, Burchett B, Service C, George LK. The association of age and depression among the elderly: an epidemiologic exploration. J Gerontol 1991;46:M210e5. 37. Andresen EM, Malmgren JA, Carter WB, Patrick DL. Screening for depression in well older adults: evaluation of a short form of the CES-D (Center for Epidemiologic Studies Depression Scale). Am J Prev Med 1994;10:77e84. https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationprojections/bulletins/nationalpopulationprojections/2016basedstatisticalbulletin#a-growing-number-of-older-people http://refhub.elsevier.com/S0033-3506(19)30024-1/sref2 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref2 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref2 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref2 https://www.rnib.org.uk/sites/default/files/Sight_loss_UK_2013 https://www.rnib.org.uk/sites/default/files/Sight_loss_UK_2013 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref4 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref4 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref4 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref4 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref4 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref5 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref6 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref7 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref8 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref9 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref9 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref9 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref9 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref9 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref10 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref10 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref10 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref10 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref11 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref11 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref11 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref11 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref11 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref12 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref12 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref12 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref12 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref12 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref13 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref13 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref13 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref13 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref14 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref14 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref14 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref14 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref15 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref15 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref15 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref15 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref15 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref16 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref16 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref16 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref16 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref16 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref17 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref17 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref18 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref18 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref18 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref18 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref19 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref19 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref19 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref19 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref20 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref20 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref20 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref20 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref21 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref21 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref22 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref22 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref22 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref22 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref23 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref23 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref23 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref23 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref23 https://digital.nhs.uk/catalogue/PUB01184 https://digital.nhs.uk/catalogue/PUB01184 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref25 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref25 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref25 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref26 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref26 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref26 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref27 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref27 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref27 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref27 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref28 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref28 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref28 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref28 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref28 https://doi.org/10.5255/UKDA-SN-5050-10 https://doi.org/10.5255/UKDA-SN-5050-10 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref30 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref30 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref30 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref30 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref31 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref31 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref31 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref31 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref31 http://hrsonline.isr.umich.edu/sitedocs/userg/dr-005 http://hrsonline.isr.umich.edu/sitedocs/userg/dr-005 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref33 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref33 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref33 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref33 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref34 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref34 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref34 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref34 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref35 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref35 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref35 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref35 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref35 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref36 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref36 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref36 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref36 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref37 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref37 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref37 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref37 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref37 https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8 147 38. Courtin E, Knapp M, Grundy E, Avendano-Pabon M. Are different measures of depressive symptoms in old age comparable? An analysis of the CES-D and Euro-D scales in 13 countries. Int J Methods Psychiatr Res 2015;24:287e304. 39. Strawbridge WJ, Deleger S, Roberts RE, Kaplan GA. Physical activity reduces the risk of subsequent depression for older adults. Am J Epidemiol 2002;156:328e34. 40. Weil E, Wachterman M, McCarthy EP, Davis RB, O'Day B, Iezzoni LI, et al. Obesity among adults with disabling conditions. J Am Med Assoc J Am Med Assoc 2002;288:1265e8. 41. Kopelman P. Health risks associated with overweight and obesity. Obes Rev Off J Int Assoc Study Obes 2007;8(Suppl 1):13e7. 42. Luppino FS, de Wit LM, Bouvy PF, Stijnen T, Cuijpers P, Penninx BW, et al. Overweight, obesity, and depression: a systematic review and meta-analysis of longitudinal studies. Arch Gen Psychiatr 2010;67:220e9. 43. Chen DS, Genther DJ, Betz J, Lin FR. Association between hearing impairment and self-reported difficulty in physical functioning. J Am Geriatr Soc 2014;62:850e6. 44. Lin FR, Yaffe K, Xia J, Xue QL, Harris TB, Purchase-Helzner E, et al. Hearing loss and cognitive decline in older adults. JAMA internal medicine 2013;173:293e9. 45. Reuben DB, Mui S, Damesyn M, Moore AA, Greendale GA. The prognostic value of sensory impairment in older persons. J Am Geriatr Soc 1999;47:930e5. 46. Reyes-Ortiz CA, Kuo YF, DiNuzzo AR, Ray LA, Raji MA, Markides KS. Near vision impairment predicts cognitive decline: data from the hispanic established populations for epidemiologic studies of the elderly. J Am Geriatr Soc 2005;53:681e6. 47. Wallhagen MI, Strawbridge WJ, Shema SJ, Kurata J, Kaplan GA. Comparative impact of hearing and vision impairment on subsequent functioning. J Am Geriatr Soc 2001;49:1086e92. 48. Hultsch DF, Hammer M, Small BJ. Age differences in cognitive performance in later life: relationships to self-reported health and activity life style. J Gerontol 1993;48:P1e11. 49. Blane D, Netuveli G, Montgomery SM. Quality of life, health and physiological status and change at older ages. Soc Sci Med (1982) 2008;66:1579e87. 50. Potter GG, Steffens DC. Contribution of depression to cognitive impairment and dementia in older adults. Neurol 2007;13:105e17. 51. Langa KM, Llewellyn DJ, Lang IA, Weir DR, Wallace RB, Kabeto MU, et al. Cognitive health among older adults in the United States and in England. BMC Geriatr 2009;9:23. 52. Karpa MJ, Gopinath B, Beath K, Rochtchina E, Cumming RG, Wang JJ, et al. Associations between hearing impairment and mortality risk in older persons: the Blue Mountains Hearing Study. Ann Epidemiol 2010;20:452e9. 53. Gopinath B, Hickson L, Schneider J, McMahon CM, Burlutsky G, Leeder SR, et al. Hearing-impaired adults are at increased risk of experiencing emotional distress and social engagement restrictions five years later. Age Ageing 2012;41:618e23. 54. Kim SY, Kim HJ, Park EK, Joe J, Sim S, Choi HG. Severe hearing impairment and risk of depression. A national cohort study 2017;12:e0179973. 55. Newman CW, Jacobson GP, Hug GA, Sandridge SA. Perceived hearing handicap of patients with unilateral or mild hearing loss. Ann Otol Rhinol Laryngol 1997;106:210e4. 56. Scherer MJ, Frisina DR. Characteristics associated with marginal hearing loss and subjective well-being among a sample of older adults. J Rehabil Res Dev 1998;35:420e6. 57. Chia EM, Mitchell P, Rochtchina E, Foran S, Golding M, Wang JJ. Association between vision and hearing impairments and their combined effects on quality of life. Arch Ophthalmol 2006;124:1465e70. 58. Boorsma M, Joling K, Dussel M, Ribbe M, Frijters D, van Marwijk HW, et al. The incidence of depression and its risk factors in Dutch nursing homes and residential care homes. Am J Geriatr Psychiatry Off J Am Assoc Geriatr Psychiatr 2012;20:932e42. 59. Netuveli G, Wiggins RD, Hildon Z, Montgomery SM, Blane D. Quality of life at older ages: evidence from the English longitudinal study of aging (wave 1). J Epidemiol Community Health 2006;60:357e63. 60. Gale CR, Cooper C, Deary IJ, Aihie Sayer A. Psychological well- being and incident frailty in men and women: the English Longitudinal Study of Ageing. Psychol Med 2014;44:697e706. 61. Li CM, Zhang X, Hoffman HJ, Cotch MF, Themann CL, Wilson MR. Hearing impairment associated with depression in US adults, national health and nutrition examination survey 2005-2010. JAMA otolaryngology– head & neck surgery 2014;140:293e302. 62. Mener DJ, Betz J, Genther DJ, Chen D, Lin FR. Hearing loss and depression in older adults. J Am Geriatr Soc 2013;61:1627e9. 63. Tambs K. Moderate effects of hearing loss on mental health and subjective well-being: results from the Nord-Trondelag Hearing Loss Study. Psychosom Med 2004;66:776e82. 64. Nachtegaal J, Smit JH, Smits C, Bezemer PD, van Beek JH, Festen JM, et al. The association between hearing status and psychosocial health before the age of 70 years: results from an internet-based national survey on hearing. Ear Hear 2009;30:302e12. 65. Lee AT, Tong MC, Yuen KC, Tang PS, Vanhasselt CA. Hearing impairment and depressive symptoms in an older Chinese population. J Otolaryngol 2010;39:498e503. 66. Gopinath B, Wang JJ, Schneider J, Burlutsky G, Snowdon J, McMahon CM, et al. Depressive symptoms in older adults with hearing impairments: the Blue Mountains Study. J Am Geriatr Soc 2009;57:1306e8. 67. Penninx BW, Guralnik JM, Ferrucci L, Simonsick EM, Deeg DJ, Wallace RB. Depressive symptoms and physical decline in community-dwelling older persons. J Am Med Assoc J Am Med Assoc 1998;279:1720e6. 68. Ribeiro MV, Hasten-Reiter Junior HN, Ribeiro EA, Juca MJ, Barbosa FT, Sousa-Rodrigues CF. Association between visual impairment and depression in the elderly: a systematic review. Arq Bras Oftalmol 2015;78:197e201. 69. Choi HG, Lee MJ, Lee SM. Visual impairment and risk of depression: a longitudinal follow-up study using a national sample cohort, vol. 8; 2018. p. 2083. 70. Carabellese C, Appollonio I, Rozzini R, Bianchetti A, Frisoni GB, Frattola L, et al. Sensory impairment and quality of life in a community elderly population. J Am Geriatr Soc 1993;41:401e7. 71. Brown RL, Barrett AE. Visual impairment and quality of life among older adults: an examination of explanations for the relationship. J Gerontol B Psychol Sci Soc Sci 2011;66:364e73. 72. Mangione CM, Gutierrez PR, Lowe G, Orav EJ, Seddon JM. Influence of age-related maculopathy on visual functioning and health-related quality of life. Am J Ophthalmol 1999;128:45e53. 73. Rein DB, Wirth KE, Johnson CA, Lee PP. Estimating quality- adjusted life year losses associated with visual field deficits using methodological approaches. Ophthalmic Epidemiol 2007;14:258e64. 74. Coyne JC. Self-reported distress: analog or Ersatz depression? Psychol Bull 1994;116:29e45. 75. Clak DA, Beck AT. Scientific foundations of cognitive theory and therapy of depression. Hoboken: John Wiley & Sons Inc; 1999. http://refhub.elsevier.com/S0033-3506(19)30024-1/sref38 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref38 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref38 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref38 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref38 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref39 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref39 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref39 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref39 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref40 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref40 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref40 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref40 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref41 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref41 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref41 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref41 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref42 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref42 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref42 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref42 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref42 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref43 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref43 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref43 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref43 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref44 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref44 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref44 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref44 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref45 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref45 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref45 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref45 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref46 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref47 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref47 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref47 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref47 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref47 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref48 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref48 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref48 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref48 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref49 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref49 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref49 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref49 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref50 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref50 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref50 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref50 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref51 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref51 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref51 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref52 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref52 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref52 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref52 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref52 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref53 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref54 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref54 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref54 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref55 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref55 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref55 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref55 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref56 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref56 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref56 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref56 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref57 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref57 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref57 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref57 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref57 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref58 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref59 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref59 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref59 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref59 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref59 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref60 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref60 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref60 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref60 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref61 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref62 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref62 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref62 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref63 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref63 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref63 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref63 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref64 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref65 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref65 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref65 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref65 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref66 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref66 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref66 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref66 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref66 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref67 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref67 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref67 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref67 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref67 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref68 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref68 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref68 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref68 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref68 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref69 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref69 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref69 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref70 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref70 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref70 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref70 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref70 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref71 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref71 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref71 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref71 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref71 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref72 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref72 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref72 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref72 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref72 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref73 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref73 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref73 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref73 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref73 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref74 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref74 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref74 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref75 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref75 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref75 https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 p u b l i c h e a l t h 1 6 9 ( 2 0 1 9 ) 1 4 0 e1 4 8148 76. Townsend N, Bhatnagar P, Wilkins E, Wickramasinghe K, Rayner M. Cardiovascular disease statistics. London: British Heart Foundation; 2015. 77. Hirai A, Takata M, Mikawa M, Yasumoto K, Iida H, Sasayama S, et al. Prolonged exposure to industrial noise causes hearing loss but not high blood pressure: a study of 2124 factory laborers in Japan. J Hypertens 1991;9:1069e73. 78. Chang TY, Liu CS, Huang KH, Chen RY, Lai JS, Bao BY. High- frequency hearing loss, occupational noise exposure and hypertension: a cross-sectional study in male workers. Environ Health 2011;10:35. 79. Wang WL, Chen N, Sheu MM, Wang JH, Hsu WL, Hu YJ. The prevalence and risk factors of visual impairment among the elderly in Eastern Taiwan. Kaohsiung J Med Sci 2016;32:475e81. 80. Weeks DG, Michela JL, Peplau LA, Bragg ME. Relation between loneliness and depression: a structural equation analysis. J Pers Soc Psychol 1980;39:1238e44. 81. Hawton A, Green C, Dickens AP, Richards SH, Taylor RS, Edwards R, et al. The impact of social isolation on the health status and health-related quality of life of older people. Qual Life Res Int J Qual Life Aspects Treat Care Rehabil 2011;20:57e67. 82. Gibson WK, Cronin H, Kenny RA, Setti A. Validation of the self-reported hearing questions in the Irish longitudinal study on ageing against the whispered voice test. BMC Res Notes 2014;7:361. 83. Zimdars A, Nazroo J, Gjonca E. The circumstances of older people in England with self-reported visual impairment: a secondary analysis of the English Longitudinal Study of Ageing (ELSA). Br J Vis Impair 2012;30:22e30. 84. Popelka MM, Cruickshanks KJ, Wiley TL, Tweed TS, Klein BE, Klein R. Low prevalence of hearing aid use among older adults with hearing loss: the Epidemiology of Hearing Loss Study. J Am Geriatr Soc 1998;46:1075e8. 85. Evans BJ, Rowlands G. Correctable visual impairment in older people: a major unmet need. Ophthalmic Physiol Optic J Bri Coll Ophthalmic Opticians (Optometrists) 2004;24:161e80. http://refhub.elsevier.com/S0033-3506(19)30024-1/sref76 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref76 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref76 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref77 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref78 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref78 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref78 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref78 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref79 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref79 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref79 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref79 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref79 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref80 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref80 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref80 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref80 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref81 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref81 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref81 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref81 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref81 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref82 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref82 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref82 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref82 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref83 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref83 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref83 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref83 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref83 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref84 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref84 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref84 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref84 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref84 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref85 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref85 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref85 http://refhub.elsevier.com/S0033-3506(19)30024-1/sref85 https://doi.org/10.1016/j.puhe.2019.01.018 https://doi.org/10.1016/j.puhe.2019.01.018 The relationship between self-reported sensory impairments and psychosocial health in older adults: a 4-year follow-up stud ... Introduction Methods Design and study population Exposures of interest: hearing impairment and vision impairment Outcomes of interest Self-rated health Quality of life Depressive symptoms Covariates Statistical analyses Results Cross-sectional associations Longitudinal associations Discussion Strengths and limitations Conclusions and implications Author statements Acknowledgements Ethical approval Funding Competing interests References Editorial-Board_2019_Public-Health www.elsevier.com/locate/puhe Editorial Office Isabel Mattar Melissa Davis Public Health Editorial Office, RSPH, John Snow House, 59 Mansell St., London, E1 8AN, Tel.: +44 (0) 207 265 7331 Fax: +44 (0) 207 265 7301 E-mail: public.health@rsph.org.uk Editorial Board Editors-in-Chief Phil Mackie Edinburgh, UK Fiona Sim London, UK Senior Associate Editors Raheelah Ahmad London, UK Cathy Johnman Glasgow, UK Andrew Lee Sheffield, UK Joanne Morling Nottingham, England, UK Associate Editors John Ford Cambridge, UK Ryan Swiers Nottingham, UK Rifat Atun Boston, USA John Beard Geneva, Switzerland Petri Bockerman Turku, Finland Noriko Cable London, UK Ann DeBaldo Florida, USA Linda Degutis Atlanta, USA Peter Donnelly St. Andrews, UK Mark Eisler Bristol, UK Brian Ferguson York, UK Robert Friis California, USA Sian Griffiths Hong Kong Jay Glasser Houston, Texas, USA John Goddeeris Michigan, USA Lawrence Gostin Washington, USA David Hunter Durham, UK Michael Kelly London, UK Giuseppe La Torre Rome, Italy Roger Magnusson Sydney, Australia Gerry McCartney Glasgow, UK George Morris Troon, Ayrshire, UK Angus Nicoll Stockholm, Sweden David Pencheon Cambridge, UK Mala Rao London, UK Devi Sridhar Edinburgh, UK Seung Wook Lee Seoul, Republic of Korea International Editorial Board

Turn in your highest-quality paper
Get a qualified writer to help you with

“ Economic Evaluation and Health Policy ”

Get high-quality paper

NEW! AI matching with writer

Hire a Writer

Client Reviews

4.9

Sitejabber

4.6

Trustpilot

4.8

Our Guarantees

100% Confidentiality

Information about customers is confidential and never disclosed to third parties.

Original Writing

We complete all papers from scratch. You can get a plagiarism report.

Timely Delivery

No missed deadlines – 97% of assignments are completed in time.

Money Back

If you're confident that a writer didn't follow your order details, ask for a refund.

New to Your Trusted Assignment Help Service? Sign up & Save

Calculate the price of your order

Type of paper needed:

Pages:

You will get a personal manager and a discount.

Academic level:

We'll send you the first draft for approval by at

Total price:

$0.00

Power up Your Academic Success with the
Team of Professionals. We’ve Got Your Back.

Power up Your Study Success with Experts We’ve Got Your Back.

Order Now Order Now