Epidemiology

1. Attributable Risk (AR) provides information about the excess risk in exposed versus un-exposed.

 True

 False

QUESTION 2

1. Relative Risk is the probability that a member of an unexposed group will develop disease relative to the probability that a member of an exposed group will develop the same disease.

 True

 False

QUESTION 3

1. Which of the following goes into cell c of the epidemiological matrix?

QUESTION 4

1. Which measure tells us the percentage of the disease among the exposed that is attributable to a particular exposure?

QUESTION 5

1. When AR > 0, the risk is less in the exposed person than unexposed person.

 True
 False

QUESTION 6

1. Which of the following goes into cell b of the epidemiological matrix?

E-D+

E+D-

E-D-

E+D+

QUESTION 7

1. Which of the following goes into cell d of the epidemiological matrix?

E-D+

E+D-

E-D-

E+D+

QUESTION 8

1. Which of the following goes into cell a of the epidemiological matrix?

E-D+

E+D-

E-D-

E+D+

QUESTION 9

1. When RR < 1.0, it indicates greater risk in the exposed group than in the unexposed group.

 True
 False

QUESTION 10

1. Incidence among the exposed = Etiologic cases + background incidence.

 True
 False

QUESTION 11

1. The incidence in the exposed is calculated by:

QUESTION 12

1. Relative Risk measures strength of association between exposure and disease.

 True
 False

QUESTION 13

1. Relative means __________, while absolute means ___________.

QUESTION 14

1. Excess cases shows a person never had exposure, so the person would not have become a D+ during the time interval.

 True
 False

QUESTION 15

1. [(a + c) / (a + b +c + d) ] x 10n is the formula for:

QUESTION 16

1. When RR > 1.0, it indicates less risk in the exposed group than the unexposed group.

 True
 False

QUESTION 17

1. 200 new cases of HPV were diagnosed in the 1000 college women who got PAP smears at WKU last year. Of those, 120 had five or more sexual partners since becoming sexually active, as compared to 360 who did not have HPV.

Calculate and interpret RR, AR, AP, PAR, and PAR%.  Please show your calculations.

Lesson 4 Practice Examples Key

1. In country A, the mortality rate from unintentional injuries for last year is 36/100,000 people. In country B, the mortality rate from unintentional injuries for the same year is 42/100,000 people.

a. Calculate the RR for death due to unintentional injuries in country B as compared to country A. Show your work, please.

RR crude = (42/100,000) / (36/100,000) = 1.17

b. State what your RR means.

This means that the risk of death due to unintentional injuries are 1.17 times greater (or 17% greater) in Country B as compared to Country A

c. What are some of the potential explanations for your findings?

One could speculate many things to explain the difference: unsafe roads, high risk-taking behaviors, inadequate emergency response system, etc. However, what I want you to think about here is that we really don’t KNOW what this increased risk of death means. It could possibly be a result of some other factor – such as age. There could be a lot more teens in country B….

2.Four hundred people were stranded in a shelter during a recent hurricane. Relief workers brought sandwiches – 200 chicken salad and 200 peanut butter and jelly — to the shelter for people to eat. That evening, 183 people were stricken with diarrhea. Of those, 141 had eaten chicken salad; the rest had peanut butter and jelly. Each person ate one sandwich.

The first step is to set up your epi matrix:

200

a. How much more or less likely were people who ate chicken salad to get diarrhea as compared to those who ate peanut butter and jelly?

This question is asking us to calculate what? (I’m pausing to let you answer) If you said RR, you are correct. The first step in doing this, therefore, is to calculate the incidence in the exposed.

IE+ = a/(a+b) = 141/(141+59) = 141/200=.705

Next, calculate the incidence among the unexposed

IE- = c/(c+d) = 42/(42+158) = 42/200 = .21

Now we can calculate RR

by comparing IE+ to IE-

RR = IE+/IE- = .705/.21 = 3.36

My interpretation is this: People who ate the chicken salad were 3.36 times more likely to have diarrhea than those who ate peanut butter and jelly.

b. How many cases per 100 could be attributed to the chicken salad?

Here, we are calculating absolute risk: AR = IE+ – IE- = .705 – .21 = .495

Because I asked for “per 100” we will multiply this by 100, which equals 49.5, and interpret our finding as such: 49.5 per 100 cases of diarrhea were attributable to chicken salad.

c. What percentage of diarrhea cases would be eliminated if we did away with the chicken salad?

Here we need to calculate the attributable proportion (AR%).

AR% = AR/IE+ = .495/.705 = .702

Note: AR% can also be calculated as such:

AR% = (RR-1)/RR = (3.36 – 1)/3.36 = 2.36/3.36 = .702

Pretty cool, eh? Anyway, my interpretation would be: If we served only peanut butter and jelly, we would eliminate 70.2% of the diarrhea among people in the shelter.

d. How many cases of diarrhea per 100 in the population could be prevented if chicken salad was not served?

Here we are looking at PAR. Recall that PAR is an absolute measure that looks at the proportion of disease that can be prevented if that exposure were eliminated.

PAR = I total – IE-

Where I total = (a+c) / (a + b + c + d)

So,

PAR = [(141+42) / (143 + 59 + 42 + 158 )] – [42 / (42 + 158)

= (183 / 400) – .21

= .458 – .21

= .248

Because we want to report this as per 100, we then multiply by 100 giving us the answer of 24.8 per 100 and interpret this to mean that 24.8 of the 70.5* cases of diarrhea per 100 could be eliminated by not serving chicken salad.

*this is the IE+ multiplied by 100 to get a per 100 measure

e. What percentage of diarrhea in the population could be prevented if chicken salad was not served?

Here you are being asked to compute and interpret the PAR%. We can do this the easy way by taking PAR and dividing it by the I total that we just computed:

PAR% = .248 / .458 = .541

Multiply by 100 to get the % and we get PAR% = 54.1%

Now we interpret by saying 54.1% of diarrhea in this population could be eliminated by not serving chicken salad.

3. In State A, the mortality rate from breast cancer for last year was 110/100,000 people. In State B, the mortality rate from breast cancer for the same year was 143/100,000 people.

a. Calculate the RR for death due to breast cancer in State B as compared to State A. Show your work, please.

RR crude = (143/100,000) / (110/100,000) = 0.00143/ 0.0011= 1.3

b. State what your RR means.

This means that the risk of death due to breast cancer is 1.3 times greater (or 30% greater) in State B as compared to State A

c. What are some of the potential explanations for your findings?

One could speculate many things to explain the difference: family history, hormone replacement therapy, age etc. However, what I want you to think about here is that we really don’t KNOW what this increased risk of death means. It could possibly be a result of some other factor – such as age. There could be a lot more women in State B who are at the specific range of age….

4. 500 people attended in a conference held in convention center. They were served by two different sandwiches – 200 ham sandwiches and 300 chicken sandwiches. That evening, 162 people were diagnosed with diarrhea. Of those, 148 had eaten ham sandwiches; the rest had chicken sandwiches. Each person ate one sandwich.

First we need to set up our table and fill in the missing data. The known data are in bold.

D+ (diarrhea)

D- (no diarrhea)

Total

200

a. How much more or less likely were people who ate ham sandwich to get diarrhea as compared to those who ate chicken sandwich?

This is asking us to calculate what? (I’m pausing to let you answer) If you said RR, you are correct. The first step in doing this, therefore, is to calculate the incidence in the exposed.

IE+ = a/(a+b) = 148/(148+52) = 148/200=.74

Next, the incidence among the unexposed

IE- = c/(c+d) = 14/(14+286) = 14/300 = .047

Now we can calculate RR

RR = IE+/IE- = .74/. 047 = 15.7

My interpretation is this: People who ate the ham sandwich were 15.7 times more likely to have diarrhea than those who ate chicken sandwiches.

b. How many cases per 100 could be attributed to the ham sandwich?

AR = IE+ – IE- =. 74- 0.047= 0.693

Because I asked for “per 100” we will multiply this by 100, which equals 69.3 and interpret our finding as such: 69.3 per 100 cases of diarrhea were attributable to ham sandwiches.

c. What percentage of diarrhea cases would be eliminated if we did away with the ham sandwich?

Here we need to calculate the attributable proportion (AR%).

AR% = AR/IE+ = 0.693/.74 = 0.93

Note: AR% can also be calculated as such:

AR% = (RR-1)/RR = (15.7 – 1)/15.7= 14.7/15.7= 0.93

My interpretation is: If we served only chicken sandwiches, we would eliminate 93% of the diarrhea among people in the shelter.

d. How many cases of diarrhea per 100 in the population could be prevented if ham sandwich was not served?

Here we are looking at PAR. Recall that PAR is an absolute measure that looks at the proportion of disease that can be prevented if that exposure were eliminated.
PAR = I total – IE-
Where I total = (a+c) / (a + b + c + d)
So,

PAR = [(148+14) / (148 + 14 + 52 + 286)] – [14 / (14 + 286)]

= (162/ 500) – 0.05

= 0.324- 0.05

= .274

Because we want to report this as a per 100, we then multiply by 100 giving us the answer of 27.4 per 100 and interpret this to mean that 27.4 of the 74 cases of diarrhea per 100 could be eliminated by not serving chicken salad.

*this is the IE+ multiplied by 100 to get a per 100 measure

e. What percentage of diarrhea in the population could be prevented if ham sandwich was not served?

Here you are being asked to compute and interpret the PAR%. We can do this the easy way by taking PAR and dividing it by the I total that we just computed:

PAR% = .274 / .324 = .84

Multiply by 100 to get the % and we get PAR% = 84 %

Now we interpret by saying 84% of diarrhea in this population could be eliminated by not serving ham sandwich.

5. In city A, the mortality rate from heart disease for last year was 110/100,000 people. In city B, the mortality rate from heart disease for the same year was 154/100,000 people.

a.

Calculate the RR for death due to heart disease in city B as compared to city A. Show your work, please.

RR crude = (154/100,000) / (110/100,000) = 0.00154/ 0.0011= 1.4

b. State what your RR means.

This means that the risk of death due to heart disease is 1.4 times greater (or 40% greater) in City B as compared to City A

c. What are some of the potential explanations for your findings?

One could speculate many things to explain the difference: smoking, high risk-taking behaviors, inadequate physical activity, etc. However, what I want you to think about here is that we really don’t KNOW what this increased risk of death means. It could possibly be a result of some other factor – such as age. There could be a lot more elderly in city B….

Lesson 4

Measures of Effect

We continue with descriptive epi and build on the information you learned during the last lesson where we looked at measures of disease frequency within a single population. This week, we begin comparing two populations. Aren’t you excited? =)
1

Overview of Lecture
Overview of Effect
Epidemiological Matrix
Relative Measures of Effect
Relative Risk
Odds Ratio (covered in a different lesson)
Absolute Measures of Effect
Absolute Risk
Attributable Risk Percent
Impact
Attributable Proportion
Population Attributable Proportion

2

Measures of Effect
How strong is the relationship between two factors
Association: how much does one factor vary according to the value of another factor
Impact: how much does one factor account for the value of the second factor

3
Measures of effect express how two groups or populations differ (or how they are the same) with regard to some measure of disease frequency (usually P, I or M: remember those from the last lesson? ). They show the strength of a relationship between two variables with regard to either association or impact.
Association looks at how much one factor – such as a disease — varies according to the value of another factor – such as an exposure. An example is the gas pedal on a car. As you push it down, the car goes faster. The speed (D) varies according to how far you press down the pedal (E).
Impact looks at how much one factor – such as an exposure– accounts for the value of the second factor – such as a disease. An example here is sales commission. If you earn 3% commission on any sale you make, the amount of the sale (E) will affect the amount of commission that you make (D).
We’ll apply these principles throughout this lecture, so don’t worry if it’s not quite clear yet. =)
NOTE: Sometimes you will see “measures of association” used interchangeably with “measures of effect”

Association & Impact
Association
Relative: Division
Absolute: Subtaction
Impact : Percentage

4
Association can be:
A)    “Relative” measures of effect are the type most commonly calculated by epidemiologists. Here, relative means “division” — that is, something to be divided by something else.
B)     “Absolute” measures of effect are less commonly calculated. Here, absolute means “subtraction” — that is, something is subtracted from something else.
Impact is usually expressed as a percentage
 

Example 1:
If you are 10 and your brother is 5, how are your ages associated?
RELATIVE: 2 times older
10/5 (age 1 relative to age 2)
ABSOULTE: 5 year difference
10-5 = 5
IMPACT:
(10-5)/ 10 = 50%
(difference relative to age 1)

5
If you are 10 and your brother is 5, how are your ages associated?
RELATIVE: 2 times older. 10/5
ABSOULTE: 5 year difference. 10-5 = 5
IMPACT: (10-5)/ 10 = 50%

Example 2:
If you are now 30 and your brother is 25, how are your ages associated?
Don’t look at the notes until you’ve calculated relative, absolute, and impact

6
RELATIVE: rather than twice, as with example 1, it is now 30/25 = 1.2 times older
ABSOULTE: still 5 year difference — 30-25 = 5 year difference
IMPACT: (30-25) / 30 = 17%

Measures of Effect
E  D
Cause/Risk Factor — Time —– Outcome/Effect

7
Measures of association establish the relationship of E to D over time as depicted on the slide. In order to determine the strength of the relationship between E and D, we have to begin by categorizing people into one of four groups so that we can compare them. These groups are:
People who were exposed and got the disease (E+, D+)
People who were exposed but did not get the disease (E+, D-)
People who were not exposed but still got the disease (E-, D+)
People who were not exposed and who did not get the disease (E-, D-)
Note: In epi articles, E- is written as E with a line on top of it — which is called “E bar”. I cannot figure out how to do this here so I will continue to represent E bar as E-. Additionally, E+ is commonly written as E in journals. Same thing holds true with D.

Epi Matrix

8

In-Class Example: Epi Matrix
There was a recent outbreak of diarrhea onboard a small cruise ship. Of the 400 passengers, 120 got diarrhea. Among those who got diarrhea, 90 had gone to a beach-themed party, w4hich was attended by a total of 250 people.
Construct your 2×2, filling in each cell.

There was a recent outbreak of diarrhea onboard a small cruise ship. Of the 400 passengers, 120 got diarrhea. Among those who got diarrhea, 90 had gone to a beach-themed party, w4hich was attended by a total of 250 people. Construct your 2×2, filling in each cell.
9

In-Class Example: Answer 1
There was a recent outbreak of diarrhea onboard a small cruise ship. Of the 400 passengers, 120 got diarrhea. Among those who got diarrhea, 90 had gone to a beach-themed party, which was attended by a total of 250 people.

D+ D- Total
E+ 90 250
E-
Total 120 400

Let’s start by filling in the numbers we know. I’ve color coded to show you the placement.
10

Calculate other cells
D+ D- Total
E+ 90 250 – 90 = 160 250
E- 120 – 90 = 30 280 – 160 = 120 400 – 250 = 150
Total 120 400 – 120 = 280 400

From here, it’s a matter of subtracting to complete the cells, Note: depending on which cells you computed first, your calculations may look a little different. However, the number should be the same.
Let’s move forward…
11

Relative Risk
Relative Measures of Effect

Note: Odds Ratios are also a relative measure of effect that is used in special circumstances. You’ll learn about odds ratios in a different lesson.
12

RELATIVE RISK
RR = Incidence in Exposed Group (E+)
Incidence in Non-Exposed Group (E-)

13
Relative risk (RR) is the probability that a member of an exposed group will develop the disease relative to the probability that a member of an unexposed group will develop the same disease. RR is a ratio measure. A ratio, recall, is a measurement of something (disease, injury, death, etc. ) in the exposed population divided by the measurement of that same thing in unexposed population. As such, RR tells us how much more (or less) likely one group is to develop disease than the other.
 RR is the generic term referring to:
1) Pb/Pa or
2) Ib/Ia or
3)  Mb/Ma
Where a and b are two populations whose disease characteristics we wish to compare. (Note: we are comparing the same disease in each population).
 
The null value of RR – meaning there is no difference between the two groups you are comparing — is 1.

Calculating RR
1. Present data in a 2 x 2 table
2. Calculate incidence among exposed:
IE+ = a/(a+b)
3. Calculate incidence among unexposed
IE- = c/(c+d)
4. Calculate relative risk
RR = IE+/IE-
5. Interpret results

14
The first step to calculating RR is to set up your epi matrix, and plug in all of your numbers.
Next, you will calculate the incidence (or prevalence, or mortality) among the exposed. This is done by dividing the a cell by the sum of cells a plus b
Then, you will calculate the incidence (or prevalence, or mortality) among the unexposed. This is done by dividing the c cell by the sum of cells c + d.
Now that you have these numbers, calculate RR by dividing the incidence among the exposed by the incidence among the unexposed.
The last step in anything you do is to interpret the results… Interpretation means that you should say WHAT IT MEANS, not merely what you found. I am going to repeat that, because it’s one of the biggest mistakes students make… When I ask you to interpret something, what I am asking is for you to be specific in reporting your findings (how much more or less is it AND what that specific finding means…. More on this in the next few slides…

How to Interpret Relative Risks
When RR = 1.0, it indicates that the risk is the same (equal) for both groups (exposed and unexposed)
When RR > 1.0, it indicates greater risk in the exposed group than in the unexposed group
When RR < 1.0 , it indicates less risk in the exposed group than the unexposed group, and that the factor or exposure may provide some protection against the problem. 15 When RR = 1 (the null), we say there is no association between exposure and disease, or that E and D are independent. Independence means that knowledge of exposure status (E- or E+) gives you no clue as to the disease status (D- or D+) of the individual.   When RR  1, there is an association between exposure and disease, meaning that people who have been exposed have more, or less, probability of getting the disease than those who have not been exposed. Thus, exposure status makes a difference.   If exposed people tend to be those who were diseased, then there is a "positive association" meaning E+ goes with D+ and E- goes with D-. RR will be greater than 1. The amount greater than 1 is how much more likely the exposed are to develop the disease (as compared to the unexposed). Let’s say that RR = 3.4 . The interpretation would be that the exposed were 3.4 times as likely to develop the disease than the unexposed. If the unexposed have less disease, then there is an "inverse association" meaning E+ goes with D- and E- goes with D+. In this situation, RR will be less than 1 and the exposure will be considered to have a protective effect on preventing D (think vaccines). The amount less than 1 is how much LESS common the disease is among the exposed as compared to the unexposed. So, if RR = .75, your interpretation would be: “the exposed were .75 times as likely to get the disease as compared to the unexposed.” IMPORTANT NOTE: Because relative risk is a simple ratio, errors tend to occur when the terms "more" or "less" are used. Because it is a ratio and expresses how many times more probable the outcome is in the exposed group, the simplest solution is to incorporate the words "times the risk" or "times as high as" in your interpretation, rather than using more than, greater than, etc. When you are interpreting RR, you will always be correct by saying: "Those who had (name the exposure) had (insert RR value here) 'times the risk' compared to those who did not have the exposure." Or "The risk of (name the disease) among those who (name the exposure) was (insert RR value here) 'times as high as' the risk of (name the disease) among those who did not (name the exposure)." Another important note: Rarely ever will you get a RR=1, where the incidence is the exact same in the unexposed and the exposed. Instead, you’ll get values near 1. One of the questions I always get asked is how far away from 1 does RR have to be in order for the difference to be meaningful or statistically significant. In a later lesson you will learn how to calculate confidence intervals which answers this question, but for now, just know that the further away from 1 you get, the stronger the association. The closer to 1, the weaker the association. Increased Risk = (RR-1) x 100 Relative Risk Increased Risk 1.0 Equally likely 0% 1.5 1.5 times as likely 50% 2.0 2 times as likely 100% 3.0 3 times as likely 200% Sometimes you may want to report results in terms of how much higher risk the risk is. To do this, you must first account for the RR null being 1. So, you subtract 1 from your RR. Most often, the increased risk is presented as a percentage, so you would then multiply that by 100. Reporting the data this way is called “increased risk” and the formula is (RR-1) X 100. Ok, so let’s do work a problem together…. 16 In-Class Example: RR Are women who take hormone replacement therapy (HRT) at greater risk for ovarian cancer? Here’s the data: Of the 104 women who had ovarian cancer, 27 were on HRT. Of the 2286 women who did not have ovarian cancer, 455 were on HRT. So, let’s work through an example together using these steps… Are women who take hormone replacement therapy (HRT) at greater risk for ovarian cancer? Here’s the data: Of the 104 women who had ovarian cancer, 27 were on HRT. Of the 2286 women who did not have ovarian cancer, 455 were on HRT. 17 Step 1: Display Data In step 1, you set up your epi matrix. Here’s what we know: Of the 104 women who had ovarian cancer, 27 were on HRT. Of the 2286 women who did not have ovarian cancer, 455 were on HRT. As you’ll notice, the scenario isn’t giving you the numbers for each cell. You do, however, have the information you need to fill in each cell. Go ahead and fill in the cells on your own, they check your cell data on the next slide. 18 Check your numbers! Of the 104 women who had ovarian cancer (total D+), 27 were on HRT (cell a: D+, E+). Because we have these two numbers, we can determine how many D+ were not on HRT by subtracting the a cell from the total. This gives us the number for our c cell Of the 2286 women who did not have ovarian cancer (total D-), 455 were on HRT (D-, E+ the “b” cell). That means 1831 were not on HRT – the “d” cell. (total D- minus b cell) With me so far? 19 Steps 2 - 4 Step 2: Calculate I among exposed Step 3: Calucate I among unexposed Step 4: Calculate RR 20 Step 2 IE+ per 1000 = [a/(a+b)] x 1000 = [27/(27 + 455)] x 1000 = (27/482) x 1000 = .0560 x 1000 =56.0 per 1000 Step 3 IE- per 1000 = [c/(c+d)] x 1000 = [77/(77 + 1831)] x 1000 = 77/1908 x 1000 = .0404 x 1000 = 40.4 per 1000 Step 4 RR = IE+/IE- = 56.0/40.4 = 1.4 Step 5: Interpret results RR = 1.4 means that women who used HRT were 1.4 times as likely to develop ovarian cancer as compared women who did not use HRT Another way of stating this is from the increased risk perspective: Women who used HRT had a 40% increased risk of ovarian cancer than those did not use HRT. 21 As I mentioned earlier, it is important that we interpret the data. Numbers– without interpretation – are meaningless. A word of caution that I did not mention earlier is to really check yourself if you use the word “caused” in your interpretations. Remember, epi is an *observational* science and doesn’t prove causation (it shows association or impact), but it’s really tempting to say “cause”… Just keep in mind the whole purpose behind measures of effect– to show how much or less common D is among two groups – and use that to guide your interpretation. Absolute (Attributable) Risk Absolute Measures of Association Absolute = subtraction 22 Absolute Risk (AR) Also known as: Rate difference Risk difference Attributable risk Excess risk 23 One of the most exasperating aspects of epi for me is the lack of common terminology. It makes reading articles very challenging at times. The *proper* term for what you’re about to learn is absolute risk. However, it is so commonly referred to in the literature and epi texts alike as “attributable risk” that I have no choice but to present it that way as well… Ok, I do have a choice, but because I want you to be able to critique epi articles, I’ll use the prevailing terminology. =) Other historical terms for absolute risk include "rate difference," "risk difference,“ and “excess risk.” Note that while "attributable risk" (the historical term for absolute risk of incidence) seems to suggest cause and effect, causation should NOT be inferred because it is a measure of association Attributable Risk Assumptions diseases have multiple causes any single risk factor may be responsible for a fraction of all cases of a given disease the incidence of disease in unexposed persons reflects the background risk (i.e., the incidence due to causes other than the exposure of interest) AR is based on three assumptions – 1. Diseases have many sufficient causes (you should know what I mean!) 2. Any single component cause (risk factor) may be responsible for a portion of all cases of a given disease. And this would be called??? =) 3. Incidence of disease in the UNEXPOSED is “background” risk – that is, it’s due to something other than the exposure of interest… 24 Three ways of E+ becoming D+ within a specified period of time E+ had no impact on becoming D+ E+ had some impact on D+ E+ is sole reason for D+ There are three ways E+ people become D+ during a specified time period: Because people can get D+ without being exposed, then it is possible that some of the people who are D+E+, are D+ for reasons other than E. That is, even though they were exposed, it wasn’t the exposure that caused them to become D+. They become D+ during the time period, but not because of the exposure. This is known as “background incidence.” E+ had some impact on the person becoming D+ during the time period. Maybe the exposure made the person become D+ earlier than he/she would have in the absence of exposure. These people would still have become D+ without the exposure, but at some later point in time, outside of the specified time period. E+ is the only reason the person became D+ during the specified time period. Had the exposure never occurred, the person would not become D+. The D+ in 2 and 3 are called etiologic cases, meaning the exposure is a contributing factor to becoming D+. The D+ in 3 are also called “excess cases” because they would not have gotten the disease had they not been exposed. All excess cases are etiologic cases, meaning the exposure played some role. But not all etiologic cases are excess cases, because in order to be “excess,” the exposure was the only thing responsible for becoming D+. =) This would make a great test question. Thus, IE+ = etiologic cases (cases due to exposure) PLUS cases due to some other factor (background incidence) despite exposure 25 Attributable (Absolute) Risk Attributable Risk (AR) compares difference in incidence, prevalence, or mortality among exposed to unexposed Provides information about the absolute effect of the exposure, or the excess risk in exposed versus non-exposed So, back to AR. Because attributable risk looks at the difference in disease frequency between the exposed and unexposed, it implies that not all of the disease incidence is due to the exposure, as even some non-exposed individuals develop the disease. Thus, we calculate AR to remove this risk of disease that would have occurred anyway due to other causes. Attributable Risk (AR) answers two basic questions that are the flip sides of the same coin: How many cases of disease Y among the exposed are attributable to exposure X? How many cases of disease Y among the exposed can be eliminated by eliminating exposure X? 26 Attributable (absolute) Risk 2 Quantifies disease incidence (or disease risk) which can be attributed to a specific exposure Quantifies risk of disease attributable to the exposure (incidence of exposed) by removing the risk of disease that would have occurred anyway (incidence of unexposed) AR = IE+ minus IE- 27 Recall that absolute refers to subtraction. As such, AR is the difference between two measures of disease frequency in the exposed and unexposed. This difference indicates how much risk can be attributed to that exposure. Thus, to calculate AR, we simply subtract the incidence (or prevalence) of the unexposed from the incidence (or prevalence) of the exposed. The difference is the “absolute risk.” It is interpreted as XX number of cases among the exposed are attributable to the exposure. Or XX number of cases among the exposed could be eliminated by removing the exposure. Interpreting AR When AR = 0, there is no difference in risk between the exposed and unexposed When AR > 0, the risk is greater in the exposed than unexposed
When AR<0, the risk is less in the exposed than unexposed When using absolute measures, the null value is zero, indicating that the measures of disease frequency (I,P,M) are identical in the two groups (E+, E-). Thus, if AR = 0, it means that none of the disease cases could be eliminated if we were able to get rid of the exposure of interest. Or, there are no cases of the disease attributable to the exposure. Numbers greater than 0 indicate that the disease is more common in the exposed as compared to the unexposed. The amount over 0 is the number of cases that are attributable to the exposure. Another way to state this is that it is the number of cases among the exposed that could be prevented if we eliminated the exposure. Numbers less than 0 indicate a protective effect (or therapeutic effect). The amount under 0 are the number of cases that were prevented by the exposure (example: fluoride and cavities) 28 Going back to the HRT and Ovarian Cancer Example.. IE+ = 56 per 1000 IE- = 40.4 per 1000 AR = IE+ - IE- = 56 – 40.4 = 15.6 per 1000 Interpretation: 15.6 cases of ovarian cancer per 1000 women who take HRT are attributable to HRT use. (Or, 15.6 cases of ovarian cancer per 1000 women who take HRT could be eliminated if we eliminated HRT use). Let’s say we want to know how many cases of ovarian cancer per 1000 women who use HRT are attributable to HRT. Because we’ve already calculated our IE+ and IE- a few slides ago, we’ll just be lazy and plug those numbers in here. So, what we do is simply subtract IE- from IE+ and then to get our answer then interpret! Just remember, that what we are looking at here is how many cases -- among the exposed -- are attributable to the exposure. Pretty easy, eh? AR is sometimes expressed as a proportion (%) of the disease incidence in the exposed, which is the proportion of disease incidence in the exposed that is due to the exposure. You’ll learn about how this is done in the next section. =) 29 Attributable Proportion Population Attributable Risk Population Attributable Risk Percent Impact Measures of Association Attributable Proportion AKA Attributable Risk Percent Proportion of the disease WITHIN the exposed group that is attributable to exposure 31 I promised you you’d see AP again! =) Remember how I said a few slides ago that attributable (AR) is sometimes expressed as a proportion (%) of the disease incidence in the exposed? Well, this is how we do it. Attributable proportion, also known as the attributable risk percent, is a measure of the public health impact of a causative factor. The calculation of this measure assumes that the occurrence of disease in the unexposed group represents the baseline or expected risk for that disease. It further assumes that if the risk of disease in the exposed group is higher than the risk in the unexposed group, the difference can be attributed to the exposure. Thus, the attributable proportion is the proportion of disease in the exposed group attributable to the exposure, and represents the expected percent reduction in disease if the exposure could be removed (or never existed). Appropriate use of attributable proportion depends on a single risk factor being responsible for a condition. When multiple risk factors interact (e.g., physical activity and age), this measure may not be appropriate. Calculating Attributable Proportion AP = (Attributable Risk ⁄ Incidence of the exposed) x 100 = [(IE+ minus IE-) / IE+] x 100 = {[(a/a+b) – (c/c+d)] / (a/a+b)} x 100 Alternative Formula AP=[ (RR-1)/ RR] x 100 32 So, to calculate AP, we divide the attributable risk (IE+ minus IE-) by the incidence of the exposed (IE+). Alternatively, you can use the alternative formula for AP which is (RR – 1) divided by RR. If you’d like to know how this alternative theory was derived, I can refer you to some resources that explain it, but for the purposes of this course, just trust me that this is an alternative way of computing AP. =) Note: Because AP shows impact, we multiply it by 100 to show the percentage of cases among the exposed that are attributable to E+. Thus, as mentioned on an earlier slide, AP is also known as an “Attributable Risk Percent” which is typically written as AR% When we interpret AP, we state that if we eliminate the exposure, we would get rid of XX% of the disease among the exposed. Let’s look at our HRT example. Now, is this the same as prevention? Well…….. HRT and Ovarian Cancer… again AP = (AR /IE+) x 100 = (15.6/ 56) X 100 = .279 X 100 = 27.9% Interpretation: If HRT was not used, we would eliminate 27.9% cases of ovarian cancer among those using HRT. Seeing as how we’ve already completed the calculations for AR and IE+, I’m just going to zip up to the slide showing those numbers then plug them in here. Or, I could have done it like this, AP = [(RR-1)/RR] x 100, and flipped up a few more slides to get my RR (1.4) So, AP =[ (1.4 – 1) / 1.4] x 100 = (.4/1.4) x 100 = .286 x 100 = 28.6% Note The difference in the answers (27.9% and 28.6%) is due to rounding at each phase of calculations. The rest is just interpretation. Remember, we are still only looking at the impact of the exposure among the exposed, so that should always be mentioned in your interpretation… I know some of you are thinking, “Well, what if we wanted to know the impact of HRT on ovarian cancer in the overall population?” 33 Population Attributable Risk (PAR) If we eliminated the exposure, what would happen to the disease burden in the whole population? Number of cases in the population that can be attributed to the exposure Public health impact of exposure 34 From a public health perspective we are interested in both the benefits of an intervention to the exposed group and to the whole community. PAR is a similar measure to AR except it is concerned not with the excess rate of disease in the exposed, but rather with the excess rate of disease in the population. PAR is the disease incidence in the population (i.e. exposed and non-exposed) that is due to the exposure. Therefore it is the disease incidence in the population that would be eliminated if the exposure were eliminated PAR estimates the public health impact of a particular exposure. It is used to estimate the number of disease cases that could be prevented if that exposure was eliminated. PARs are thus important in judging public health priorities because it helps determine which exposures have the most relevance to the health of a community. This of interest to policy makers and those responsible for funding prevention programs. Calculating PAR PAR = Incidence in total population MINUS Incidence in unexposed = [(a + c) / (a + b +c + d) ] - [c /( c+d)] To calculate PAR, we subtract the incidence of the unexposed (E-) from the incidence in the total population. PAR = I total – IE- The long formula on this slide relates to the epidemiological matrix... Let’s take a look at the HRT example… 35 Ovarian Cancer & HRT The incidence per 1000 in the total population = [(a + c) / (a + b +c + d) ] x 1000; thus, I total = 104/2390 x 1000 = .0435 x 1000 = 43.5 per 1000 Interpretation: 43.5 cases of ovarian cancer per 1,000 persons taking HRT The incidence of the unexposed = [c /( c+d)] x 1000 IE- = 77/1908 x 1000 = .0404 x 1000 = 40.4 Interpretation: 40.4 cases of ovarian cancer per 1,000 persons not taking HRT PAR = I total - IE- = 43.5 – 40.4 = 3.1 Interpretation: 3.1 cases of the 43.5 per 1000 of ovarian cancer in the population are attributable to HRT…. Or 3.1 cases of the 43.5 cases per 1000 could be eliminated in the population by not using HRT. 36 Population Attributable Risk Percent (PAR%) Proportion of cases in the population that can be attributed to the exposure Just as AR is sometimes expressed as a proportion (%) of the disease incidence in the exposed, so too can PAR… PAR% is the proportion disease incidence in the population that is due to the exposure. Therefore, it is the percentage of disease incidence in the population that would be eliminated if the exposure were eliminated 37 [(a + c) / (a + b +c + d) ] - [c /( c+d)] (a + c) / (a + b +c + d) This gets multiplied by 100 Calculating PAR% PAR % = [(Incidence in total population MINUS Incidence in unexposed) divided by incidence in total population] x 100 38 Despite the long scary formula on the slide, PAR% is just PAR divided by the incidence of disease in the total population. Using our HRT data, we’d just take the PAR and divide it by the Itotal and multiply it all by 100. PAR% = PAR/Itotal X 100 =3.1/43.5 x 100 = .071 x 100 = 7.1 To interpret, we say that 7.1% of ovarian cancer in the population is attributable to HRT. Or, 7.1% of ovarian cancer in the population could be prevented if HRT were not used. PUTTING IT ALL TOGETHER In-Class Example: Aspiration & Pneumonia Of the 143 hospitalized patients with pneumonia, 102 aspirated. In comparison, 298 of the 1274 hospitalized patients without pneumonia aspirated. After setting up your epi matrix, please calculate and interpret the following per 100: Relative Risk Absolute Risk Attributable Proportion (AR%) Population Attributable Risk Population Attributable Risk Percent . Ok, so let’s put everything together and look at some fake data on pneumonia (D) among people who aspirated (E+) and those who didn’t (E-). Grab a pencil, paper, and calculator and once you’re finished, move on to the next slide to check your answers. Of the 143 hospitalized patients with pneumonia, 102 aspirated. In comparison, 298 of the 1274 hospitalized patients without pneumonia aspirated. After setting up your epi matrix, please calculate and interpret the following per 100: Relative Risk Absolute Risk Attributable Proportion (AR%) Population Attributable Risk Population Attributable Risk Percent 40 Step 1: Set Up Epi Matrix Of the 143 hospitalized patients with pneumonia (total D+), 102 aspirated (cell a). In comparison, 298 (cell b) of the 1274 hospitalized patients without pneumonia (total D-) aspirated 41 RR (IE+, IE-) AR AR% PAR PAR% Calculations 42 IE+ = (a/a+b) x 100 = 102/400 x 100 = 0.255 x 100 = 25.5 cases of pneumonia per 100 patients in the exposed patients IE- = (c/c+d) x 100 = 41/1017 x 100 = .040 x 100 = 4.0 cases of pneumonia per 100 patients in the unexposed group RR = IE+/ IE- = 25.5/4 = 6.4 Patients with aspiration were 6.4 times as likely to develop pneumonia than patients without aspiration AR = IE+ - IE- =25.5 - 4 = 21.5. Among those with pneumonia who had aspirated, 21.5 cases of pneumonia per 100 patients are attributed to aspiration. Put another way, if we eliminated aspiration we would eliminate 21.5 cases of pneumonia per 100 patients among those who had aspirated. AP (AR%) = AR/ IE+ X 100 = 21.5 /25.5 x 100 = 84.3%: If we eliminated aspiration, we would eliminate 84.3% of the pneumonia among those who had aspirated. Note: It would be WRONG to say that aspiration is RESPONSIBLE for (caused, etc.) 84.3% of pneumonia in patients. PAR = I total – IE- (where, I total = [(a + c) / (a + b +c + d) ] x 100 = 143/1417 x 100 = .1 x 100 = 10) = 10 - 4 = 6 6 cases per 100 cases of pneumonia in the population are attributable to aspiration…. Or 6 cases of pneumonia per 100 in the population could be prevented by eliminating aspiration PAR% = PAR/I total x 100 = 6/10 x 100 = .6 x 100 = 60% If we eliminated aspiration, we would eliminate 60% of pneumonia in the population. Had enough? Go do your practice exercises. =) 43