Measures of Association

Question 1

How can a 2x2 table be used to measure the strength of association of exposure to illness?

Answer 1

• calculate the risk of those exposed and diseased compared to those exposed and not diseased
• calculate the risk of those not exposed and diseased compared to those not exposed and not diseased

Question 2

What is the risk upon exposure and no exposure? What is the overall incidence risk?

Answer 2

overall incidence risk (or prevalence) is 0.3%, or about 3 per 1000 cats

Question 3

What is a risk ratio? How is it calculated?

Answer 3

relative risk, incidence risk ratio comparing the risk of exposure to no exposure - compares the risk of a health event (disease, injury, risk factor, or death) among one group with the risk among another group

Question 4

What is the risk ratio? How is the result translated?

Answer 4

the risk of FLUTD is 4x higher for cats on dry food diet

Question 5

What is an odds ratio? How is it calculated?

Answer 5

odds of disease in exposed group compared to odds of disease in a non-exposed group giving the magnitude of disease

Question 6

What is the odds ratio? How is the result translated?

Answer 6

the odds of FLUTD is 4x higher for cats on a dry food diet

Question 7

How do odds ratios compare to risk ratios?

Answer 7

can be applied to all study designs

Question 8

What does it mean when a risk ratio or odds ratio is equal to 1, > 1, and < 1?

Answer 8

no effect

increased risk

reduced risk (protetive)

Question 9

In what 2 situations is it common for the risk ratio = odds ratio?

Answer 9

1. disease is rare in underlying population with a prevalence < 5%
2. risk ratio is close to null (R=1)

Question 10

How are measures of effect expressed? What do they assume?

Answer 10

absolute effect measures related more to a number of cases an exposure caused, rather than the relative strength (ratio)

causality - “effect” will be the result of exposure

Question 11

What are measures of effect computed for?

Answer 11

• exposed group
• population

Question 12

What is the risk difference? How is it calculated?

Answer 12

attributable risk - difference in risk of disease in exposed and not exposed groups

Question 13

What is the risk difference/attributable risk? How are the results translated?

Answer 13

for every 1000 cats on dry food, 4.5 had FLUTD because of the dry food

Question 14

What is attributable fraction? How is it calculated?

Answer 14

preventable fraction, proportion of D+ in E+ due to exposure

Question 15

What is the attributable fraction? How is the result translated?

Answer 15

75% of FLUTD cases in the cats on dry food diets were due to them being on dry food

Question 16

How should the attributable fraction be calculated when the “exposure” is actually protective? In what situations is this commonly done?

Answer 16

when E+ is protective, it becomes a preventable fraction, so E+ and E- should be flipped so that AF/PrevF will always be between 0 and 1

vaccines

Question 17

When are population-level estimates done? When is it useful?

Answer 17

extrapolate results to a whole population, instead of focusing on the effect on the exposed group - remove exposure from all of the population and see how much disease can be prevented

population-level recommendations and policy making

Question 18

When are population-level estimates appropriate? What happens if this situation is not met?

Answer 18

if the proportion of exposed is the same in the study as in the population - source population exposure information must be available

if exposure is rare, applying mitigation strategies will have a very small impact on the population

Question 19

What is population attributable risk? How is it calculated?

Answer 19

increase or decrease in risk of disease in a population that is attributable to exposure

Question 20

What is the best way to express population attributable risk?

Answer 20

absolute units - easily misinterpreted as a percentage

Question 21

What is the population attributable risk? How are the results expressed?

Answer 21

we would expect the risk of FLUTD to decrease by 1.8 cases per 1000 if dry food was not fed to cats

for every 1000 cats in this population, 1.8 had FLUTD because they ate dry food

Question 22

What is population attributable fraction? How is it calculated?

Answer 22

etiological fraction

proportion of disease in the study population that is due to exposure

Question 23

What is the populaiton attributable fraction? How are the results expressed?

Answer 23

54% of FLUTD cases in the cat population was attributable to them eating dry food

Question 24

How should each result be expressed?

Answer 24

Question 25

What kind of associations do 2x2 tables calculate? What binomial distribution statistical tests can be used to calculate the significance?

Answer 25

unconditional associations not accounting for other exposures or confounders

• Pearson’s χ2
• Fischer’s Exact test
• ensures what was observed was not due to chance

Question 26

What model is used to account for multiple predictors?

Answer 26

logistic regression models - estimate ORs on a ln scale

Question 27

What binomial distribution is preferred when using 2x2 tables? Why? What is assumed to be the null hypothesis?

Answer 27

Fisher’s Exact test - appropriate for small numbers and even zero counts (1-tailed = RR >/< 1; 2-tailed = unknown RR)

no significant difference (p<0.05 = significance!)

Question 28

What is the p-value conventionally set at? What does this mean?

Answer 28

< 0.05

less than 1/20 chance that the observed values occurred from chance if there were truly no differences between exposure groups

Question 29

What can be assumed if the 2x2 table is statistically different using the Fisher’s Exact test?

Answer 29

all measures of association (effect) will be statistically significant —> CIs will also show significance