Module 14

Question 1

Stratification analysis

Answer 1

Performed to evaluate the effect of an exposure within strata (levels) of the confounder

Question 2

Advantages of stratification

Answer 2

Performing analyses within strata is a direct and logical strategy
Minimum statistical assumptions must be satisfied for the analysis to be appropriate
The computational procedure is straightforward

Question 3

Disadvantages of stratification

Answer 3

Small numbers of observations in some strata
A variety of ways to form strata with continuous variables (like age)
Categorization produces loss of information
Difficulty in interpretation when several confounding factors must be evaluated

Question 4

Sometimes the confounding is only _____, or can ________

Answer 4

Partial, go in the opposite direction

Question 5

What is the key with stratification?

Answer 5

Do separate analyses within categories (or strata) of the confounder and compare these odds ratios (or relative risks) to the original (unadjusted) one

Question 6

Interaction

Answer 6

An interaction effect between two variables is defined as one in which the effect of one variable depends on the level of the other variable

Question 7

What type of term is interaction?

Answer 7

Statistical

Question 8

What type of term is effect measure modification?

Answer 8

Epidemiologic

Question 9

Effect measure modification

Answer 9

When the magnitude of effect of an exposure varies by levels of a third variable

Question 10

Interaction vs confounding

Answer 10

In both situations, you have a third variable, and your crude results are different than the results you obtain when you stratify on this third variable
If it is confounding, the stratified results are close to each other
If it is interaction or effect modification, they are substantially different from both the crude and each other

Question 11

How do you tell when the difference between the stratum-specific measures of association are big enough to suggest interaction?

Answer 11

You often make a subjective judgment, although there are statistical tests to help

Question 12

Do we want to eliminate interaction as much possible?

Answer 12

No. Whereas we want to prevent or eliminate confounding, interaction is something we want to discover, if it exists. It is not a bias but a genuine part of the results.

Question 13

What to do when there is both confounding and interaction

Answer 13

Confounding is irrelevant
Report the interaction

Question 14

The additive model

Answer 14

Departures from additive effects are of public health interest
Can be expected to fit (not exceed) the additive model if causal mechanisms are completely independent

Question 15

Synergism

Answer 15

The combined effect is greater than what one would expect if the exposures acted independently (i.e., the sum of the effects of both)

Question 16

How to determine if there is synergism

Answer 16

Add both of the relative risks for only one factor then subtract the background risk, then compare it to the relative risk for those who have both factors. If the combined relative risks for one factor are less than the relative risk for both factors, then there is a synergistic effect.

Question 17

What are the two models to determine whether there is a synergistic effect between two risk factors?

Answer 17

Additive and multiplicative

Question 18

How to use the multiplicative model to determine synergistic effect/interaction?

Answer 18

Multiply each relative risk for one individual factor and compare the total to the relative risk for both combined.