Confounding And Effect Measure Modification

Front 1

Confounding

Back 1

Systematic difference between the groups being compared that distorts the true association between an exposure and disease; unlike bias, it is an inherent characteristic of the population

Front 2

Counterfactual Ideal

Back 2

Ideal comparison group would be the exact same people who are in the exposed group had they not been exposed

Front 3

How is confounding related to the counterfactual ideal?

Back 3

Confounding can be thought of as a failure to come close to the counterfactual ideal

Front 4

Three Criteria for a Confounder

Back 4

(1) Independent predictor of the outcome →

The confounder is a risk factor for disease among people who are unexposed

(2) Associated with the exposure <-->

The confounder occurs more or less often among the exposed than the unexposed.

(3) Cannot be an intermediate on the causal pathway between exposure and disease

The confounder cannot be caused by the exposure

Front 5

Effect Measure Modification

Back 5

The strength of the association between an exposure and disease differs according to the level of another variable; biological process worthy of investigation; evaluated, like confounding, using stratified analysis; occurs when stratum-specific estimates are meaningfully different from one another

Front 6

Randomization

Back 6

Randomly allocate study subjects to treatment groups so each subject has an equal chance of being assigned to the treatment group or comparison group

Front 7

Residual Confounding

Back 7

Confounding that persists despite efforts to control or adjust for confounding

Front 8

Restriction

Back 8

Limit study to people who are within one category of the confounder

Front 9

Stratification

Back 9

Stratify (separate) your study population into subgroups where one group has the confounder characteristic and one group does not. Then calculate a measure of association for each subgroup. Always use stratification!

Front 10

6 Steps of Stratification

Back 10

(1) Calculate crude measure of association

(2) Divide subjects into strata (of the confounder)

(3) Calculate stratum-specific measures of association

(4) Calculate adjusted measure of association

(5) Determine whether crude measure differs from adjusted measure of association, and by how much (magnitude)

(6) Determine if there is effect measure modification

Front 11

Matching

Back 11

Select study subjects so that confounders are distributed identically among the exposed and unexposed (cohort study) or cases and controls (case-control study)

Front 12

Multivariate Regression

Back 12

Involves construction of a statistical model (requires computer) that describes the association between exposure, disease, and confounder