Lectures 11-12: Confounding and Effect Modification

Question 1

Confounding variable

Answer 1

A 3rd variable that distorts an observed relationship between exposure and outcome. A confounder has a relationship with both exposure and outcome, but is not part of the causal pathway

We CONTROL for confounders

Question 2

Effect modifier

Answer 2

A 3rd variable that modifies the magnitude of an association by varying it within different levels

We REPORT and DESCRIBE effect modifiers as they can not be controlled

Question 3

Randomization

Answer 3

Method to control for confounding: Allocates equal number of subjects with confounders in each group by random sampling

Not always practical, effective when sample size is large

Question 4

Restriction

Answer 4

Method to control for confounding: study participation restricted to only subjects who do not fall within confounding categories

Straight forward, convenient, inexpensive; may limit ability to extrapolate upon results

Question 5

Matching

Answer 5

Method to control for confounding: subjects are selected in matched pairs related to confounding variable to equally distribute confounder among each study group

Intuitive, but difficult to accomplish/expensive/time consuming

Question 6

Stratification

Answer 6

Method to control for confounding: statistical analysis evaluating association between exposure/outcome within strata of confounding variable

Intuitive/straight forward; not successful when multiple confounding variables are present

Question 7

Multivariate analysis

Answer 7

Method to control for confounding: mathematically factoring out effects of confounding variables

Controls for multiple confounders; data is more difficult to interpret