Confounding

Question 1

Confounding

Answer 1

Observation correct, but explanation wrong bc investigators failed to measure other variables. Highly associated w/ exposure, but also independently associated w/ outcome (but not in causal pathway)

Question 2

5 Things that Control for Confounding (3 in design phase, 2 in analysis phase)

Answer 2

Design: Randomization, restriction (inclusion/exclusion criteria), matching
Analysis: Stratification, multivariate analysis

Question 3

Stratum Specific Analysis

Answer 3

Separate out E and O tables by 3rd potentially confounding variable (so just do your 2x2, but do it for 3rd variable exposed group and 3rd variable non-exposed group). 1st test that 3rd variable is associated w/ both exposure and outcome w/ their own 2x2 tables

Question 4

2 Criteria to be Confounding by SS Analysis

Answer 4

Crude OR/RR different from SS ones

SS OR/RR’s similar to each other

Question 5

Adjusted RR

Answer 5

Takes confounders into account - the RR from the stratum specific groups

Question 6

Effect Modification

Answer 6

Change in strength of association b/w E and O based on level of a 3rd variable. SS OR/RRs differ from crude, but not similar to each other

Question 7

Mediation

Answer 7

Variable in the causal pathway b/w E and O. Not a confounder