Confounding & Effect Modification

This lesson is part of Premium

Confounder = associated with exposure AND independently with outcome, not on causal pathway (classic: coffee–smoking–CHD). Control: design (randomisation, restriction, matching) and analysis (stratification, multivariable regression, adjustment). Effect modification/interaction (effect genuinely differs across subgroups) — distinct from confounding. Diagram: confounding triangle. Exam tip: randomisation handles unknown confounders; adjustment only known ones.

Premium unlocks all 40 lessons, every calculator and the timed mock.