Confounding in health research

Consideration of confounding is fundamental to the design, analysis, and in terpretation of studies intended to estimate causal effects. Unfortunately, the word confounding has been used synonymously with several other terms, and it has been used to refer to at least four distinct concepts. This pape r provides an overview of confounding and related concepts based on a count erfactual model of causation. In this context, which predominates in nonexp erimental research, confounding is a source of bias in the estimation of ca usal effects. Special attention is given to the history of definitions of c onfounding, the distinction between confounding and confounders, problems i n the control of confounding, the relations of confounding to exchangeabili ty and collapsibility, and confounding in randomized trials.