Confounding by Indication and Related Concepts

Abstract

The term confounding by indication is increasingly used in the literature, although the concept has lost much of its original meaning. The literature includes instances where confounding by indication is equated with confounding in general or reverse causality, and other instances where it is used to refer to confounding by contraindication. In this paper, we review concepts related to confounding by indication in studies of intended effects, confounding by indication and contraindication in studies of unintended effects, and confounding by extraneous aspects of the indication. We also discuss non-experimental methods to address confounding by indication, including design strategies (e.g., restriction to a domain free of the indication) and statistical techniques (e.g., propensity score matching). We argue for greater conceptual and semantic clarity with regard to the different forms of confounding by indication and conclude that non-experimental methods do not adequately address this bias in studies of intended effects when the indication defies quantification.