Confounding as a Result of Differential Diagnosis

Abstract

Confounding occurs when a competing cause of illness is associated with one or more of the factors being examined under the study hypothesis. Factors that influence the likelihood of diagnosis and are associated with exposure will also have confounding properties in registry-based follow-up studies without ad hoc diagnosis, and in case-control studies where cases are ascertained from medical settings. Such confounders can be controlled in the same manner as other confounders if the necessary information has been collected in the study. Oftentimes sufficient control of confounders will not be possible, especially in registry-based studies. This article gives guidelines for estimating the direction and magnitude for both forms of confounding. The formulas may be needed for the interpretation of positive or negative results in studies where control for confounding has been incomplete.