Methods for analyzing secondary outcomes in public health case-control studies

Abstract

Case-control studies are common in public health research. In these studies, cases are chosen based on the primary outcome but there are usually many other related variables which are collected. While the analysis of the association between the primary outcome and exposure variables is generally the main focus of the study, the association between secondary outcomes and exposure variables may also be of interest. Since the experiment was designed for the analysis of the primary outcome, the analysis of secondary outcomes may suffer from selection bias. In this chapter we will introduce the problem and the potential biased inference that can result from ignoring the sampling design. We will discuss and compare a design-based and model-based approach to account for the bias, and demonstrate the methods using a public health data set.