A Review of Disease Risk Scores and Their Application in Pharmacoepidemiology

Abstract

Summary scores that reduce baseline covariate information to a single value have become standard tools for confounding control in pharmacoepidemiologic studies. The propensity score (PS) summarizes covariate associations with treatment assignment and has been the most widely used summary score for confounding control. An alternative to the PS is the prognostic score, often referred to as the disease risk score (DRS). Instead of summarizing covariate associations with treatment, the DRS summarizes covariate associations with potential outcomes. Adjustment based on the DRS has unique challenges and limitations compared to PS adjustment, but the DRS also has advantages over the PS in certain settings. In this paper, we review the recent developments and applications of DRSs. We discuss differences between the PS and the DRS as well as the benefits and challenges of using the DRS for confounding control. Finally, we discuss areas for future research and development for the application of risk scores in pharmacoepidemiology and nonexperimental medical studies.