Propensity Score Matching: Retrospective Randomization?

• DOI: 10.1053/j.jfas.2017.01.013
• Publication Date: 2017-02-22T05:50:38Z
• Authors: Daniel C. Jupiter
• Keywords: Random Allocation; 03 medical and health sciences; 0302 clinical medicine; Research Design; Case-Control Studies; Humans; Propensity Score; Randomized Controlled Trials as Topic; 3. Good health

Randomized controlled trials are viewed as the optimal study design. In this commentary, we explore the strength of this design and its complexity. We also discuss some situations in which these trials are not possible, or not ethical, or not economical. In such situations, specifically, in retrospective studies, we should make every effort to recapitulate the rigor and strength of the randomized trial. However, we could be faced with an inherent indication bias in such a setting. Thus, we consider the tools available to address that bias. Specifically, we examine matching and introduce and explore a new tool: propensity score matching. This tool allows us to group subjects according to their propensity to be in a particular treatment group and, in so doing, to account for the indication bias.