Online Program

Comparisons of Propensity Score Analysis for Analyzing Rare Binary Outcome (303626)

 View Presentation

Keywords: observational study, propensity score, rare outcome, stratification, matching, doubly-robust estimation

Propensity score methods are being actively implemented to reduce selection bias in observational studies. The most frequently used methods are: covariate adjustment using the propensity score, propensity score stratification, propensity score matching (1:1 matching and many-to one matching), and doubly-robust estimation. Comparison of these methods have been carried out in the existing literature. However, there has been relatively little information on the comparisons in the case of rare event despite the fact that these are commonly occurring situations. Therefore, this study aims to compare the performance of the aforementioned propensity score based methods in the rare event setting. Each method will be introduced in detail along with the corresponding balance diagnosis. Other relevant issues such as whether odds ratio or risk difference should be used or not as the treatment effect measure will also be discussed. We use real data example for illustration to compare anastomotic leak after laparoscopic and robotic gastrectomy revision. Practical guide in performing propensity score analysis using rare events data is provided.