Across 4 large cardiovascular datasets, propensity score methods were not necessarily superior to conventional covariate adjustment, with stratification performing poorly with few outcome events.
Observational
Yes
Propensity score methods are not necessarily superior to conventional covariate adjustment in observational studies, highlighting the need to carefully select the most suitable statistical method.
Propensity scores (PS) are an increasingly popular method to adjust for confounding in observational studies. Propensity score methods have theoretical advantages over conventional covariate adjustment, but their relative performance in real-word scenarios is poorly characterized. We used datasets from 4 large-scale cardiovascular observational studies (PROMETHEUS, ADAPT-DES the Assessment of Dual AntiPlatelet Therapy with Drug-Eluting Stents, THIN The Health Improvement Network, and CHARM Candesartan in Heart Failure-Assessment of Reduction in Mortality and Morbidity) to compare the performance of conventional covariate adjustment with 4 common PS methods: matching, stratification, inverse probability weighting, and use of PS as a covariate. We found that stratification performed poorly with few outcome events, and inverse probability weighting gave imprecise estimates of treatment effect and undue influence to a small number of observations when substantial confounding was present. Covariate adjustment and matching performed well in all of our examples, although matching tended to give less precise estimates in some cases. PS methods are not necessarily superior to conventional covariate adjustment, and care should be taken to select the most suitable method.
Elze et al. (2017) conducted an observational in Cardiovascular disease. Propensity score methods (matching, stratification, inverse probability weighting, PS as covariate) vs. Conventional covariate adjustment was evaluated on Performance of statistical adjustment methods. Across 4 large cardiovascular datasets, propensity score methods were not necessarily superior to conventional covariate adjustment, with stratification performing poorly with few outcome events.