Performance of propensity score methods in observational studies: a systematic review

Area of Science:

• Epidemiology and Biostatistics
• Health Research Methodology
• Statistical Inference

Background:

• Observational studies are crucial in health research, but estimating treatment effects requires robust methods.
• Existing research on propensity score (PS) methods is fragmented, necessitating a comprehensive synthesis.
• Guidance for researchers on selecting appropriate PS methods is critically needed.

Purpose of the Study:

• To systematically review and synthesize evidence on the performance of various propensity score methods.
• To provide practical recommendations for researchers using PS methods in observational studies.
• To identify the most effective PS techniques for estimating treatment effects.

Main Methods:

• A narrative synthesis systematic review adhering to PRISMA guidelines.
• Inclusion of peer-reviewed studies published between 2000 and 2025.
• Evaluation of studies based on treatment effect estimation, bias, mean squared error (MSE), and confidence interval (CI) coverage.

Main Results:

• Propensity score matching (PSM) and inverse probability of treatment weighting (IPTW) generally enhance covariate balance and CI coverage.
• Caliper-based optimal and full matching effectively reduce bias for odds ratios (ORs) and marginal hazard ratios (MHR).
• Doubly robust (DR) methods combined with IPTW demonstrate low MSE and reliable inference, while PS stratification often shows poorer performance.

Conclusions:

• Propensity score matching (PSM), doubly robust inverse probability of treatment weighting (DR-IPTW), and full matching are highly effective.
• These methods excel at reducing bias, achieving covariate balance, minimizing MSE, and improving precision.
• The findings offer practical guidance for selecting optimal PS methods in observational research.