Propensity Score Methods: A Robust Approach for Causal Inference in Observational Data (Motivated by the Comparative Study on Anesthesia Type and Hip Arthroplasty Outcomes by Lee et al.)

Propensity score methods provide a robust statistical framework for drawing causal inferences in observational healthcare data where randomized controlled trials are infeasible. By balancing covariates between treatment groups, these methods mitigate confounding bias in non-randomized comparative effectiveness studies and simulate the effects of randomization. This report outlines key concepts in propensity score methodology, including matching, stratification, weighting, and covariate adjustment, and their practical application in healthcare research. Using the study by Lee et al. (2025), which compares general versus regional anesthesia in patients undergoing total hip arthroplasty, the report demonstrates how these methods can enhance validity and improve causal interpretations in real-world data analyses.In Lee et al.’s study, propensity score matching (PSM) successfully reduced baseline differences across treatment groups, allowing for more credible comparisons of postoperative outcomes. The authors utilized logistic regression to estimate treatment probabilities, applied nearest-neighbor matching with a caliper, and verified covariate balance through standardized mean differences (SMDs). Results showed a lower incidence of complications in the regional anesthesia group, supported by sensitivity analyses using multivariable models and inverse probability weighting (IPTW). The report concludes with a discussion of strengths, limitations, and future directions, including integrating machine learning and real-world electronic health record (EHR) data to improve balance and generalizability. Propensity score methods remain essential for causal inference in modern clinical, surgical, and epidemiological research.