The article by Peter C. Austin explores the optimal caliper width for propensity-score matching in observational studies, focusing on estimating differences in means and risk differences. Propensity-score matching is a method used to adjust for systematic differences between treated and untreated subjects when estimating treatment effects. The study uses Monte Carlo simulations to examine the impact of different caliper widths on bias reduction, mean squared error (MSE), confidence interval coverage, and type I error rates. Key findings include: - For estimating differences in means, calipers of width 0.2 times the standard deviation of the logit of the propensity score minimize MSE and eliminate at least 98% of bias. - For estimating risk differences, calipers of width 0.05 to 0.30 times the standard deviation of the logit of the propensity score minimize MSE and improve confidence interval coverage. - When all covariates are binary, calipers of width 0.8 times the standard deviation of the logit of the propensity score minimize MSE. - The choice of caliper width has minimal impact on performance when all covariates are binary. The study also includes a case study using data from patients with heart failure, where different caliper widths resulted in qualitatively similar estimates of the absolute reduction in mortality due to β-blocker prescription. The article concludes with recommendations for researchers to optimize the performance of propensity-score matching.The article by Peter C. Austin explores the optimal caliper width for propensity-score matching in observational studies, focusing on estimating differences in means and risk differences. Propensity-score matching is a method used to adjust for systematic differences between treated and untreated subjects when estimating treatment effects. The study uses Monte Carlo simulations to examine the impact of different caliper widths on bias reduction, mean squared error (MSE), confidence interval coverage, and type I error rates. Key findings include: - For estimating differences in means, calipers of width 0.2 times the standard deviation of the logit of the propensity score minimize MSE and eliminate at least 98% of bias. - For estimating risk differences, calipers of width 0.05 to 0.30 times the standard deviation of the logit of the propensity score minimize MSE and improve confidence interval coverage. - When all covariates are binary, calipers of width 0.8 times the standard deviation of the logit of the propensity score minimize MSE. - The choice of caliper width has minimal impact on performance when all covariates are binary. The study also includes a case study using data from patients with heart failure, where different caliper widths resulted in qualitatively similar estimates of the absolute reduction in mortality due to β-blocker prescription. The article concludes with recommendations for researchers to optimize the performance of propensity-score matching.