Table 2. Comparison of Commonly Used Propensity Score Analysis Methods.

PS Method	Advantages	Disadvantages
Matching	Allows intuitive analysis and transparent presentation of covariate balance Removes covariate imbalance more effectively (less bias) than other stratification or covariate adjustment Does not require specification of the PS-outcome association	Limits generalizability and reduces power by excluding unmatched patients Subject to residual and unmeasured confounding
Weighting	Can be extended to account for time-dependent confounding and censoring Removes covariate imbalance more effectively (less bias) than other stratification or covariate adjustment Does not require specification of the PS-outcome association Analyzes study population within range of common support	May be subject to the influence of few patients with extreme weights PS model misspecification can result in extreme weights. Subject to residual and unmeasured confounding
Stratification	Allows relatively straightforward analysis and transparent presentation of covariate balance for each stratum Does not require specification of the PS-outcome association Analyzes study population within range of common support	In the presence of non-uniform treatment effects across strata, a single summary estimate is not meaningful. Subject to residual and unmeasured confounding
Covariate Adjustment	Analyzes the entire study population	Requires specification of the PS-outcome association; if misspecified, the estimated effect can be biased. Does not allow transparent assessment of covariate balance Subject to residual and unmeasured confounding Due to these limitations, this method is no longer considered a best practice.

Abbreviation: PS, propensity score.