Theory and practice of propensity score analysis.

Area of Science:

• Epidemiology
• Biostatistics
• Observational Research Methods

Background:

• Propensity score (PS) analysis is a common method for causal inference in observational studies.
• Addressing confounding is crucial for valid causal claims from non-randomized data.

Purpose of the Study:

• To outline the fundamental assumptions for causal inference.
• To present a structured five-step methodology for conducting propensity score analysis.
• To compare propensity score methods with traditional regression techniques.

Main Methods:

• The study details three core assumptions: conditional exchangeability, positivity, and consistency.
• A five-step process is described: PS model construction, overlap assessment, appropriate weighting/matching, covariate balance checking, and effect estimation.
• Various weighting methods (IPTW, SMRW, overlap weights) and matching are discussed.

Main Results:

• The structured approach ensures rigorous application of propensity score methods.
• Effective covariate balance and overlap are critical for reliable causal estimates.
• Propensity score analysis offers advantages over conventional multivariable regression for causal inference.

Conclusions:

• Adherence to assumptions and the outlined steps enhances the validity of causal inferences from observational data.
• Propensity score methods provide a powerful framework for estimating treatment effects in the presence of confounding.
• This methodology facilitates more reliable causal conclusions compared to standard regression models.