Propensity score weighting for causal subgroup analysis

Area of Science:

• Epidemiology
• Biostatistics
• Health Services Research

Background:

• Estimating treatment effects in specific patient subpopulations is crucial for comparative effectiveness research.
• Existing causal inference methods for subgroup analysis (SGA) in observational studies are underdeveloped.
• Accurate SGA is vital for personalized medicine and evidence-based healthcare decisions.

Purpose of the Study:

• To develop and validate novel analytical methods and visualization tools for causal SGA in observational research.
• To improve the estimation of treatment effects within prespecified patient subgroups.
• To address limitations in existing methods for bias and variance in SGA.

Main Methods:

• Introduction of the subgroup weighted average treatment effect estimand and propensity score weighting estimator.
• Application of overlap weighting (OW) for exact covariate balance within subgroups.
• Combination of OW and LASSO for bias-variance tradeoff optimization in SGA.
• Development of the Connect-S plot for visualizing subgroup covariate balance.

Main Results:

• Balancing covariates within subgroups effectively bounds the bias of subgroup causal effect estimators.
• The proposed OW and OW-LASSO methods demonstrate improved performance in simulations.
• The Connect-S plot provides a useful tool for assessing covariate balance in SGA.
• Simulation studies confirm the advantages of the proposed methods over existing approaches.

Conclusions:

• The developed suite of methods and tools significantly advances causal SGA in observational studies.
• The proposed techniques enhance the reliability and interpretability of subgroup treatment effect estimates.
• These advancements have direct implications for improving patient care and comparative effectiveness research outcomes.