Machine learning-based subgroup analysis via covariate balancing using genetic matching in observational studies

Subgroup analysis has received increasing attention from practitioners in personalized medicine.In observational studies, one of the main difficulties in subgroup analysis is the imbalance of covariates between treatment and control groups.In this paper, we develop a new method for identifying subgroups in observational studies by first using genetic matching to construct comparable groups and then performing imputation and variable selection to identify covariates that drive treatment heterogeneity.The performance of the proposed treatment effect estimation will be assessed in simulation studies.Right-heart catheterization (RHC) data analysis shows the validity and usefulness of our methodology.