Estimation of Optimal Individualized Treatment Rules Using a Covariate-Specific Treatment Effect Curve With High-Dimensional Covariates

With a large number of baseline covariates, we propose new semiparametric modeling strategy for heterogeneous treatment effect estimation and individualized selection, which are two major goals in personalized medicine. We achieve the first goal through estimating covariate-specific (CSTE) curve modeled as an unknown function weighted linear combination all covariates. The weight or coefficient each covariate is estimated by fitting sparse logistic single-index model. CSTE spline-backfitted kernel procedure, enables us to further construct simultaneous confidence band (SCB) under desired level. Based on SCB, find subgroups patients that benefit from treatment, so can make selection. innovations proposed method 3-fold. First, quantify variability associated with optimal rule high-dimensional Second, very flexible depict both local global associations between covariates presence thus it enjoys flexibility while achieving dimensionality reduction. Third, SCB achieves nominal level asymptotically, provides uniform inferential tool making decisions. Supplementary materials this article available online.