Robust Statistical Method For Finding Optimal Treatment Regimes

• Activity Number: 232
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Baqun Zhang*+ and Marie Davidian and Anastasios Tsiatis
• Companies: North Carolina State University and North Carolina State University and North Carolina State University
• Keywords: Optimal treatment regime ; Inverse probability weighting ; Potential outcomes ; Classification

Abstract

A treatment regime is a rule that assigns a treatment, among a set of possible treatments, to a patient as a function of his/her observed covariates. The goal is to find the optimal treatment regime defined as that, if followed by a population of patients, would lead to the best outcome on average. For a single treatment decision, the optimal treatment regime can be found by developing a regression model for the expected outcome as a function of treatment and baseline covariates, where, for a given set of covariates, the optimal treatment is the one which yields the largest expected outcome. This, however can lead to biased results if the regression model is misspecified. Realizing that the parameters in a regression model induce different treatment regimes, we instead consider estimating the mean outcome for such treatment regimes directly using doubly-robust augmented inverse propensity score estimators of the mean outcome for the parameter-induced treatment regimes, which we then maximize across the parameter values to obtain our optimal treatment regime estimator. We also show how this problem can be viewed as a classification problem. Simulations and application are presented