Skip to content

Abstract Search

Primary Submission Category: Weighting

Transfer Learning for Individualized Treatment Rules

Authors: Andong Wang, Johnny Rajala, Kelly Wentzlof, Miontranese Green,

Presenting Author: Andong Wang*

Modern precision medicine aims to utilize real-world data to provide the best treatment for an individual patient. An individualized treatment rule (ITR) maps individual characteristics to a recommended treatment that maximizes the expected outcome of each patient. A problem facing modern medicine is that studies on the effect of treatment are conducted for a source population that may be different from the population of interest. Our research goal is to investigate a transfer learning algorithm to obtain targeted, optimal, and interpretable ITRs. We develop a calibrated augmented inverse probability weighting (CAIPW) estimator by maximizing the value function for the target population to estimate an optimal ITR. Additionally, we investigate transfer learning methods based on two large medical databases, eICU Collaborative Research Database (eICU-CRD) and Medical Information Mart for Intensive Care III (MIMIC-III), identifying the important covariates, treatment options, and outcomes of interest to estimate the optimal linear and tree-based ITRs for patients with sepsis. This project introduces new techniques for data merging to provide data-driven optimal ITRs, catering to each patient’s individual medical needs. These techniques extend beyond medicine, applying to a wide range of areas such as marketing, technology, social services, and education.