🤖 AI Summary
This paper addresses the estimation of optimal individualized treatment rules (ITRs) for multicategory treatments in observational studies—a setting where existing methods are restricted to binary treatments and require fully observed outcomes. We propose a novel framework accommodating both fully observed and right-censored outcomes (e.g., survival data). Our approach integrates matching-based learning with ensemble machine learning (random forests, gradient boosting) and unifies causal inference with survival analysis via inverse probability of censoring weighting–Cox regression (IPCW-Cox). To our knowledge, this is the first method for multicategory ITRs with theoretical guarantees of consistency and convergence rate. Simulation studies demonstrate substantial improvements over state-of-the-art alternatives. Applied to real-world hepatocellular carcinoma data, our method identifies clinically meaningful, stratified treatment policies, enhancing both the applicability and interpretability of precision medicine in complex, real-world settings.
📝 Abstract
One primary goal of precision medicine is to estimate the individualized treatment rules that optimize patients' health outcomes based on individual characteristics. Health studies with multiple treatments are commonly seen in practice. However, most existing individualized treatment rule estimation methods were developed for the studies with binary treatments. Many require that the outcomes are fully observed. In this article, we propose a matching-based machine learning method to estimate the optimal individualized treatment rules in observational studies with multiple treatments when the outcomes are fully observed or right-censored. We establish theoretical property for the proposed method. It is compared with the existing competitive methods in simulation studies and a hepatocellular carcinoma study.
