Maximin Learning of Individualized Treatment Effect on Multi-Domain Outcomes

2026-03-27

AI Summary

This work addresses the limited generalizability of existing individualized treatment effect estimation methods, which typically rely on a single or specific observed outcome and struggle to extend to unobserved yet clinically relevant multidimensional outcomes. To overcome this challenge, the authors propose the DRIFT framework, which uniquely integrates minimax robust optimization with generalized factor analysis. By extracting latent factors and constructing an anchored uncertainty set, DRIFT learns treatment effects that are robust to unseen clinical domains. The method enjoys parametrization invariance, admits a closed-form solution, and comes with theoretical guarantees. Evaluated on the EMBARC randomized controlled trial for depression, DRIFT significantly outperforms existing approaches on external multidomain outcomes—such as side effects and self-reported symptoms—that were unobserved during training, demonstrating superior generalization capability.

Abstract

Precision mental health requires treatment decisions that account for heterogeneous symptoms reflecting multiple clinical domains. However, existing methods for estimating individualized treatment effects (ITE) rely on a single summary outcome or a specific set of observed symptoms or measures, which are sensitive to symptom selection and limit generalizability to unmeasured yet clinically relevant domains. We propose DRIFT, a new maximin framework for estimating robust ITEs from high-dimensional item-level data by leveraging latent factor representations and adversarial learning. DRIFT learns latent constructs via generalized factor analysis, then constructs an anchored on-target uncertainty set that extrapolates beyond the observed measures to approximate the broader hyper-population of potential outcomes. By optimizing worst-case performance over this uncertainty set, DRIFT yields ITEs that are robust to underrepresented or unmeasured domains. We further show that DRIFT is invariant to admissible reparameterizations of the latent factors and admits a closed-form maximin solution, with theoretical guarantees for identification and convergence. In analyses of a randomized controlled trial for major depressive disorder (EMBARC), DRIFT demonstrates superior performance and improved generalizability to external multi-domain outcomes, including side effects and self-reported symptoms not used during training.

Problem

• individualized treatment effect
• multi-domain outcomes
• robustness
• generalizability
• latent factors

Innovation

• maximin learning
• individualized treatment effect
• latent factor model
• adversarial learning
• multi-domain outcomes