Proximal Learning for Individualized Treatment Regimes Under Unmeasured Confounding

TL;DR

This paper introduces proximal learning methods for estimating optimal individualized treatment regimes in observational studies with unmeasured confounding, balancing assumptions and decision-making value.

Contribution

It develops new proximal causal inference techniques for ITRs under unmeasured confounding, with theoretical guarantees and practical algorithms.

Findings

Proposed methods outperform existing approaches in simulations.

Theoretical properties established for the new proximal learning approaches.

Successful application demonstrated on real observational data.

Abstract

Data-driven individualized decision making has recently received increasing research interests. Most existing methods rely on the assumption of no unmeasured confounding, which unfortunately cannot be ensured in practice especially in observational studies. Motivated by the recent proposed proximal causal inference, we develop several proximal learning approaches to estimating optimal individualized treatment regimes (ITRs) in the presence of unmeasured confounding. In particular, we establish several identification results for different classes of ITRs, exhibiting the trade-off between the risk of making untestable assumptions and the value function improvement in decision making. Based on these results, we propose several classification-based approaches to finding a variety of restricted in-class optimal ITRs and develop their theoretical properties. The appealing numerical performance of our proposed methods is demonstrated via an extensive simulation study and one real data application.