Efficient and Robust Semi-supervised Estimation of ATE with Partially Annotated Treatment and Response

TL;DR

This paper introduces a semi-supervised method for estimating average treatment effects using electronic health records, effectively handling partially annotated data with improved efficiency and robustness.

Contribution

It develops a semi-supervised estimator for ATE that is semi-parametric efficient and robust, incorporating high-dimensional models and machine learning techniques.

Findings

SMMAL estimator outperforms supervised benchmarks in simulations.

The method is semi-parametric efficient with B-spline regression.

Adaptive sparsity estimation enhances high-dimensional model performance.

Abstract

A notable challenge of leveraging Electronic Health Records (EHR) for treatment effect assessment is the lack of precise information on important clinical variables, including the treatment received and the response. Both treatment information and response often cannot be accurately captured by readily available EHR features and require labor intensive manual chart review to precisely annotate, which limits the number of available gold standard labels on these key variables. We consider average treatment effect (ATE) estimation under such a semi-supervised setting with a large number of unlabeled samples containing both confounders and imperfect EHR features for treatment and response. We derive the efficient influence function for ATE and use it to construct a semi-supervised multiple machine learning (SMMAL) estimator. We showcase that our SMMAL estimator is semi-parametric efficient with B-spline regression under low-dimensional smooth models. We develop the adaptive sparsity/model doubly robust estimation under high-dimensional logistic propensity score and outcome regression models. Results from simulation studies support the validity of our SMMAL method and its superiority over supervised benchmarks.