Improving RCT-Based CATE Estimation Under Covariate Mismatch via Double Calibration

2026-03-17

AI Summary

This study addresses the challenge of estimating conditional average treatment effects (CATE) when covariates in randomized controlled trials (RCTs) and observational studies do not fully overlap. The authors propose MR-OSCAR, a novel method that integrates covariate imputation with double calibration: in the first stage, missing covariates in the RCT are imputed using observational data; in the second stage, predictions from the observational model are calibrated to align with the RCT covariate distribution, thereby preserving causal structure while improving CATE estimation accuracy. Theoretical analysis provides finite-sample error decomposition guarantees. Empirical results demonstrate that MR-OSCAR substantially outperforms benchmarks relying solely on shared covariates, particularly when missing covariates are highly predictable and the RCT sample size is moderate. The method has been successfully applied to the Greenlight Plus pediatric obesity trial and Vanderbilt electronic health records data.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Causality
• Intelligent Robots: State Estimation

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

We develop estimators that improve precision of heterogeneous treatment effect estimates that allow borrowing information from observational studies when the available covariates in each data source do not perfectly match. Standard data-borrowing methods often assume perfectly matched covariates. We propose MR-OSCAR, an RCT-calibrated, two-stage estimation approach that first predicts the trial-missing variables using the observational data via imputation and then calibrates observational outcome predictions to the randomized trial, preserving the causal contrast, unlike the results for generalization, where imputation does not improve performance. Our theory gives finite-sample guarantees with a transparent error decomposition including an imputation error that shrinks as the observational mapping becomes more predictable. Simulations show that imputation almost always outperforms naively using only the shared covariates and clarifies when borrowing helps (strong predictability of the missing block, moderate trial size) and when it does not (poor predictability or dominant trial-only moderators). We motivate the approach with the Greenlight Plus trial on early childhood obesity and outline a forthcoming EHR analysis at Vanderbilt, highlighting the use of our method in common scenarios where data do not perfectly align.

Problem

• CATE estimation
• covariate mismatch
• data borrowing
• randomized controlled trial
• observational study

Innovation

• CATE estimation
• covariate mismatch
• double calibration
• data borrowing
• imputation