Distributionally Robust Transfer Learning with Structurally Missing Covariates, with Application to Cross-National Cardiac Arrest Prediction

2026-05-22

AI Summary

Clinical prediction models often suffer performance degradation when deployed across healthcare systems due to structural missingness of covariates in the target domain and the absence of labeled data. To address this challenge, this work proposes DRUM, a robust unsupervised transfer learning framework that avoids imputing missing covariates or relying on unverifiable distributional assumptions. Instead, DRUM optimizes worst-case predictive performance over the conditional distribution of missing covariates. The method integrates distributionally robust optimization with neural network–based generative modeling, conditional distribution estimation, and a bias-correction mechanism, further incorporating a tunable robustness parameter to mitigate sensitivity to perturbation estimation errors. In both simulated experiments and a real-world cross-national cardiac arrest prediction task, DRUM substantially improves predictive accuracy, calibration, and clinical classification performance under both average and worst-case scenarios.

Abstract

Deploying clinical prediction models across healthcare systems often fails when key training covariates are unavailable at deployment and labeled outcomes are limited in the target domain. For example, high-performing models for out-of-hospital cardiac arrest (OHCA) rely on detailed prehospital measurements routinely collected in high-resource settings but unavailable in many international registries. Existing methods either discard missing covariates, sacrificing predictive information, or rely on untestable assumptions about their target distribution. We propose DRUM (\underline{D}istributionally \underline{R}obust \underline{U}nsupervised transfer learning with structurally \underline{M}issing covariates), a framework that transfers prediction models to target populations where certain covariates are structurally absent and outcome labels are unavailable. DRUM partitions covariates into shared components ($X$), observed across all settings, and missing components ($A$), observed only in the source. Rather than imputing missing covariates, DRUM optimizes worst-case predictive performance over the unknown target distribution of $A \mid X$ using a neural network generator, with a robustness parameter controlling allowable deviation from the source conditional. We further develop a bias correction procedure that reduces sensitivity to nuisance estimation error. Simulations show substantial improvements in both mean and worst-case prediction error under distribution shift. Applied to cross-national OHCA prediction, transferring models from a US registry to multiple Asian registries where prehospital variables are unrecorded, DRUM yields better-calibrated predictions and improved clinical classification performance across sites.

Problem

• transfer learning
• structurally missing covariates
• distributional robustness
• cross-national prediction
• clinical risk prediction

Innovation

• distributionally robust optimization
• transfer learning
• structurally missing covariates
• unsupervised domain adaptation
• clinical prediction models