Algorithms Trained on Normal Chest X-rays Can Predict Health Insurance Types

2025-11-14

AI Summary

This study challenges the conventional assumption that medical imaging data reflect purely biological phenomena by demonstrating that deep learning models can infer patients’ insurance type—a proxy for socioeconomic status—from routine, normal chest X-rays. Method: Using state-of-the-art architectures—including DenseNet121, SwinV2-B, and MedMamba—we trained and evaluated models on the MIMIC-CXR-JPG and CheXpert datasets. We controlled rigorously for age, sex, race, and other clinical covariates to isolate socioeconomic signal. Patch occlusion analysis localized predictive features across lung fields and mediastinum. Results: Models achieved AUCs of 0.67 and 0.68 on the two datasets, respectively, with statistically significant performance even after covariate adjustment. The distributed, anatomy-agnostic signal—termed a “social fingerprint”—reflects systemic sociotechnical disparities embedded in imaging equipment, acquisition protocols, and clinical workflows. This work provides the first empirical evidence that structural social bias is encoded diffusely in routinely acquired negative radiographs, establishing a novel paradigm and methodological foundation for auditing implicit inequities in medical AI systems.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Security and Privacy: Data transparency and provenance

Abstract

Artificial intelligence is revealing what medicine never intended to encode. Deep vision models, trained on chest X-rays, can now detect not only disease but also invisible traces of social inequality. In this study, we show that state-of-the-art architectures (DenseNet121, SwinV2-B, MedMamba) can predict a patient's health insurance type, a strong proxy for socioeconomic status, from normal chest X-rays with significant accuracy (AUC around 0.67 on MIMIC-CXR-JPG, 0.68 on CheXpert). The signal persists even when age, race, and sex are controlled for, and remains detectable when the model is trained exclusively on a single racial group. Patch-based occlusion reveals that the signal is diffuse rather than localized, embedded in the upper and mid-thoracic regions. This suggests that deep networks may be internalizing subtle traces of clinical environments, equipment differences, or care pathways; learning socioeconomic segregation itself. These findings challenge the assumption that medical images are neutral biological data. By uncovering how models perceive and exploit these hidden social signatures, this work reframes fairness in medical AI: the goal is no longer only to balance datasets or adjust thresholds, but to interrogate and disentangle the social fingerprints embedded in clinical data itself.

Problem

• AI models predict health insurance types from normal chest X-rays
• Deep networks detect socioeconomic status embedded in medical images
• Medical AI fairness requires disentangling social fingerprints in data

Innovation

• AI predicts health insurance from chest X-rays
• Deep vision models detect socioeconomic status signals
• Models learn diffuse social fingerprints in medical images