ChronoMedicalWorld: A Medical World Model for Learning Patient Trajectories from Longitudinal Care Data

2026-05-20

AI Summary

This study addresses the challenge of accurately modeling the long-term physiological evolution of chronic disease patients under sustained clinical interventions, a task where existing electronic health record models often fall short. To this end, the authors propose an action-conditioned latent world model that jointly embeds clinical states and a broad-action encoder to integrate structured interventions with physician–patient dialogue transcripts. The recursive latent transition module is trained under a closed-loop replay-prefix protocol and enhanced with six distinct loss objectives, SIGReg regularization, and physiologically informed priors—including slope constraints, continuity enforcement, and penalties for implausible large jumps. Evaluated on a cohort of 2,232 chronic kidney disease patients, the model achieves an annual eGFR prediction MAE of 7.384 and RMSE of 10.256, outperforming a fine-tuned GPT-5.5 baseline by 7.28% and 7.35%, respectively, with performance gains largely attributed to effective modeling of clinical dialogue.

Abstract

Long-horizon clinical simulation -- predicting how a patient's physiology evolves over years under specified interventions -- is central to chronic-disease care, yet existing electronic health record (EHR) models are predominantly discriminative, and general-purpose large language models drift under repeated interventions. We propose the \textbf{ChronoMedicalWorld Model (CMWM)}, an action-conditioned latent world-model framework for learning patient trajectories from longitudinal care data. CMWM couples a joint-embedding state encoder with a wide action encoder that admits both structured intervention indicators and free-text communication embeddings, and trains a recurrent latent transition module under a six-term objective: next-observation supervision, next-latent prediction, SIGReg latent regularisation, and three physiology-aware shape priors (slope, continuity, large-jump penalty). A closed-loop rollout-prefix protocol matches training to deployment, so the model is optimised against the same multi-step error it exhibits at inference. As a concrete case study, we instantiate CMWM for annual estimated glomerular filtration rate (eGFR) trajectory forecasting in chronic kidney disease (CKD). On a 2{,}232-patient nephrology cohort, the CKD instantiation achieves a dynamic-50\% history rollout test mean absolute error (MAE) of 7.384 and root-mean-square error (RMSE) of 10.256, against 7.964 and 11.069 for a tuned GPT-5.5 structured-prompting baseline ($-7.28\%$ MAE, $-7.35\%$ RMSE), with the gain dominated by the dialogue portion of patient--health-coach communication. The framework is not CKD-specific: its architecture, loss design, and training protocol apply to any chronic condition that can be cast as periodic clinical state interleaved with structured and conversational interventions.

Problem

• long-horizon clinical simulation
• patient trajectories
• chronic disease care
• electronic health records
• physiological evolution

Innovation

• world model
• longitudinal patient trajectory
• action-conditioned modeling
• physiology-aware priors
• closed-loop rollout training