ChronoMedicalWorld: A Medical World Model for Learning Patient Trajectories from Longitudinal Care Data
Researchers have developed the ChronoMedicalWorld Model (CMWM), a novel framework designed to predict patient health trajectories over long periods using longitudinal electronic health record data. This action-conditioned latent world model incorporates both structured interventions and free-text communication to forecast physiological changes. In a study focusing on chronic kidney disease, CMWM demonstrated improved accuracy in predicting estimated glomerular filtration rate compared to a GPT-5.5 baseline, with gains attributed partly to the analysis of patient-health coach dialogue. AI
IMPACT This model could enhance long-term patient care by providing more accurate predictions of disease progression and intervention effectiveness.