Introduction

Many existing clinical AI models remain static or retrospective: they predict diagnosis, deterioration, readmission, mortality, or treatment response under observed care, but do not explicitly model how a patient may evolve under alternative interventions. We define medical world models as AI systems that learn patient-state dynamics from multimodal clinical data, simulate future disease trajectories under alternative interventions, and support safe clinical decision-making under uncertainty.

This workshop brings together researchers and practitioners from artificial intelligence, machine learning, medicine, biomedical informatics, medical imaging, causal inference, clinical decision support, and digital health, through invited talks, paper spotlights, posters, and a panel discussion, to define an emerging interdisciplinary research agenda for dynamic, simulation-based, decision-grade healthcare AI.