aiqbench analysis
This research proposes a novel world model for electronic health records that decouples patient states from clinical actions, achieving competitive forecasting performance with a significantly smaller parameter footprint than current foundation models. By explicitly modeling the evolution of latent trajectories conditioned on actions, the framework offers a more interpretable and efficient approach to predicting patient outcomes in longitudinal healthcare data. This work advances machine learning applications in specialized domains by demonstrating that explicit dynamic modeling can rival larger black-box architectures while retaining critical predictive signals.
Forecasting Patient Trajectories from the Electronic Health Records with Action-Conditioned World Models
arXiv:2610.23862v1 Announce Type: new Abstract: We introduce a predictive world model for electronic health records (EHR) that represents patient histories as action conditioned latent trajectories. Rather than modeling clinical events solely through autoregressive token prediction, our framework explicitly separates patient states from observed clinical actions and learns their evolution through a state and action conditioned transition operator. At its core is a state dependent linearization predictor that propagates local changes in latent patient state, enabling time resolved prediction, multi step rollout, and action conditioned simulation. Empirically, we show that generated latent trajectories preserve downstream predictive signal, temporal structure, and subgroup level statistical relationships present in encoder derived representations. We further decode free running rollouts into the original clinical feature space and evaluate future clinical measurements against observed values using task matched forecasting baselines, demonstrating that the learned dynamics retain predictive information beyond latent space agreement. Across binary classification and time to event tasks, EHR WM achieves competitive, though task dependent, performance relative to substantially larger EHR foundation models while using a 30x smaller parameter footprint. Finally, perturbing observed action sequences produces corresponding changes in generated trajectories, demonstrating the action sensitivity of the learned dynamics. Together, these results support world modeling as a complementary framework for learning explicit, time indexed representations of patient evolution from longitudinal EHR.
Abstract:We introduce a predictive world model for electronic health records (EHR) that represents patient histories as action conditioned latent trajectories. Rather than modeling clinical events solely through autoregressive token prediction, our framework explicitly separates patient states from observed clinical actions and learns their evolution through a state and action conditioned transition operator. At its core is a state dependent linearization predictor that propagates local changes in latent patient state, enabling time resolved prediction, multi step rollout, and action conditioned simulation. Empirically, we show that generated latent trajectories preserve downstream predictive signal, temporal structure, and subgroup level statistical relationships present in encoder derived representations. We further decode free running rollouts into the original clinical feature space and evaluate future clinical measurements against observed values using task matched forecasting baselines, demonstrating that the learned dynamics retain predictive information beyond latent space agreement. Across binary classification and time to event tasks, EHR WM achieves competitive, though task dependent, performance relative to substantially larger EHR foundation models while using a 30x smaller parameter footprint. Finally, perturbing observed action sequences produces corresponding changes in generated trajectories, demonstrating the action sensitivity of the learned dynamics. Together, these results support world modeling as a complementary framework for learning explicit, time indexed representations of patient evolution from longitudinal EHR.
Machine Learning (cs.LG)
arXiv:2610.23862 [cs.LG]
(or arXiv:2610.23862v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.23862 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 303 words
Related coverage
Graph-Enhanced Disease Trajectory Discovery and Temporal Endpoint Forecasting from Longitudinal Electronic Health Records
arXiv · Oct 15, 2026
Learning Continuous Patient Trajectories from Electronic Health Records
arXiv · Sep 30, 2026
EHR2Path: Comprehensive Pathway-Level Modeling of Longitudinal Patient Trajectories from Multimodal Electronic Health Records
arXiv · Aug 4, 2026
A Two-Level Clinical Novelty Prioritization Framework for Disease Trajectories Discovered from Longitudinal Electronic Health Records
arXiv · Oct 15, 2026