From Static Risk to Dynamic Trajectories: Toward World-Model-Inspired Clinical Prediction

2026-05-16

AI Summary

This work addresses a critical limitation of existing static clinical prediction models, which conflate disease biology with clinician behavior and struggle to account for treatment feedback, time-varying confounding, and non-random observation patterns. The authors propose the first unified framework for intervention-aware disease trajectory modeling that seamlessly integrates discrete and continuous time. This framework jointly models disease progression, treatment assignment, and observation processes, enabling both factual prediction and counterfactual inference as well as policy evaluation. It synthesizes multi-state/joint models, temporal point processes, deep sequential architectures, and longitudinal causal inference, while incorporating overlapping diagnoses, uncertainty quantification, and target trial emulation. The approach captures individualized treatment-sensitive trajectories and supports pre-deployment stress testing of clinical strategies, thereby generating reliable, decision-grade evidence for safe implementation in closed-loop learning health systems.

Abstract

Clinical decision-making is a feedback system where risk estimates influence treatment, which in turn changes disease trajectories, and both shape clinicians' measurement practices. Static prediction often fails clinically: models trained on observational care logs conflate disease biology with clinician behavior, particularly under treatment confounder feedback and irregular or informative observation. This Review focuses on intervention-aware disease trajectory modeling in clinical AI--methods estimating patient-specific longitudinal disease evolution and assessing trajectory changes under alternative treatments. We organize the field around six linked components: three decision tasks (factual forecasting, counterfactual estimation, policy evaluation) and three data-generating mechanisms (disease evolution, treatment assignment, observation process) that determine identifiability. We present the first unified framework bridging forecasting, counterfactual trajectories, and policy evaluation across discrete/continuous time, explicitly addressing treatment assignment, time-varying confounding, and observation bias. We synthesize key method families (multistate/joint models, temporal point-process, deep sequence architectures, longitudinal causal inference), map them to relevant components, and align evaluation with claim strength via overlap diagnostics, uncertainty quantification, off-policy robustness, and target-trial validation. This synthesis advances benchmark prediction to decision-grade clinical evidence, enabling treatment-sensitive individualized futures, pre-deployment policy stress-testing, and safer closed-loop learning health systems that adapt/abstain when evidence is insufficient.

Problem

• clinical prediction
• disease trajectory
• treatment confounding
• observation bias
• counterfactual estimation

Innovation

• world-model-inspired
• intervention-aware trajectory modeling
• counterfactual estimation
• time-varying confounding
• decision-grade clinical evidence