LifeLong digital twin models whole-person health over a lifetime
LifeLong Digital Twin: A Unified Modeling Paradigm and Agent Harness for Event-Driven Lifelong Health State Trajectories
Artificial Intelligence
Summary
Human health changes every day based on many factors like biology, behavior, and environment. The authors created a system called LifeLong Digital Twin to track health events and states continuously throughout life. This system uses advanced language models combined with other data types to better predict diseases and their risks over time. Their tests show improved accuracy in forecasting diseases and ranking health risks for individuals.
What this means in practice
- •For clinical data teams: Create integrated patient health records that continuously update health status and predict disease risks over years for better personalized care.
- •For healthcare ai developers: Develop AI systems combining text-based language models with diverse health data to improve long-term disease forecasting and risk assessments.
- •For insurance analytics teams: Enhance disease risk ranking models for setting insurance premiums based on detailed lifelong health event trajectories.$Commercial implications: Enables personalized insurance products by accurately predicting individual disease risks from integrated lifelong health data.
Authors
Jin Jiang, Sean Yates, Jasper Chong, Raymond Brooks, Alex Lawson, Yuqin Qin, Liangcai Gao
Abstract
Human health is a continuous, dynamic trajectory shaped by the cumulative interplay of biological processes, clinical events, behaviors and environmental exposures across the life course. Unifying the full breadth of lifelong health information, including longitudinal records, genetic variation, molecular profiles and environmental histories, is essential for whole-person modeling and remains a major challenge. We introduce LifeLong Digital Twin, a unified, event-driven modeling paradigm that organizes Life Events into daily Health States and accumulates them into Lifelong Health Context. An accompanying Agent Harness incorporates multimodal evidence beyond the language model's textual context. We evaluate four language models across 25 disease endpoints on three tasks: Disease Trajectory Forecasting, Disease Risk Ranking and Multi-horizon Disease Prediction. The approach yields marked gains over the reference condition: model-averaged F1 increases by 22.0% for disease identification in trajectory forecasting and 18.3% for five-year disease outcomes; thyroid-disease F1 reaches 0.669. The framework provides a foundation for whole-person digital twins and research on personalized lifelong disease prevention.