Abstract
The prevention of chronic disease is a long-term combat with continual fine-tuning to adapt to the course of disease. Without comprehensive insights, prescriptions may prioritize short-term gains but deviate from trajectories toward long-term survival. Here we introduce Duramax, an evidence-based framework empowered by reinforcement learning to optimize long-term preventive strategies. Duramax learned from real-world treatment trajectories involving over 200 lipid-modifying drugs across more than 3.6 million months, becoming specialized in cardiovascular disease (CVD) prevention. Duramax demonstrated a superior performance in model validation using an independent cohort encompassing over 29.7 million treatment months. Specifically, Duramax achieved policy value of 93, outperforming clinicians with value of 68. When clinicians’ decisions aligned with Duramax’s suggestions, CVD risk reduced by 6%. Moreover, post hoc analysis confirmed that Duramax’s decisions were transparent and reasonable. Our research showcases how tailored computational analysis on well-curated health records can achieve high nuance in personalized disease prevention.
| Original language | English |
|---|---|
| Article number | 553 |
| Number of pages | 12 |
| Journal | npj Digital Medicine |
| Volume | 8 |
| Issue number | 1 |
| Early online date | 27 Aug 2025 |
| DOIs | |
| Publication status | Published - 27 Aug 2025 |
Bibliographical note
Copyright © The Author(s) 2025. This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted materialderived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit https://creativecommons.org/licenses/by-
nc-nd/4.0/
Data Access Statement
Sensitive patient data is not available. Restricted access for validation is available upon request. Please write to R.L. ([email protected]) and C.S.L.C. ([email protected]) for details.The code used in the article is available upon request to R.L. ([email protected]).