Dementia-R1: Reinforced Pretraining and Reasoning from Unstructured Clinical Notes for Real-World Dementia Prognosis

2026-01-06

arXiv.org

AI Summary

This work addresses the challenge that existing large language models struggle to capture the non-monotonic symptom progression across multiple clinical visits in long-term dementia prognosis, primarily due to sparse binary reward signals and the absence of explicit symptom trajectory annotations. To overcome these limitations, the authors propose Dementia-R1, a novel framework that integrates cold-start reinforcement learning with pretraining grounded in verifiable clinical metrics. The approach first extracts longitudinal clinical signals from unstructured electronic health records for pretraining and then refines predictions of final clinical outcomes through reinforcement learning. This strategy effectively mitigates the issues of sparse rewards and complex trajectory modeling, achieving an F1 score of 77.03% on real-world data. Notably, the 7B-parameter variant matches GPT-4o’s performance on the ADNI benchmark while significantly enhancing the model’s ability to capture fluctuating cognitive trajectories.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

While Large Language Models (LLMs) have shown strong performance on clinical text understanding, they struggle with longitudinal prediction tasks such as dementia prognosis, which require reasoning over complex, non-monotonic symptom trajectories across multiple visits. Standard supervised training lacks explicit annotations for symptom evolution, while direct Reinforcement Learning (RL) is hindered by sparse binary rewards. To address this challenge, we introduce Dementia-R1, an RL-based framework for longitudinal dementia prognosis from unstructured clinical notes. Our approach adopts a Cold-Start RL strategy that pre-trains the model to predict verifiable clinical indices extracted from patient histories, enhancing the capability to reason about disease progression before determining the final clinical status. Extensive experiments demonstrate that Dementia-R1 achieves an F1 score of 77.03% on real-world unstructured clinical datasets. Notably, on the ADNI benchmark, our 7B model rivals GPT-4o, effectively capturing fluctuating cognitive trajectories. Code is available at https://anonymous.4open.science/r/dementiar1-CDB5

Problem

• dementia prognosis
• longitudinal prediction
• clinical notes
• symptom trajectories
• non-monotonic reasoning

Innovation

• Reinforcement Learning
• Longitudinal Prognosis
• Clinical Reasoning
• Cold-Start RL
• Unstructured Clinical Notes