JMIR Publications

ABSTRACT

Background:

With the development of artificial intelligence technology, large language models (LLMs) have, through their powerful natural language processing and interaction capabilities, provided a new technical path for the intelligent and precise transformation of chronic disease management. However, the effectiveness, adaptability and potential risks of these models in clinical practice still require systematic evaluation.

Objective:

To systematically evaluate the efficacy and existing challenges of Large Language Models (LLMs) in the field of chronic disease management and care, with the aim of synthesizing current evidence to provide a foundation and clear guidance for future clinical practice, technological iteration, policy planning, and further scientific research.

Methods:

Literature on the application of LLMs in chronic disease management and care was retrieved from databases including PubMed, Web of Science, Embase, the Cochrane Library, CINAHL, Wiley Online Library, SpringerLink, ScienceDirect, CNKI, Wanfang, VIP, and CBM, from inception to December 2025. Two researchers independently performed literature search, study selection, quality assessment, data extraction, and statistical analysis.

Results:

A total of 11 studies were included, comprising cross-sectional studies (n=5), quasi-experimental studies (n=3), mixed-methods studies (n=2), and qualitative studies (n=1). Quality assessment indicated that all studies were of moderate to high quality. LLMs demonstrated potential in clinical decision support, patient education, self-assessment, and risk prediction, with notable improvements in accuracy and adaptability particularly when enhanced by model customization and retrieval-augmented generation techniques. However, evidence indicates that their application still faces challenges such as insufficient adaptation to health literacy levels, risk of hallucination, and issues regarding data security and ethical accountability.

Conclusions:

Large Language Models provide technical support for the precision and personalization of chronic disease management and care, yet their clinical translation faces multiple challenges. Future efforts should focus on health literacy adaptation, hallucination mitigation, construction of safety frameworks, and standardization of evaluation to promote their safe and effective clinical application. Clinical Trial: PROSPERO CRD420251208327; https://www.crd.york.ac.uk/PROSPERO/view/CRD420251208327