JMIR Publications

ABSTRACT

Background:

Accounting for nearly three-quarters of deaths worldwide, chronic diseases are a major global health burden. Large language models (LLMs) are advanced artificial intelligence systems, possessing transformative potential to optimise chronic disease management, yet robust evidence is lacking.

Objective:

To synthesise evidence on the feasibility, opportunities, and challenges of LLMs across the disease management spectrum–from prevention to screening, diagnosis, treatment, and long-term care.

Methods:

Following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines, eleven databases (Cochrane Central Register of Controlled Trials, CINAHL, Embase, IEEE Xplore, Medline via Ovid, ProQuest Health & Medicine Collection, ScienceDirect, Scopus, Web of Science Core Collection, China National Knowledge Internet, and SinoMed) were searched on 17 April 2024. Intervention and simulation studies were included if they examined LLMs in managing chronic diseases. Narrative analysis with descriptive figures were utilised to synthesise study findings. Random-effects meta-analyses were conducted to assess pooled effect estimates for LLM feasibility in chronic disease management.

Results:

Twenty studies were eligible examining general-purpose (n = 17) and fine-tuned LLMs (n = 3) in managing chronic diseases, including cancer, cardiovascular diseases, and metabolic disorders. LLMs demonstrated feasibility across the chronic disease management spectrum by generating relevant, comprehensible, and accurate health recommendations (71%; 95% confidence interval [CI] = 0.59, 0.83; I2 = 88.32%) with fine-tuned LLMs having higher accurate rates compared to general-purpose LLMs (odds ratio = 2.89; 95% CI = 1.83, 4.58; I2 = 54.45%). LLMs facilitated equitable information access, increased patient awareness of ailments, preventive measures, and treatment options, and promoted self-management behaviours in lifestyle modification and symptom coping. Additionally, LLMs facilitated compassionate emotional support, social connections, and healthcare resource to improve health outcomes of chronic diseases. However, LLMs faced challenges in addressing privacy, language, and cultural issues, undertaking advanced tasks, including diagnostic, medication, and comorbidities management, and generating personalised regimens with real-time adjustments and multiple modalities.

Conclusions:

LLMs demonstrated potential to transform chronic disease management at individual, social, and healthcare levels, yet their direct application in clinical settings is still in its infancy. A multifaced approach–incorporating robust data security, domain-specific model fine-tuning, multimodal data integration, and wearables–is crucial to evolve LLMs into invaluable adjuncts for healthcare professionals to transform chronic disease management. Clinical Trial: PROSPERO (CRD42024545412).