Application of Large Language Models in Chronic Disease Care: Mixed Methods Systematic Review and Thematic Synthesis

Area of Science:

• Artificial Intelligence in Healthcare
• Digital Health
• Nursing Informatics

Background:

• Chronic diseases represent a significant global health burden, necessitating continuous and personalized long-term management.
• Traditional healthcare models struggle to provide the sustained support required for chronic disease care.
• Large language models (LLMs) offer potential for scalable, interactive support in chronic disease management.

Purpose of the Study:

• To systematically review evidence on LLM technical performance, applications, and challenges in chronic disease care.
• To critically evaluate LLMs using a theory-driven, 3D framework based on Orem's Self-Care Theory.

Main Methods:

• A mixed-methods systematic review following PRISMA 2020 and SWiM guidelines.
• Comprehensive literature search across multiple databases from inception to April 2026.
• Thematic synthesis of 20 moderate-to-high quality studies using NVivo 14, guided by Orem's 3D framework.

Main Results:

• Three core themes emerged: foundational safety/privacy, self-care enablement, and design/integration challenges.
• LLMs showed strength in patient education and decision support, enhanced by retrieval-augmented generation (RAG) and fine-tuning.
• Barriers included content readability, hallucination risks, and ethical ambiguities, limiting real-world readiness.

Conclusions:

• This review provides the first theory-driven, nursing-informed evaluation of LLMs in chronic care.
• Evidence primarily supports LLM capabilities in self-care enablement, with deficiencies in safety and integration.
• Future nursing research should focus on safety, health-literacy-adaptive design, and implementation science for equitable patient-centered care.