Using Large Language Models to Enhance Exercise Recommendations and Physical Activity in Clinical and Healthy...

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Exercise Science

Background:

• Regular exercise recommendations (ERs) and physical activity (PA) are vital for chronic disease management.
• Developing personalized exercise programs requires significant time and expertise.
• Large language models (LLMs) offer a potential solution for creating personalized ERs, but their application is nascent.

Purpose of the Study:

• To systematically review and classify LLM applications in ERs and PA.
• To identify research gaps and future directions for LLM integration in personalized health interventions.

Main Methods:

• A scoping review methodology was employed.
• Searches were conducted across major scientific databases (Web of Science, PubMed, IEEE, arXiv) up to March 2024.
• Thematic analysis synthesized findings from 11 included studies, with two independent reviewers screening papers.

Main Results:

• LLMs, particularly ChatGPT-based models (55% of studies), demonstrated potential in generating tailored ERs.
• Studies showed LLMs can save clinical time, enhance safety with patient data, boost engagement, and support behavior change.
• LLM-generated PA guidance increased accessibility, especially for remote or underserved populations.

Conclusions:

• LLMs show promise in ERs and PA but should supplement, not replace, human expertise.
• Expert validation is essential to ensure safety and mitigate risks associated with LLM use.
• Future research should focus on pilot testing, clinician training, and large-scale trials for ethical and effective integration.