Do ChatGPT and Gemini's Recommendations Align With Established Guidelines for Hand and Upper Extremity Surgery?

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Decision Support Systems
• Healthcare Informatics

Background:

• Large language models (LLMs) are increasingly used in healthcare for administrative tasks and patient communication.
• Concerns exist regarding the clinical accuracy and reliability of LLM recommendations in medical practice.
• This study investigates the adherence of LLM outputs to established clinical practice guidelines (CPGs).

Purpose of the Study:

• To evaluate the concordance of recommendations from ChatGPT and Gemini with American Academy of Orthopedic Surgeons (AAOS) clinical practice guidelines (CPGs).
• To compare the performance of ChatGPT and Gemini in terms of accuracy and citation transparency across different orthopedic conditions.

Main Methods:

• ChatGPT (version 4o) and Gemini (version 1.5 Flash) were prompted with text aligned to AAOS CPGs for carpal tunnel syndrome, distal radius fractures, and glenohumeral joint osteoarthritis.
• Blinded reviewers assessed the LLM outputs for concordance with CPGs.
• Concordance rates were compared between models, topics, and guideline strengths, with citation transparency also evaluated.

Main Results:

• An overall concordance rate of 62.1% was observed across 174 recommendations.
• No statistically significant difference in concordance was found between ChatGPT (66.7%) and Gemini (57.5%) (P = .131).
• Both models exhibited low citation transparency, with Gemini providing sources more frequently (39.1%) than ChatGPT (3.5%) (P < .0001).

Conclusions:

• Current LLMs demonstrate modest concordance with AAOS CPGs, indicating significant limitations for clinical integration.
• Variability in performance across topics and guideline strengths, coupled with poor citation transparency, necessitates further refinement.
• Thorough evaluation and improvement are crucial before adopting LLMs in hand surgery and other clinical specialties.