Use and Application of Large Language Models for Patient Questions Following Total Knee Arthroplasty

Area of Science:

• Orthopaedic Surgery
• Artificial Intelligence in Healthcare
• Patient Education

Background:

• Consumer-driven healthcare models increase patient access to information, necessitating accurate health data.
• Nurses significantly impact patient satisfaction post-total knee arthroplasty (TKA), but incur costs.
• The rapid adoption of AI, like ChatGPT, prompts evaluation of its role in patient care.

Purpose of the Study:

• To compare orthopaedic surgeons' assessment of AI-generated answers versus nurse-provided answers for common TKA patient questions.
• To evaluate patient comfort levels and trust in AI for postoperative TKA inquiries compared to nurse responses.

Main Methods:

• Prospective creation of 60 common TKA patient questions.
• Evaluation of answers from arthroplasty-trained nurses and ChatGPT-4 by 3 fellowship-trained orthopaedic surgeons.
• Assessment of patient comfort and trust in AI using REDCap surveys.

Main Results:

• Surgeons deemed 73.3% of nurse responses and 73.3% of ChatGPT responses appropriate.
• ChatGPT had fewer "inappropriate" or "unreliable" responses compared to nurses.
• More patients (53.8%) were comfortable with ChatGPT answers than nurse answers (34.0%), though many patients (92.1%) expressed uncertainty about trusting AI.

Conclusions:

• ChatGPT demonstrates potential for providing appropriate answers to TKA patient questions.
• Utilizing AI could minimize healthcare costs and maintain patient satisfaction.
• Successful AI implementation hinges on delivering credible information aligned with physician and patient objectives.