High patient and surgeon satisfaction with ChatGPT-generated responses to real patient questions regarding total knee arthroplasty

Comprehensive Summary

This study evaluates the quality of ChatGPT answers to total knee arthroplasty (TKA) questions by patients. 220 questions from 44 patients were collected. The 10 most common questions were entered into ChatGPT-3.5 which was prompted to act like an arthroplasty surgeon responding to a patient preparing for TKA surgery. Two types of questions were asked with one requesting “an easy-to-understand answer” and the other without any additional request. The answers to the questions were sent to 81 surgeons and 53 patients to evaluate the answer. The mean satisfaction rating from the surgeons was 4.72/5 and that of the patients was 4.99/5. Moreover 55.6% surgeons reported that they would have used 80-100% of the AI’s answers as a response to their patients. The results therefore reflect that ChatGPT generates sufficient responses to TKA questions. Nevertheless, they cannot be a replacement for a healthcare professional.

Outcomes and Implications

Patients do not have access to a healthcare professional 24/7. Especially those undergoing major surgeries such as total knee arthroplasty (TKA) have many questions to ask. Therefore AI may be a helpful tool and accessible large language models such as ChatGPT can be a useful tool to help digest complex medical knowledge about their condition. Therefore there is a need to evaluate ChatGPT’s accuracy and educational performance. The study indicates that, overall, the model’s answers provide great satisfaction to the patients, indicating its potential as an educational and communicational tool for medical knowledge.