ChatGPT Can Provide Satisfactory Answers to Assessments of Mock Clinical Photos Following Anterior Cruciate Ligament...

To evaluate the ability of ChatGPT-4o to provide clinical assessments of mock photos of knees and shoulders that had undergone arthroscopic or open surgery and answer common questions related to anterior cruciate ligament reconstruction and rotator cuff repair.

The authors created a series of mock postoperative knee and shoulder photos and common questions about anterior cruciate ligament reconstruction and rotator cuff repair. These clinical photos and questions were presented to GPT in isolated chat windows and without prior training, and its responses were graded by orthopaedic surgeons, physician assistants, and residents using a previously established 4-point grading system.

When grading ChatGPT's assessments of mock clinical photos, raters gave 18.8% to 75.0% total excellent grades and 0% to 25.0% unsatisfactory grades. When grading ChatGPT question answers, raters gave 4.5% to 77.3% total excellent grades and 0% to 9.1% unsatisfactory grades. Mean grades ranged from excellent to satisfactory, requiring moderate clarification, depending on the photo or question. There was wide variability in rater grading agreement, with attendings being the most critical graders.

ChatGPT-4o can provide some satisfactory assessments of simple postoperative clinical photos and responses to common questions after orthopaedic knee and shoulder surgery, as graded by orthopaedic surgeons, residents, and physician assistants. While this study highlights how ChatGPT could serve as a tool for generating draft responses to patient concerns after undergoing orthopaedic knee and shoulder surgery, it cannot yet provide reliable, independent clinical advice.

As digital interactions in health care continue to expand, large language models could play a role in enhancing communication between patients and the orthopaedic care team, potentially improving patient engagement and access to information.