Critical Analysis of ChatGPT's Responses on Anterior Cruciate Ligament Reconstruction: Accuracy and the Need for Specialist Supervision
R.E. De Paula, P.Ê.F. Bezerra, P.A.V. Maia, N.T. Cavanellas, R.S.P.E. Albuquerque, J.L.R. De FariaObjectives:
To evaluate the accuracy and clarity of answers provided by a generative language model (ChatGPT) to common questions about anterior cruciate ligament (ACL) reconstruction, and to determine the need for specialist supervision in interpreting this information.
Methods:
Ten questions frequently posed by patients and laypeople regarding ACL injury, including its definition, common surgical procedures, clinical features, and postoperative recovery, were submitted to a language model (ChatGPT, version 5.1, on 10/12/2025). The generated responses were independently evaluated by an orthopedic surgery specialist from a tertiary hospital, with over ten years of experience and a master's degree. Answers were classified for appropriateness (excellent, satisfactory, inadequate) and for the need for additional clarification (none, minimal, moderate).
Results:
Regarding appropriateness, 9 out of 10 responses (90%) were rated as "excellent." One response (10%) was classified as "satisfactory." This specific answer, concerning the typical length of hospital stay after surgery, demonstrated inaccuracies and omissions, resulting in a "moderate" need for clarification. The model inaccurately indicated discharge within a few hours, contrary to the common clinical practice of an overnight stay for observation.
Conclusion:
The evaluated language model demonstrated high accuracy (90% excellent responses) in answering common questions about ACL reconstruction. However, the identified inaccuracy in one response underscores the potential for occasional errors or oversimplifications. Therefore, while this technology is a useful informational tool, its outputs still require critical evaluation and supervision by a healthcare specialist to ensure patient safety and information reliability.