Photo Credit: Delmaine Donson

The following is a summary of “Management of Lower Extremity Traumas: Comparing Appropriate Use Criteria with ChatGPT Recommendations,” published in the April 2025 issue of American Journal of Surgery by Lu et al. 

High-energy lower extremity injuries often present significant clinical challenges, particularly when determining the optimal course of treatment. The decision to attempt limb salvage versus early amputation is fraught with complexity, as the goal is typically to preserve the limb whenever possible; however, amputation may be necessary in certain cases to prevent further complications, such as infection or irreversible tissue damage. As AI tools like ChatGPT-4 gain popularity in clinical decision-making, the potential for these technologies to assist in such high-stakes decisions is an area of growing interest. The primary objective of this study was to assess whether ChatGPT-4 can accurately recommend limb salvage or early amputation based on various patient scenarios involving lower extremity trauma. 

To achieve this, a series of scenarios were developed that represented common and complex lower leg injuries, which were then compared to the limb salvage and amputation criteria provided by the American Academy of Orthopaedic Surgeons (AAOS). Both AAOS guidelines and ChatGPT-4 were tasked with providing treatment recommendations based on the scenarios, and a recommendation score was assigned for both limb salvage and early amputation. The results were analyzed for statistical significance to determine how closely ChatGPT-4’s recommendations aligned with established medical guidelines. A total of 196 patient scenario combinations were utilized in the analysis. The mean errors for limb salvage and early amputation recommendations were found to be -0.3 and -0.2, respectively. 

Statistical analysis revealed that both AAOS and ChatGPT-4 showed significant positive correlations in predicting the appropriateness of limb salvage and early amputation treatment options. The effect size for limb salvage was -0.094, while the effect size for early amputation was -0.14, both indicating moderate levels of agreement between the AI tool and clinical guidelines. While ChatGPT-4 generally underestimated the appropriateness scores for both treatment options—limb salvage and early amputation—its recommendations were still consistent with those of AAOS, suggesting that ChatGPT-4 is a potentially useful tool for clinicians to consider when making treatment decisions. However, given the slight underestimation in its scoring, healthcare providers should exercise caution and supplement AI-generated recommendations with their own clinical judgment and expertise. 

Further research is needed to refine AI’s role in complex clinical decision-making and explore its integration into real-time clinical workflows for trauma care.

Source: americanjournalofsurgery.com/article/S0002-9610(25)00051-0/abstract