推进上肢急诊护理：关于ChatGPT在诊断和处理手及腕部创伤中潜在作用的一项试点研究。
Advancing emergency upper extremity care: A pilot study of ChatGPT's potential role in diagnosing and managing hand and wrist trauma.
作者信息
Rothchild Evan, Jung Geena, Aiello Christopher, Tanna Neil, Ricci Joseph A
机构信息
Division of Plastic Surgery, Montefiore Medical Center/Albert Einstein College of Medicine, New York, NY, USA.
University of Notre Dame, Notre Dame, IN, USA.
出版信息
J Hand Microsurg. 2025 Apr 23;17(4):100260. doi: 10.1016/j.jham.2025.100260. eCollection 2025 Jul.
PURPOSE
Hand and wrist trauma is a frequent cause of emergency room (ER) visits. However, hospitals often lack immediate hand specialist coverage. This study aims to evaluate the efficacy of Artificial Intelligence (AI) platforms like ChatGPT in aiding in the diagnosis and patient management of upper extremity trauma.
METHODS
Ten clinical vignettes depicting common hand and wrist emergency clinical situations were created by the senior author to represent a broad range of common upper extremity injuries. These were presented to plastic surgery residents and ChatGPT (version 4.0). The responder was tasked to provide a diagnosis, ER management, and definitive treatment plans for each vignette. Responses were collected and scored by two attending plastic surgeons, blinded to the source, on a scale of 0 (poor) to 30 (excellent). Univariate and linear regression models were utilized for analysis.
RESULTS
A total of 16 resident responses (9 junior and 7 senior) and 16 ChatGPT responses were collected for each of the 10 clinical scenarios. ChatGPT had significantly higher total average scores (mean = 26.6 vs. 22.7, p < 0.05) and ER management scores (mean = 9.9 vs. 6.7, p < 0.05) when compared to residents. We did not find any notable differences in diagnosis or definitive treatment scores between residents and ChatGPT responses. However, the study was not sufficiently powered to detect smaller effect sizes in these areas. No apparent correlations between scores and resident year of training were observed.
CONCLUSIONS
ChatGPT provided clinically accurate diagnosis and management plans for upper extremity trauma. Implementing AI in trauma management has the potential to improve the management of hand and wrist trauma in emergency settings by serving as a diagnostic and clinical reference tool for emergency medical providers. However, their integration into clinical practice should be carefully evaluated and focused on complementing, and not replacing, traditional consults. Ultimately, these tools could alleviate the burden placed on ERs and limit reliance on hand consults.
目的
手和腕部创伤是急诊室就诊的常见原因。然而，医院往往缺乏手外科专家的即时会诊。本研究旨在评估ChatGPT等人工智能（AI）平台在辅助上肢创伤的诊断和患者管理方面的效果。
方法
资深作者创建了10个临床病例，描述常见的手和腕部急诊临床情况，以代表广泛的常见上肢损伤。将这些病例呈现给整形外科住院医师和ChatGPT（版本4.0）。要求回答者为每个病例提供诊断、急诊处理和确定性治疗方案。由两名不知情的整形外科主治医师收集回答并评分，评分范围为0（差）至30（优）。采用单变量和线性回归模型进行分析。
结果
针对10个临床场景中的每一个，共收集了16份住院医师的回答（9名初级和7名高级住院医师）以及16份ChatGPT的回答。与住院医师相比，ChatGPT的总平均分显著更高（平均分为26.6对22.7，p<0.05），急诊处理评分也更高（平均分为9.9对6.7，p<0.05）。我们未发现住院医师和ChatGPT回答在诊断或确定性治疗评分上有任何显著差异。然而，该研究的样本量不足以检测这些领域中较小的效应量。未观察到评分与住院医师培训年份之间存在明显相关性。
结论
ChatGPT为上肢创伤提供了临床准确的诊断和管理方案。在创伤管理中应用人工智能有潜力通过作为急诊医疗提供者的诊断和临床参考工具，改善急诊环境下手和腕部创伤的管理。然而，应仔细评估它们融入临床实践的情况，并侧重于补充而非取代传统会诊。最终，这些工具可以减轻急诊室的负担，并减少对手部会诊的依赖。
Zotero 插件