Department of Orthopedic Surgery, Vanderbilt University Medical Center, Nashville, Tennessee.
J Knee Surg. 2024 Jul;37(9):664-673. doi: 10.1055/s-0044-1782233. Epub 2024 Mar 5.
The internet has introduced many resources frequently accessed by patients prior to orthopaedic visits. Recently, Chat Generative Pre-Trained Transformer, an artificial intelligence-based chat application, has become publicly and freely available. The interface uses deep learning technology to mimic human interaction and provide convincing answers to questions posed by users. With its rapidly expanding usership, it is reasonable to assume that patients will soon use this technology for preoperative education. Therefore, we sought to determine the accuracy of answers to frequently asked questions (FAQs) pertaining to total knee arthroplasty (TKA).Ten FAQs were posed to the chatbot during a single online interaction with no follow-up questions or repetition. All 10 FAQs were analyzed for accuracy using an evidence-based approach. Answers were then rated as "excellent response not requiring clarification," "satisfactory requiring minimal clarification," satisfactory requiring moderate clarification," or "unsatisfactory requiring substantial clarification."Of the 10 answers given by the chatbot, none received an "unsatisfactory" rating with the majority either requiring minimal (5) or moderate (4) clarification. While many answers required nuanced clarification, overall, answers tended to be unbiased and evidence-based, even when presented with controversial subjects.The chatbot does an excellent job of providing basic, evidence-based answers to patient FAQs prior to TKA. These data were presented in a manner that will be easily comprehendible by most patients and may serve as a useful clinical adjunct in the future.
互联网为患者在进行骨科就诊之前提供了许多经常访问的资源。最近,基于人工智能的聊天应用程序 Chat Generative Pre-Trained Transformer 已公开且免费提供。该界面使用深度学习技术模拟人机交互,并为用户提出的问题提供令人信服的答案。随着其用户群的迅速扩大,可以合理地假设患者将很快使用这项技术进行术前教育。因此,我们试图确定该技术对全膝关节置换术(TKA)相关常见问题(FAQ)的回答的准确性。
在一次在线互动中,我们向聊天机器人提出了 10 个 FAQ,没有后续问题或重复。使用基于证据的方法对所有 10 个 FAQ 进行了准确性分析。然后,根据需要澄清的程度,将答案评为“无需澄清的优秀回复”、“需要最小澄清的满意回复”、“需要适度澄清的满意回复”或“需要大量澄清的不满意回复”。
聊天机器人给出的 10 个答案中,没有一个被评为“不满意”,其中大多数需要最小程度(5)或中等程度(4)的澄清。虽然许多答案需要细微的澄清,但总体而言,答案往往是客观的和基于证据的,即使是在涉及有争议的主题时也是如此。
聊天机器人在 TKA 之前为患者提供基本的、基于证据的 FAQ 答案方面做得非常出色。这些数据以大多数患者都能轻松理解的方式呈现,将来可能成为有用的临床辅助手段。