ChatGPT和百度作为骨科相关疾病在线健康信息咨询工具的潜能评估  

Assessment of the potential of ChatGPT and Baidu as online health information consultation tools for orthopedic diseases

作　　者：陈阳[1] 王洁[2,3,4] 尹相杰 滕涛 张哲平[1] 仉建国 吴南[2,3,4,5] CHEN Yang;WANG Jie;YIN Xiangjie;TENG Tao;ZHANG Zheping;ZHANG Jianguo;WU Nan(Department of Orthopaedics,Beijing Puren Hospital,Beijing 100062,China;Department of Orthopaedics,Peking Union Medical College Hospital,Chinese Academy of Medical Sciences and Peking Union Medical College,Beijing 100730,China;Beijing Key Laboratory for Genetic Research of Skeletal Deformity,Beijing 100730,China;Key Laboratory of Big Data for Spinal Deformities,Chinese Academy of Medical Sciences,Beijing 100730,China;State Key Laboratory of Complex Severe and Rare Diseases,Beijing 100730,China)

机构地区：[1]北京市普仁医院骨科,北京100062 [2]中国医学科学院北京协和医学院北京协和医院骨科,北京100730 [3]骨骼畸形遗传学研究北京市重点实验室,北京100730 [4]中国医学科学院脊柱畸形大数据研究与应用重点实验室,北京100730 [5]疑难重症及罕见病全国重点实验室,北京100730

出　　处：《中华骨与关节外科杂志》2024年第9期837-843,共7页Chinese Journal of Bone and Joint Surgery

基　　金：中央高水平医院临床科研业务费(2022-PUMCH-D-004,2022-PUMCH-C-033)。

摘　　要：目的:通过模拟患者的网络检索,评估ChatGPT和百度作为患者寻求骨科相关疾病在线健康信息资源工具的潜能。方法:选取2017年6月至2023年6月于北京协和医院或北京市普仁医院就诊并被诊断为腰椎间盘突出症、膝骨关节炎、先天性脊柱侧凸、Klippel-Feil综合征的患者,通过在线问卷的方式收集患者关心的问题,每种疾病收集满50份有效问卷为止。将每种疾病所收集的问题按照病因、症状、病情进展、治疗、预后进行分类,最终每个方面筛选出2个问题,即每种疾病归纳为10个问题,最终4种疾病共计40个问题。分别用ChatGPT和百度进行信息检索,并对检索结果进行比较。结果:对于40个骨科相关疾病问题,ChatGPT的回答总体优于百度搜索,差异有统计学意义(P<0.001)。对于骨科常见病,ChatGPT回答优于百度搜索(P=0.042),且主要体现在预后相关问题上(P=0.040);对于骨科罕见病,ChatGPT回答优于百度搜索(P<0.001),且主要体现在病因(P=0.017)、症状(P=0.013)、预后(P=0.036)相关问题上。百度的检索结果对于骨科常见病相关问题的结果评级高于骨科罕见病(P<0.001),ChatGPT回答骨科常见疾病和罕见病的结果评级差异无统计学意义(P=0.445)。结论:相较于百度的检索结果,ChatGPT回答骨科相关疾病问题的可信度较高,可以满足大部分患者的一般医疗咨询需求。Objective:To assess the potential of ChatGPT and Baidu as tools for online health information consultation related to orthopedic conditions by simulating patient searches.Methods:Patients diagnosed with lumbar disc herniation,knee osteoarthritis,congenital scoliosis,and Klippel-Feil syndrome(n=50 per disease)at Peking Union Medical College Hospital and Beijing Puren Hospital were selected to complete an online questionnaire.The responses were categorized by etiology,symptoms,disease progression,treatment,and prognosis.Finally,two questions were selected for each category,resulting in 10 questions per disease and a total of 40 questions.These questions were used for searches on ChatGPT and Baidu,and the results were compared.Results:For the 40 orthopedic-related disease questions,the overall responses from ChatGPT were superior to those from Baidu,with a statistically significant difference(P<0.001).For common orthopedic diseases,ChatGPT performed better than Baidu did(P=0.042),particularly in questions about prognosis(P=0.040).For rare orthopedic diseases,ChatGPT's responses were better than those of Baidu(P<0.001),especially for questions related to etiology(P=0.017),symptoms(P=0.013),and prognosis(P=0.036).The accuracy of Baidu's responses for common diseases was significantly greater than that for rare diseases(P<0.001),whereas there was no statistically significant difference in ChatGPT responses between common and rare diseases(P=0.445).Conclusions:Compared to Baidu,ChatGPT offers more credible answers for orthopedic-related disease questions and can meet the general medical consultation needs of the majority of patients.

关 键 词：人工智能 大语言模型 腰椎间盘突出症 膝骨关节炎 先天性脊柱侧凸 KLIPPEL-FEIL综合征 

分 类 号：R68[医药卫生—骨科学]