Assessing the Accuracy and Reliability of AI-Generated Medical Responses: An Evaluation of the Chat-GPT Model

Abstract

Background: Natural language processing models such as ChatGPT can generate text-based content and are poised to become a major information source in medicine and beyond. The accuracy and completeness of ChatGPT for medical queries is not known. Methods: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes/no) or descriptive answers. The physicians then graded ChatGPT-generated answers to these questions for accuracy (6-point Likert scale; range 1 – completely incorrect to 6 – completely correct) and completeness (3-point Likert scale; range 1 – incomplete to 3 - complete plus additional context). Scores were summarized with descriptive statistics and compared using Mann-Whitney U or Kruskal-Wallis testing. Results: Across all questions (n=284), median accuracy score was 5.5 (between almost completely and completely correct) with mean score of 4.8 (between mostly and almost completely correct). Median completeness score was 3 (complete and comprehensive) with mean score of 2.5. For questions rated easy, medium, and hard, median accuracy scores were 6, 5.5, and 5 (mean 5.0, 4.7, and 4.6; p=0.05). Accuracy scores for binary and descriptive questions were similar (median 6 vs. 5; mean 4.9 vs. 4.7; p=0.07). Of 36 questions with scores of 1-2, 34 were re-queried/re-graded 8-17 days later with substantial improvement (median 2 vs. 4; p<0.01). Conclusions: ChatGPT generated largely accurate information to diverse medical queries as judged by academic physician specialists although with important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Assessment of Artificial Intelligence Chatbot Responses to Top Searched Queries About Cancer

“…Such artificial intelligence (AI) chatbots generate responses based on a large corpus of data including articles, websites, and other publicly available text data. 1 Artificial intelligence chatbots are rapidly becoming a primary source of information for patients, and chatbot responses to medical queries may influence health-related behavior. Among AI chatbots, ChatGPT (OpenAI) has shown promise in accurately answering medical questions, even US Medical Licensing Examination-style questions.…”

ChatGPT and Physicians’ Malpractice Risk

“…They rated the output not “so incorrect as to cause patient harm” 91% to 93% of the time, but concordance with the results generated by a consultation service run by physicians and informatics experts analyzing aggregated electronic health record data was just 21% to 41%. Another study in which physicians evaluated ChatGPT 3.5 output on 180 clinical queries found that the mean score was 4.4 of 6 for accuracy and 2.4 for completeness, with 8% of answers scored as completely incorrect . In a third study, ChatGPT 3.5 responses to 36 clinical vignettes, compared with the clinical manual from which the vignettes were drawn, were scored as 72% accurate on average.…”

“…Another study in which physicians evaluated ChatGPT 3.5 output on 180 clinical queries found that the mean score was 4.4 of 6 for accuracy and 2.4 for completeness, with 8% of answers scored as completely incorrect. 6 In a third study, ChatGPT 3.5 responses to 36 clinical vignettes, compared with the clinical manual from which the vignettes were drawn, were scored as 72% accurate on average. The researchers characterized this as "impressive accuracy," but acknowledged that even small errors can harm patients.…”

ChatGPT and other artificial intelligence chatbots and biomedical writing

“…The daunting challenge requires critical and insightful analysis of the data by expert minds to discern reliable data, which is not simply possible by text mining. In clinical medicine, the preliminary studies suggest that ChatGPT could generate reasonably but not perfectly accurate responses to medical questions, write patient clinic letters, or summarize discharge notes [2][3][4][5][6] . However, all ChatGPTgenerated medical documents must be reviewed and verified by experts.…”