How Does ChatGPT Perform on the Italian Residency Admission National Exam Compared to 15,869 Medical Graduates?

Purpose: The study aims to assess ChatGPT performance on the Residency Admission National Exam to evaluate ChatGPT's level of medical knowledge compared to graduate medical doctors in Italy. Methods: ChatGPT3 was used in June 2023 to undertake the 2022 Italian Residency Admission National Exam—a 140 multiple choice questions computer-based exam taken by all Italian medical graduates yearly, used to assess basic science and applied medical knowledge. The exam was scored using the same criteria defined by the national educational governing body. The performance of ChatGPT was compared to the performance of the 15,869 medical graduates who took the exam in July 2022. Lastly, the integrity and quality of ChatGPT's responses were evaluated. Results: ChatGPT answered correctly 122 out of 140 questions. The score ranked in the top 98.8^th percentile among 15,869 medical graduates. Among the 18 incorrect answers, 10 were evaluating direct questions on basic science medical knowledge, while 8 were evaluating candidates' applied clinical knowledge and reasoning under the form of case presentation. Errors were logical (2 incorrect answers) and informational in nature (16 incorrect answers). Explanations to the correct answers were all evaluated as "appropriate." Comparison to national statistics related to the minimal score needed to match into each specialty, demonstrated that the performance of ChatGPT would have granted the candidate a match into any specialty. Conclusion: ChatGPT proved to be proficient in basic science medical knowledge and applied clinical knowledge. Future research should assess the impact and reliability of ChatGPT in clinical practice. [ABSTRACT FROM AUTHOR]