ChatGPT-4 Omni Performance in USMLE Disciplines and Clinical Skills: Comparative Analysis

Abstract

Background: Recent studies, including those by the National Board of Medical Examiners, have highlighted the remarkable capabilities of recent large language models (LLMs) such as ChatGPT in passing the United States Medical Licensing Examination (USMLE). However, there is a gap in detailed analysis of LLM performance in specific medical content areas, thus limiting an assessment of their potential utility in medical education.

Objective: This study aimed to assess and compare the accuracy of successive ChatGPT versions (GPT-3.5, GPT-4, and GPT-4 Omni) in USMLE disciplines, clinical clerkships, and the clinical skills of diagnostics and management.

Methods: This study used 750 clinical vignette-based multiple-choice questions to characterize the performance of successive ChatGPT versions (ChatGPT 3.5 [GPT-3.5], ChatGPT 4 [GPT-4], and ChatGPT 4 Omni [GPT-4o]) across USMLE disciplines, clinical clerkships, and in clinical skills (diagnostics and management). Accuracy was assessed using a standardized protocol, with statistical analyses conducted to compare the models' performances.

Results: GPT-4o achieved the highest accuracy across 750 multiple-choice questions at 90.4%, outperforming GPT-4 and GPT-3.5, which scored 81.1% and 60.0%, respectively. GPT-4o's highest performances were in social sciences (95.5%), behavioral and neuroscience (94.2%), and pharmacology (93.2%). In clinical skills, GPT-4o's diagnostic accuracy was 92.7% and management accuracy was 88.8%, significantly higher than its predecessors. Notably, both GPT-4o and GPT-4 significantly outperformed the medical student average accuracy of 59.3% (95% CI 58.3-60.3).

Conclusions: GPT-4o's performance in USMLE disciplines, clinical clerkships, and clinical skills indicates substantial improvements over its predecessors, suggesting significant potential for the use of this technology as an educational aid for medical students. These findings underscore the need for careful consideration when integrating LLMs into medical education, emphasizing the importance of structured curricula to guide their appropriate use and the need for ongoing critical analyses to ensure their reliability and effectiveness.

Keywords: AI in medical education; ChatGPT; ChatGPT 3.5; ChatGPT 4; ChatGPT 4 Omni; LLM; USMLE; United States Medical Licensing Examination; artificial intelligence in medicine; clinical skills; educational technology; large language model; medical education; medical licensing examination; medical student resources; medical students.

Figure 1.. Analysis of ChatGPT models’ and medical students’ performance on USMLE questions. This figure displays the comparative accuracies of ChatGPT 3.5 (GPT-3.5), ChatGPT 4 (GPT-4), ChatGPT 4 Omni (GPT-4o), and medical students in answering a set of 750 USMLE-style questions. The overall accuracy, preclinical accuracy, and clinical accuracy are shown. Asterisks (*) denote statistically significant differences (P<.05), highlighting the advancements in newer models of the GPT series. The number of questions is indicated for each category: n=750 for overall accuracy, n=375 for preclinical accuracy, and n=375 for clinical accuracy. GPT: Generative Pre-trained Transformer; USMLE: United States Medical Licensing Examination.

Figure 2.. Influence of question difficulty on response accuracy compared to medical student performance. This figure illustrates the effect of clinical vignette difficulty on the response accuracy of ChatGPT 3.5 (GPT-3.5), ChatGPT 4 (GPT-4), and ChatGPT 4 Omni (GPT-4o) in comparison to medical students. The bar graph represents the percentage of correct responses across different tiers of difficulty, ranging from tier 1 (most difficult) to tier 5 (easiest). The number of questions for each difficulty tier is n=10 for tier 1, n=89 for tier 2, n=263 for tier 3, n=302 for tier 4, and n=81 for tier 5.

Figure 3.. Performance of ChatGPT models in diagnostics and management compared to medical students. This figure compares the performance of ChatGPT 3.5 (GPT-3.5), ChatGPT 4 (GPT-4), and ChatGPT 4 Omni (GPT-4o) in the clinical domains of diagnostics and management. The bar graph shows the percentage of correct responses for each model and medical students in the diagnosis (n=164) and management (n=178) categories. GPT-4o exhibits the highest accuracy in both categories, followed by GPT-4, with GPT-3.5 showing the lowest performance. Asterisks (*) denote statistically significant differences (P<.05), emphasizing the advancements in newer models of the GPT series. GPT: Generative Pre-trained Transformer.