Assessing ChatGPT's Diagnostic Accuracy and Therapeutic Strategies in Oral Pathologies: A Cross-Sectional Study.

Abstract

The rapid adoption of artificial intelligence (AI) models in the medical field is due to their ability to collaborate with clinicians in the diagnosis and management of a wide range of conditions. This research assesses the diagnostic accuracy and therapeutic strategies of Chat Generative Pre-trained Transformer (ChatGPT) in comparison to dental professionals across 12 clinical cases. ChatGPT 3.5 was queried for diagnoses and management plans for 12 retrospective cases. Physicians were tasked with rating the complexity of clinical scenarios and their agreement with the ChatGPT responses using a five-point Likert scale. Comparisons were made between the complexity of the cases and the accuracy of the diagnoses and treatment plans. ChatGPT exhibited high accuracy in providing differential diagnoses and acceptable treatment plans. In a survey involving 30 attending physicians, scenarios were rated with an overall median difficulty level of 3, showing acceptable agreement with ChatGPT's differential diagnosis accuracy (overall median 4). Our study revealed lower diagnosis scores correlating with decreased treatment management scores, as demonstrated by univariate ordinal regression analysis. ChatGPT's rapid processing aids healthcare by offering an objective, evidence-based approach, reducing human error and workload. However, potential biases may affect outcomes and challenge less-experienced practitioners. AI in healthcare, including ChatGPT, is still evolving, and further research is needed to understand its full potential in analyzing clinical information, establishing diagnoses, and suggesting treatments.