Can ChatGPT Function as a Virtual Multidisciplinary Team? A Proof-of-Concept Study in Vascular Malformation Syndromes (Preprint)

Abstract

<sec> <title>BACKGROUND</title> Sturge-Weber syndrome (SWS) and Klippel-Trenaunay syndrome (KTS) are complex vascular malformation syndromes that require multidisciplinary team (MDT) management. However, the traditional MDT approach faces challenges such as time coordination, geographical barriers, and inefficiencies in cross-disciplinary communication. </sec> <sec> <title>OBJECTIVE</title> This study aims to evaluate ChatGPT’s potential in simulating MDT decision-making for SWS and KTS by comparing its diagnostic and treatment recommendations with traditional MDT conclusions. </sec> <sec> <title>METHODS</title> A case-based proof-of-concept design was employed, retrospectively analyzing MDT records of two SWS and two KTS patients. Clinical data, imaging, and laboratory results were input into ChatGPT, and its outputs were evaluated by two dermatology experts using a 1-5 Likert scale across five dimensions: accuracy, completeness, appropriateness, insight, and safety. </sec> <sec> <title>RESULTS</title> ChatGPT performed well in most dimensions, particularly in appropriateness, but showed occasional uncertainty in handling complex or rare cases, such as in gene-phenotype associations. Inter-rater reliability ranged from negligible to moderate (ICC -0.00 to 0.63), with no significant differences observed between the experts’ ratings (p &gt; 0.05). </sec> <sec> <title>CONCLUSIONS</title> ChatGPT shows strong potential as a tool for simulating MDT decision-making, particularly in diagnostic accuracy, completeness of recommendations, and cross-disciplinary insights. However, it has limitations in managing complex cases and ensuring the feasibility of its recommendations. Future studies with larger sample sizes and multi-center validation are needed to fully assess its clinical value. </sec>

Ähnliche Arbeiten