DOI: 10.30565/medalanya.1979219 ISSN: 2587-0319

Guideline Adherence, Clinical Safety, and Source Accuracy of Commercial Large Language Models in Perioperative GLP-1 and GLP-1/GIP Receptor Agonist Management: A Scenario-Based Comparative Study

Ebru Aladag
Show PDF Cite
Aim: This study compared the guideline adherence, clinical safety, risk stratification, source accuracy, and practical applicability of three commercial large language models in perioperative GLP-1 receptor agonist (GLP-1 RA) and dual GLP-1/glucose-dependent insulinotropic polypeptide receptor agonist (GLP-1/GIP RA) management scenarios.Methods: In this non-interventional scenario-based comparative study, 20 English-language perioperative clinical scenarios were submitted once to ChatGPT Plus, Claude Pro, and Gemini Advanced/Google AI Pro. A total of 60 first complete responses were evaluated. Responses were scored using a predefined 10-point composite scale across five domains and predefined binary quality and safety outcomes. Paired non-parametric analyses were used for model comparisons.Results: ChatGPT achieved uniformly maximal total scores (mean 10.00), followed by Gemini (mean 8.00) and Claude (mean 5.85). Total scores differed significantly across models (χ²(2)=37.333, p

More from our Archive