Credibility of ChatGPT in the assessment of obesity in type 2 diabetes according to the guidelines

• Published in: International Journal of Obesity Vol. 48; no. 2; pp. 271 - 275
• Main Authors: Barlas, Tugba, Altinova, Alev Eroglu, Akturk, Mujde, Toruner, Fusun Balos
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 01.02.2024; Nature Publishing Group
• Subjects: 692/163/2743/137/773; 692/163/2743/393; Body weight loss; Chatbots; Compatibility; Cross-Sectional Studies; Diabetes; Diabetes mellitus (non-insulin dependent); Diabetes Mellitus, Type 2 - diagnosis; Diabetes Mellitus, Type 2 - epidemiology; Diabetes Mellitus, Type 2 - therapy; Endocrinology; Epidemiology; Evaluation; Guidelines; Health care; Health Personnel; Health Promotion and Disease Prevention; Humans; Internal Medicine; Medicine; Medicine & Public Health; Metabolic Diseases; Obesity; Obesity - complications; Obesity - epidemiology; Patients; Public Health; Questions; Referral and Consultation; Weight control; Weight loss
• ISSN: 0307-0565; 1476-5497; 1476-5497
• DOI: 10.1038/s41366-023-01410-5

Background The Chat Generative Pre-trained Transformer (ChatGPT) allows students, researchers, and patients in the medical field to access information easily and has gained attention nowadays. We aimed to evaluate the credibility of ChatGPT according to the guidelines for the assessment of obesity in type 2 diabetes (T2D), which is one of the major concerns of this century. Materials and method In this cross-sectional non-human subject study, experienced endocrinologists posed 20 questions to ChatGPT in subsections, which were assessments and different treatment options for obesity according to the American Diabetes Association and American Association of Clinical Endocrinology guidelines. The responses of ChatGPT were classified into four categories: compatible, compatible but insufficient, partially incompatible and incompatible with the guidelines. Results ChatGPT demonstrated a systematic approach to answering questions and recommended consulting a healthcare provider to receive personalized advice based on the specific health needs and circumstances of patients. The compatibility of ChatGPT with the guidelines was 100% in the assessment of obesity in type 2 diabetes; however, it was lower in the therapy sections, which included nutritional, medical, and surgical approaches to weight loss. Furthermore, ChatGPT required additional prompts for responses that were evaluated as “compatible but insufficient” to provide all the information in the guidelines. Conclusion The assessment and management of obesity in T2D are highly individualized. Despite ChatGPT’s comprehensive and understandable responses, it should not be used as a substitute for healthcare professionals’ patient-centered approach.