261-OR: Validation of Mitraa, a GPT-Based Conversational Model for Enhancing Diabetes Self-Management and Education



Introduction and Objective: The growing prevalence of diabetes underscores the need for scalable, patient-centric tools to support diabetes self-management and education (DSME). Mitraa, a GPT-based conversational AI model, was developed to provide personalized diabetes care, including guidance on nutrition, medication adherence, glucose monitoring, and lifestyle adjustments. This study validates Mitraa’s effectiveness in delivering accurate, patient-relevant recommendations compared to certified diabetes educators (CDEs).Methods: A cohort of 250 adults with diabetes (both Type 1 and Type 2) participated in a validation study. Participants interacted with Mitraa and a certified diabetes educator on identical scenarios covering glucose management, dietary advice, and medication adjustments. The accuracy of Mitraa’s responses was assessed using a predefined scoring rubric (accuracy, relevance, and safety) by an independent panel of diabetes specialists. Secondary outcomes included patient satisfaction, usability (via System Usability Scale, SUS), and comprehension of provided recommendations.Results: Mitraa achieved an 89% accuracy score in providing safe and clinically relevant recommendations, comparable to CDEs (92%, p=0.08). Patient satisfaction with Mitraa was high, with a mean SUS score of 82. Time efficiency was superior with Mitraa, reducing interaction time by 35% compared to CDE consultations. Additionally, 95% of participants reported understanding Mitraa’s recommendations, indicating its suitability for patient education. Feedback highlighted Mitraa’s empathetic tone and adaptability to diverse cultural contexts as strengths.Conclusion: Mitraa demonstrates high accuracy, usability, and patient satisfaction in supporting diabetes management. Its scalability and cost-effectiveness position it as a promising adjunct to healthcare providers in delivering DSME. Further integration with electronic medical records and real-time glucose data could enhance its clinical utility.

Disclosure

B. Saboo: None. S. Saboo: None. H.A. Hirani: None. A.D. Modi: None.



Source link