(443) Agentic AI in Gestational Diabetes Care: An Analysis of Patient Communication and Concerns

Objective:
To categorize patient-initiated messages to a virtual GDM assistant, identifying key concerns and communication patterns.

Study Design:
This prospective observational study (June 1–July 30, 2025) included ten newly diagnosed diet-controlled GDM patients. We developed a patient-facing, agentic AI tool for GDM that received FDA approval as an investigational device to support self-management. The WhatsApp-based assistant offered real-time support, logged glucose readings, provided immediate feedback, answered dietary and general health questions, and escalated concerning symptoms to clinicians. Patient messages were classified using a predefined taxonomy: glucose readings, dietary comments and questions, general health questions, symptom reports, greetings, acknowledgments, logistical inquiries, and “other.”

Results:
A total of 710 messages were analyzed. Most (50.7%) were glucose readings. Dietary content (17.1%) highlighted nutrition as a key educational need. General health questions comprised 3.2% of messages (e.g., “Does stress raise blood sugar?”). Symptom reports were infrequent (1.5%) but clinically important; two triggered clinician callbacks within 15 minutes. Social interactions (greetings, acknowledgments) totaled 18.7%. Logistical messages accounted for 1.0%, and “other” represented 7.7%. No manual overrides were required for safety alerts.

Conclusion:
Patients primarily used the agentic AI-powered assistant for glucose reporting and nutritional queries. This demonstrates an educational opportunity in dietary management and confirms the potential of AI to continuously support and adapt patient education strategies in real-time GDM management. A randomized controlled trial comparing traditional nutritional counseling with agentic AI–based diabetes management is underway.