Abstract

Diabetes affects 12% of Americans, yet only 5-7% receive diabetes self-management education and support (DSMES) within the first year of diagnosis, and 55% fail to meet glycemic targets. While digital DSMES interventions improve access, current programs lack personalization and real-time data integration, which are key features for sustained engagement and optimal outcomes. Continuous glucose monitoring (CGM) has demonstrated improved glycemic control, yet uptake and long-term adherence remain low despite expanded Medicare coverage. Artificial intelligence (AI) and multimodal machine learning (MML) offer promising approaches for personalizing diabetes management, but their integration with existing digital health platforms remains underdeveloped.

My research aims to transform diabetes care by developing and testing Dulce-SMART, an enhanced version of an evidence-based text-messaging DSMES program (Dulce Digital, by integrating CGM technology with an AI/MML system to deliver adaptive, personalized messaging based on real-time glucose patterns. This knowledge is critically needed to bridge the gap between available diabetes technologies and their real-world utilization, particularly among insulin-treated adults with type 2 diabetes (T2D). I will conduct a pilot randomized trial comparing Dulce-SMART to standard Dulce Digital with CGM alone. Through three specific aims, my study will: (1) develop and validate an AI/MML system with real-time feedback loops that synthesizes CGM data, user interactions, and contextual information; (2) compare CGM wear time and platform engagement metrics between groups at baseline, 3, and 6 months; and (3) evaluate feasibility, acceptability, and implementation potential using the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework. An exploratory aim will examine effect sizes for changes in HbA1c and CGM metrics to inform the design of future large-scale trials. Ultimately, this research addresses the gap between effective digital health technologies and their sustained real-world use. By leveraging AI/MML to deliver personalized, adaptive support that responds to individual glucose patterns, Dulce-SMART has the potential to transform diabetes management from one-size-fits-all solutions to adaptive systems that maximize patient engagement and improve health outcomes for people living with type 2 diabetes.