MEDBRIDGE: AI-Driven Risk Stratification and Care Transition Intervention to Improve Diabetes Medication Management - PROJECT SUMMARY/ABSTRACT This 5-year K01 proposal supports Seung-Yup “Joshua” Lee, PhD, in developing an independent research program to improve diabetes outcomes through AI-supported interventions targeting high-risk care transitions. Patients with type 2 diabetes (T2D) often manage complex medication regimens (e.g., insulins and GLP-1 receptor agonists), and medication changes during transitions from hospital to home create critical vulnerability in diabetes care. Despite medication reconciliation efforts, unresolved medication discrepancies across providers remain a major barrier to glycemic control, particularly among socioeconomically disadvantaged populations. Communication gaps between inpatient and outpatient providers further compound these challenges. Current healthcare systems lack effective, data-driven tools to identify and support high-risk patients following hospital discharge. To address this gap, Dr. Lee received a Forge AHEAD Center Pilot Award (funded through an NIH research initiative) to descriptively examine medication variability across care transitions and its association with HbA1c elevation (......), diabetes-related emergency department visits, and hospitalizations using retrospective electronic health record (EHR) data. This proposed K01 will support the development of MEDBRIDGE (MEDication BRIDGE at care transition), an AI-driven risk stratification tool to identify patients at elevated risk of adverse diabetes outcomes by integrating medication data (e.g., insulin types, GLP-1 receptor agonists, and medication changes at discharge), clinical factors (e.g., HbA1c and comorbidities), and social determinants of health (e.g., area-level composite indices, insurance status, and self-reported social history). It will also support pilot feasibility testing of a tailored intervention delivered by a nurse case manager (NCM) and community health worker (CHW), informed by MEDBRIDGE risk predictions. The study includes three Specific Aims: Aim 1: Develop, assess, and improve MEDBRIDGE models. Aim 2: Co-design a MEDBRIDGE-supported NCM/CHWdelivered intervention and train NCM/CHW. Aim 3: Pilot test feasibility, acceptability, and fidelity of the intervention post-discharge in a single-arm study. Dr. Lee’s training objectives are to 1) gain knowledge and expertise in diabetes medication management and discrepancy analysis; 2) develop skills and experience in qualitative and mixed methods research approaches; 3) build capacity in intervention design and pilot trial implementation; 4) advance professional development and progress to research independence. This 5-year K01 award will provide Dr. Lee with the mentorship, training, and research experience needed to become a leader in AI-supported patient-centered diabetes care. By the end of the award period, Dr. Lee will have generated novel data on reducing disparities in T2D outcomes and submitted a competitive R01 to evaluate scalable, datainformed interventions that improve glycemic control and reduce acute care utilization in high-risk populations.
