Using Artificial Intelligence to Improve Diabetes Medication Safety After Hospital Discharge
A newly funded project led by Alok Kapoor, MD, will explore how artificial intelligence can help people with diabetes better understand changes to their medications when leaving the hospital.
Hospital discharge can be a particularly vulnerable time for people with diabetes. Insulin and other glucose-lowering medications require careful attention to dosing, timing, meals and blood glucose levels, and medication regimens are often changed during a hospital stay. Patients may leave the hospital with new medications, different doses or new instructions, increasing the potential for misunderstanding and medication-related problems.
The project, funded by the Herman G. Berkman Diabetes Clinical Innovation Fund, will develop and test an interactive artificial intelligence tool using a communication technique known as teach-back. Rather than simply giving patients instructions, teach-back asks them to explain in their own words how they will take their medications. This allows misunderstandings to be identified and corrected before they lead to potential harm.
Using a secure artificial intelligence system restricted to carefully selected, safety-focused information, the tool will generate questions based on each patient's diabetes medication instructions. It will ask patients to explain important information such as medication changes, doses, timing and potential side effects, then identify areas they may have misunderstood and provide clarification appropriate to their reading level.
The project will first be tested and refined with input from patients and clinicians. It will then be evaluated with adults who are starting diabetes medications or whose diabetes medications have changed during hospitalization. Researchers will examine how effectively the tool identifies and corrects misunderstandings, as well as patients' satisfaction, engagement and confidence using it. A clinician will be present during all patient testing to identify and correct any errors made by the artificial intelligence tool.
The project brings together expertise in medication safety, diabetes care and health informatics. Dr. Kapoor, Professor of Medicine and a hospitalist with expertise in medication safety and care transitions, will collaborate with Adrian Zai, MD, PhD, Chief Research Informatics Officer at UMass Chan Medical School, and Alexandra Albert, MD, PhD, Director of the Inpatient Diabetes Service.
The goal is to develop a safe, scalable way to reinforce diabetes medication education so patients leave the hospital with a clearer understanding of how to manage their medications and greater confidence in caring for their diabetes at home.
Herman G. Berkman Diabetes Clinical Innovation Fund Recipients
On-Demand Diabetes Education Videos for Patients and Families
A series of short, on-demand videos addressing common diabetes topics, will be created, including blood glucose monitoring, insulin use, sick-day management and diabetes technology. The evidence-based videos are designed to reinforce, not replace, education provided by Certified Diabetes Care and Education Specialists. Patients, families and caregivers will be able to watch, pause and replay the videos as needed, providing convenient online access to reliable diabetes information. The project, led by nurse practitioner Clare Foley, DNP, and Adam Edelstein, also aims to make the videos available broadly throughout UMass Memorial Health so providers can easily share them electronically.
Using Artificial Intelligence to Improve Diabetes Medication Safety After Hospital Discharge
This project will develop and test an artificial intelligence tool designed to help patients better understand changes to their diabetes medications when leaving the hospital. The goal is to reduce misunderstandings and help patients return home with greater knowledge and confidence to safely manage their diabetes medications. The team consisting of Alok Kapoor, MD, Adrian Zai, MD, PhD, and Alexandra Albert, MD, PhD, will evaluate the tool's accuracy, safety and usability with patients starting new diabetes medications or whose medications changed during hospitalization.
Clinical Study on Diabetes Management During Pregnancy
A clinical study at UMass Memorial Medical Center is comparing continuous glucose monitoring (CGM) to multiple daily fingersticks for pregnant women with type 2 diabetes. The randomized study, led by Gianna Wilkie, MD, Assistant Professor of Obstetrics and Gynecology, was awarded funds to conduct includes maternal blood glucose control, patient satisfaction, and other perinatal outcomes.
Implementing a Liver Disease Screening Process in the Adult Diabetes Clinic
Liver disease is strongly associated with type 2 diabetes and obesity. It remains underdiagnosed and undertreated, and many people living with T2D are unaware they have it. This project, led by endocrinologist Madona Azar, MD, implemented a new process in the UMass Memorial diabetes clinic that uses a screening tool to analyze clinically available data to determine patients' risk of liver fibrosis.
AI Diabetic Retinopathy Screening in Primary Care
This project implemented an artificial intelligence (AI)- based diabetic retinopathy screening program in Family Medicine clinics to detect eye disease and improve comprehensive care for people living with diabetes. Recent studies have identified AI-based algorithms as promising tools for screening and early detection of diabetic retinopathy, helping those at risk. This study, led by optometrist Juan Ding, OD, PhD, tested the diagnostic accuracy of a hand-held AI-assisted camera used by primary care physicians to screen at-risk individuals for retinal changes indicative of diabetic retinopathy.
Analying the Benefits of Continuous Glucose Monitors to Reduce Hospitalizations and Diabetic Complications
This randomized clinical trial provided continuous glucose monitors (CGM) to people with diabetes who were currently not using one and arrived at the Emergency Room with high or low blood sugar, or other diabetes-related complications. The recently completed study, led by endocrinologist Dr. Mark O’Connor, analyzes whether CGM successfully prevents people from returning to the ER with diabetes-related issues, compared with the control group who do not wear a device to monitor their blood sugar.
Improving Inpatient Blood Glucose Management
This project aimed to implement a carbohydrate-counting system for hospitalized inpatients with diabetes across the UMass Memorial Health system. Endocrinologist Leslie Domalik, MD, evaluated whether adopting a flexible meal dosing option based on carb counting would improve the outcomes of hospitalized patients with diabetes. By coordinating the timing of blood glucose testing, insulin dosing, and the administration of rapid-acting mealtime insulin, she wanted to ensure carbohydrate counts are listed for all food served to hospitalized patients and to better coordinate insulin delivery with meal delivery.
Improving Care Access for the Highest Risk Diabetes Patients
The inaugural Herman Berkman Diabetes Clinical Innovation funding was awarded to Daniel Amante, PhD, and Adarsha Bajracharya, MD, in 2019. It led to Dr. Amante receiving a three-year KL2 Mentored Career Development Training grant to develop a Diabetes Mellitus program using Behavioral economics to Optimize Outreach and Self-management support with Technology (DM-BOOST).
ID PLUS Care was a multidisciplinary, collaborative approach to improve care access, quality, and management for at-risk patients with diabetes. The program monitored Electronic Health Record data to identify UMass Medicare Accountable Care Organization patients at risk for negative outcomes and proactively contacted them to nudge them toward recommended services.