$2.5 million funding helps researchers to advance tools for predicting readmission in diabetic patients

Each year in the United States, more than 1 million patients with diabetes make return trips to the hospital for diabetes-related illness, often being readmitted within 30 days of their initial hospitalization. The costs of these return visits add up, in terms of dollars and in terms of the toll on patient health that comes with prolonged or chronic illness and repeated hospitalization.

Some patients are especially prone to hospital readmission, and now, with $2.5 million in funding from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), researchers at the Lewis Katz School of Medicine at Temple University (LKSOM) are poised to better predict which patients with diabetes are most likely to experience additional hospital visits. Building from a model previously generated at Temple and known as the Diabetes Early Readmission Risk Indicator (DERRI^TM), the researchers aim to develop improved models that accurately and automatically predict 30-day readmission risk among patients with diabetes, based on electronic health record (EHR) data.

"The DERRI^TM app originally was developed for point-of-care use, with patients manually inputting information on a small number of variables, such as insulin use, co-existing conditions and recent hospital discharge," explained Daniel J. Rubin, MD, MSc, FACE, Associate Professor of Medicine at LKSOM, Chair of the Glycemic Control Taskforce at Temple University Hospital and lead investigator on the new five-year NIDDK grant. "Now, we want to upgrade DERRI^TM to build models that bypass manual input and instead draw directly on EHR data, allowing for faster and more accurate prediction."

Dr. Rubin's collaborators on the project include Dr. Zoran Obradovic, Laura H. Carnell Professor of Data Analytics, Director of the Center for Data Analytics and Biomedical Informatics and Professor in the Computer and Information Sciences Department at Temple University; and Anuradha Paranjape, MD, MPH, FACP, Vice Chair for Clinical Affairs in the Department of Medicine and Professor of Medicine and Professor of Clinical Sciences at LKSOM.

The new models, named eDERRI^TM, will incorporate data from the PaTH Clinical Data Research Network (CDRN), a multi-center, 40-plus hospital member of the National Patient-Centered Clinical Research Network, of which Temple University Health System is a member. With Dr. Obradovic's expertise, the researchers will be able to apply state-of-the-art deep-learning methods to optimize predictive modeling.

Once the optimized eDERRI^TM models are completed, the researchers aim to translate them into a readmission risk prediction tool that automatically captures EHR data. Using this data, the researchers anticipate that the tool will accurately identify risk factors for individual patients with diabetes and thereby predict readmission risk. Dr. Rubin's team plans to validate the eDERRI^TM models with PaTH CDRN data collected subsequent to the model's development. The eDERRI^TM tool additionally will be tested in patients with diabetes who are admitted to Temple University Hospital.

The novel tool stands to greatly advance the care and treatment of patients with diabetes who are especially vulnerable to severe illness.

Our proposed eDERRI^TM tool could be widely used for predicting risks, reducing costs and improving care. We are building and testing it at Temple, but the EHR component will be built in EPIC, which is used at numerous institutions across the country. Our tool could be rolled out to any hospital or treatment center that uses EPIC systems."

Dr. Daniel J. Rubin, Associate Professor of Medicine, LKSOM