Dartmouth Institute-led team developing universal toolkit to predict hospital readmission risk
An estimated 95% of hospitals in the United States use electronic medical records (EMR) to manage and report on patient medical history, including diagnoses and treatment. However, despite the widespread adoption of EMRs, the ability to leverage EMR data for patient safety monitoring and predictive analytics hasn't realized its full potential for monitoring patients hospitalized for acute myocardial infarction (AMI). Commonly known as heart attacks, AMI is one of the many conditions hospitals face penalties for if their readmission rates exceed the national average. Now a team of health systems researchers, epidemiologists, biostatisticians and computer scientists have come together to develop a new EMR toolkit to help improve continuity of care, especially for those at highest risk of returning to the hospital.
Led by Jeremiah Brown, PhD, MS, associate professor at The Dartmouth Institute for Health Policy and Clinical Practice, the research team has begun working on a four-year project to develop a universal toolkit that could be implementable in any EMR system and used to predict the risk of hospital readmission in real-time. The toolkit will focus on extracting complex information about patient health and healthcare factors, including social risk factors such as living status and social support at home. In particular, the team will work with natural language processing experts to turn narrative notes entered by doctors, nurses and social workers into health measures that can also be used to help identify risk and health status.
Supported by a $3.1 million R01 grant from the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01HL130828, the national collaborative includes Wendy Chapman, PhD, at the University of Utah and Michael Matheny, MD, MS, MPH, at Vanderbilt University.
"The focus of this project is to help hospitals improve continuity of care in directing and allocating resources to patients who face the highest risk of readmission," Brown said.
The new collaborative is moving the fields of medical informatics and predictive analytics into a new era. Each site will each develop a toolkit simultaneously and cross-validate it against different EMR systems.
"This is one of the first projects to develop and validate an informatics toolkit in multiple health systems," Brown said. The final toolkit will also be externally validated in the health system used by the national Veterans Administration, led by Michael Matheny in Nashville, Tenn.
"Overall, this project is designed to provide healthcare professionals with tools to maximize patient safety and health, especially after patients leave the hospital, and to help prevent unnecessary hospital stays in the future," Brown said.