Mining data on hospital readmissions to improve care
One of the biggest drains on the U.S. healthcare system is the revolving door of hospital readmissions.
Since October 2012, the Centers for Medicare & Medicaid Services has taken steps to reduce the high rates of readmission for patients with heart failure, heart attack, pneumonia and other chronic conditions. Hospitals are incentivized to improve care and prevent a patient’s return trip within 30 days of discharge.
The work of Senjuti Basu Roy, assistant professor of computer science at NJIT, aims to help hospitals more accurately predict the risk of readmission for heart-failure patients and to focus on quality patient care and resource utilization. Dr. Roy and her colleagues devised a novel data-mining framework that predicts the risk of readmission in a series of stages and integrates different variables such as gender, race, marital status and severity of illness at each stage. Using real-world hospital patient data, their analytic model significantly outperformed other predictive methods. Members of the research team also created an innovative framework of algorithms to help clinicians employ personalized treatment plans to minimize the 30-day readmission risk for individual patients.