Most hospitalized patients are eager to get their discharge papers and go home — but far too many return to the hospital too soon. Nationally, about 20 percent of patients discharged from the hospital return within 30 days, according to the Agency for Healthcare Research and Quality.
The cost to their health — and the nation’s health care system — is considerable, notes Daniel Polsky, Bloomberg Distinguished Professor of Health Economics at Johns Hopkins Carey Business School.
“No one really wants to be in the hospital, and readmissions drive up health care spending,” says Polsky, whose research explores how health care is organized, managed, financed, and delivered, especially for low-income people.
To date, most research on hospital readmission has relied on data gathered in the hospital, up until the time of a patient’s discharge, he notes. In a new study published in Scientific Reports, he and his colleagues instead focused on the 30-day period after patients returned home. The clinical trial tracked patient activity levels using either a smartphone or wearable device. The researchers crunched the flow of data coming from these devices using machine learning tools, building on their earlier work that used more traditional regression techniques.
“We found that prediction of 30-day hospital readmission significantly improved when using remotely monitored patient data and machine learning approaches for analysis,” says Polsky.
A way to intervene
Attention to hospital readmission rates has intensified in the decade since the launch of the Affordable Care Act, which established penalties for hospitals that experience “excess” readmissions when compared to “expected” levels of readmissions. Since that time, hospitals have incurred nearly $1.9 billion in penalties, according to the American Hospital Association.