Nearly 20 percent of hospital patients are readmitted within 30 days of discharge, a $15 billion problem for both patients and the health-care system. Under the federal Affordable Care Act, Medicare is reducing its payments to hospitals with excessive readmission rates.
Yixin Chen, PhD, associate professor of computer science and engineering in the School of Engineering & Applied Science at Washington University in St. Louis, has received a $718,042 grant from the National Science Foundation to mine data from hospital records to improve an early-warning system that has been tested at Barnes-Jewish Hospital for several years. He is collaborating with Chenyang Lu, PhD, professor of computer science and engineering; Thomas Bailey, MD, and Marin Kollef, MD, both professors of medicine at WUSTL’s School of Medicine.
With the funding, Chen and his colleagues will develop a large database gathering data from various sources, including 34 vital signs, from routine clinical processes, real-time bedside monitoring and existing electronic data sources from patients in general wards at Barnes-Jewish Hospital. Then they will develop algorithms that will mine and analyze the data looking for any signs of potential deterioration or life-threatening event in a patient, such as a heart attack, stroke or septic shock.
First, they will apply their algorithms to the patient data, such as blood pressure, heart rate and oxygen saturation, to identify patients at high risk for their condition to worsen. Those identified as being at risk will then be attached to a commercial sensor that provides data on vital signs every minute, then transmits the data wirelessly to a server, where a second algorithm will analyze it to predict deterioration. The system will also provide an alert to physicians on the patients’ deteriorating condition with an explanation of the cause and suggest possible interventions.
“Our algorithms can detect potential deterioration by finding hidden patterns in large amounts of data,” Chen said. “These hidden patterns are hard to be detected manually.”
Although early-warning systems exist, Chen said they are inadequate because they require monitoring by overburdened clinical staff. But the team’s early-warning system would not require additional work by patient-care staff because it uses existing data, Kollef said.
Kollef and Bailey have been working on such a system for about eight years in response to a mandate by the Institute for Healthcare Improvement that hospitals reduce cardiac arrests and other sudden, life-threatening events in patients on general medical floors by implementing a system of rapid response teams. Because they wanted to expand the early-warning system and make improvements, they brought in Chen and Lu for their engineering expertise.
“Being physicians, this is something for which we need a lot of support from the engineering school,” Kollef said. “It’s a nice example of taking the clinical side and the engineering side and bringing them together to come up with a solution for a problem that hasn’t had a good solution in the past.”
Together, they plan to conduct a clinical study to evaluate the proposed system, with the goal of using the technology in clinical practice to reduce patient mortality rates and hospital readmissions as well as to improve administration of the U.S. health-care system.
The data will be kept secure, Chen said, through the hospital’s security standards and through a secure VPN connection with state-of-the-art encryption. No personal information will be included with the data.
For more information, visit here.