The coronavirus surge continues unabated. People fill airplanes for holiday travel despite urgent pleas to hunker down at home. Hospitals weigh how to ration critical care because they’re swamped with patients.
As emergency rooms brace for a surge atop a surge, UC Irvine has unveiled a free, online tool to help health care pros predict which patients might be safely sent home, and which might need the most critical care.
“The goal is to give an earlier alert to clinicians to identify patients who may be vulnerable at the onset,” said Daniel S. Chow, an assistant professor in residence in radiological sciences and first author of a study published in PLOS ONE, in a prepared statement.
The machine-learning model does this by weighing a detailed list of vitals — age, gender, body mass index, co-morbidities, respiratory rate, blood cell counts and more — to predict the probability that a patient’s condition will worsen within 72 hours, requiring a ventilator or ICU care.
Researchers began collecting COVID-19 patient data at UCI Health in January, and produced a prototype of the tool by March. Then they began to test it, finding that its predictions were accurate about 95% of the time.
“Interestingly, variables which have previously been reported to be associated with worse COVID-19 disease, most notably including older age and hypertension, were less predictive in our sample than body mass index, total number of comorbidities and several laboratory values,” the study reported.
But while the tool proved highly accurate with UCI Health patients — who lived in the same area and were primarily Asian American, Latino and Caucasian — would it work in other places, with different patient populations? They tested that with a random sample of 40 patients at Atlanta’s Emory University. It did.
“Front-line medical providers have been inundated with critically ill COVID-19 patients,” the paper said. “A simple web-based tool utilized at patient presentation may facilitate decision making by simplifying integration of numerous clinical variables … which can increase physician confidence in determining which patients may be discharged safely.”
That’s of particular utility when critical-care beds are in short supply, when physicians are treating higher-than-expected numbers of patients or are working outside of their standard practice, it said.
It’s not magic in and of itself, though. “You have to talk to your specialists, your doctors; you have to assess how many beds you have available and come together as a group to figure out how you want to use the tool,” said Peter Chang, assistant professor in residence in radiological sciences and designer of the machine-learning model.
The work was a collaboration between the School of Medicine, the Sue and Bill Gross School of Nursing, the Program in Public Health and the Department of Computer Science.
“(R)eliable prediction of disease severity is essential for critical care capacity management and may enable earlier targeted interventions to improve patient outcomes,” the paper said.
