Covid-19: How to Predict a Patient's Trajectory
Real-time analysis of COVID-19 patient data yields insights doctors and health systems can use right away.
COVID-19 can feel like a gamble. Some of the afflicted never even know they had it, while others die within days of falling ill. A new tool developed by a multidisciplinary team based at the Bloomberg School and the Johns Hopkins School of Medicine is helping doctors know the odds, if not beat them.
Doctors in the five hospitals of the Johns Hopkins Health System can open their patient’s electronic health records and see a number predicting the likelihood that the disease will progress to a severe status (requiring intubation) or death on each of the next seven days.
The numbers help doctors provide fine-tuned treatment and identify patients they may need to spend more time with and when. It also enables hospitals to allocate resources—such as ICU beds or ventilators—most effectively.
“Everyone agrees the future of medicine is to a capture the data being generated in the clinical practice, analyze it, then take what you learn to improve on the practice,” says Scott Zeger, PhD, MS, John C. Malone Professor of Biostatistics and Medicine, who helped develop the tool, called the COVID Inpatient Risk Calculator. His team had already been working on the problem of amalgamating massive amounts of clinical data in order to perform real-time analyses for immediate clinical use, he says, but the eruption of COVID-19 ratcheted efforts to a new level.
In just four months, they coded a bespoke application to work in the Hopkins version of EPIC, the medical records system.
The latest version of the tool updates every six hours using data pooled from all the roughly 4,000 past and current COVID patients in the system. Doctors can hover over a patient’s risk calculation to also see the factors that determined it, such as demographics like age and sex, as well as all the vital information and lab test results recorded about them since the time they checked into the hospital—everything from blood pressure to lung function.
Other Hopkins groups are adapting a similar approach to build calculators for many other diseases, including diabetes, multiple sclerosis, and scleroderma. Beyond improving treatment, this way of learning from data in general clinical practice has the potential to save our country hundreds of billions of dollars by eliminating inefficiencies, Zeger adds.
“That’s the public health approach to improving the practice of medicine,” he says. “My dream is that the kind of tools we’re developing will both improve patients’ outcomes and reduce the costs of health care.”