A team of Johns Hopkins University biomedical engineers and heart specialists have developed an algorithm that warns doctors several hours before hospitalized COVID-19 patients experience cardiac arrest or blood clots.
The COVID-HEART predictor can forecast cardiac arrest in COVID patients with a median early warning time of 18 hours and an predict blood clots three days in advance. It was developed with data from 2,178 patients treated at the five hospitals in the Johns Hopkins Health System between March 1 and Sept. 27.
“It’s an early warning system to predict in real time these two outcomes in hospitalized COVID patients,” said senior author Natalia Trayanova, the Murray B. Sacks professor of Biomedical Engineering and a professor of medicine. “The continuously updating predictor can help hospitals allocate the appropriate resources and proper interventions to attain the best outcomes for patients.”
The paper is posted on the pre-print website called MedRxiv where scientists throughout the pandemic have been sharing urgent COVID-related findings to improve understanding of the new disease. Trayanova’s lab began the research in April with one of the first grants awarded from the National Science Foundation’s Rapid Response Research effort.
Julie K. Shade, lead author and a doctoral student in the Department of Biomedical Engineering in the Whiting School of Engineering, built the machine-learning algorithm with more than 100 clinical data points, demographic information and laboratory results obtained from the JH-CROWN registry that Johns Hopkins established to collect COVID data from every patient in the hospital system. She added other variables that she found reported by doctors on Twitter and from other pre-print papers.
“Throughout the summer I would see anecdotal reports on Twitter or in pre-prints of certain cardiovascular variables in COVID patients that doctors had found that might be significant and I would add them into the model,” Shade said. “It very much evolved as we learned about COVID. We didn’t know everything that would be important because it’s such a new disease.”
For example, the team did not anticipate that electrocardiogram data would play a critical role in the prediction of blood clotting. But once it was added, ECG data became one of the most accurate indicators for the condition, Trayanova said
The next step for the researchers is to develop the best method for setting up the technology in hospitals to aid with the care of COVID patients.
“The COVID-HEART predictor tool could help in the rapid triage of COVID-19 patients in the clinical setting especially when resources are limited,” said Allison Hays, associate professor of medicine in the Johns Hopkins University School of Medicine and the project’s clinical collaborator. “This may have implications for the treatment and closer monitoring of Covid-19 patients to help prevent these poor outcomes.”
In addition to Hays, Shade and Trayanova, other authors included: Ashish N. Doshi from the Division of Pediatric Cardiology; Eric Sung from biomedical engineering; Dan M. Popescu from the Department of Applied Mathematics and Statistics; and Anum S. Minhas, Nisha A. Gilotra and Konstantinos N. Aronis from the School of Medicine’s Division of Cardiology
The research is supported by grants to Trayanova’s lab from the National Science Foundation’s RAPID initiative (2029603) and the Lowenstein Foundation. In addition, Shade was supported by her National Science Foundation Graduate Research Fellowship (DGE-1746891).