Former Department of Orthopaedic Surgery Research Fellow Eli Levitt, M.S., worked alongside leaders of the field to conduct research on predicting SARS-CoV-2 infection (COVID-19) among US adults, using machine learning.

Levitt was listed as an author on the research publication “Clinical Characterization and Prediction of Clinical Severity of SARS-CoV-2 Infection Among US Adults Using Data From the US National COVID Cohort Collaborative,” which was published in the JAMA Network Open.

Authors aimed to identify what risk factors are associated with COVID-19 severity and severity trajectory over time, and if machine learning models could predict clinical severity. 174,568 adults with SARS-CoV-2 were included in the study.

As if a patient was being admitted to the hospital, the machine learning models were given select inpatient information about each COVID-19 patient and then “asked” to predict the severity of the case. Intriguingly, even on day one of illness, machine-learning models were able to accurately predict COVID-19 clinical severity of these patients, from being hospitalized, severely ill, or eventually deceased.

Using these models, the cohort study found that COVID-19 mortality decreased over time during 2020 and that patient demographic characteristics and comorbidities were associated with higher clinical severity. Authors concluded that the model may be a clinically useful, machine learning–based predictor of SARS-CoV-2 severity.

Levitt was a partner in conducting the research across the database of patients, as well as drafting the manuscript of the study. He worked with the UAB Center for Clinical and Translational Science (CCTS).

“It was a wonderful opportunity to further my research experience, while studying an extremely impactful topic,” said Levitt. “I’m grateful for the resources of the CCTS and the department, and I look forward to the clinical impact that this study may have.”

To read the full study and its methods, click here.