New tool can help doctors predict likelihood of severe COVID-19

A simple tool is now available that can help doctors predict which of their COVID-19 patients are more likely to need hospitalization.

The tool, known as a clinical risk score, successfully predicts which patients are at high risk of hospitalization or death based on the symptoms they present with and other key independent factors, including age and having conditions like chronic lung or heart disease.

The risk scores, which were developed to require either no labs or labs that could be run in under an hour, were first developed in 2021 by a team of researchers, led by University of Georgia professor Mark Ebell, using data from over 5,000 COVID-19 outpatients who visited clinics before March 2021.

The risk scores were prospectively validated this risk score using data from patients affected by the Alpha, Delta and Omicron variants.  It has now been tested with a new set of over 1,000 patients and found to be accurate at predicting COVID-19 deaths among hospitalized patients, as reported in a study appearing in the Annals of Family Medicine.

“The two simple risk scores, now successfully externally validated, offer clinicians a reliable way to quickly identify low-risk inpatients who could potentially be managed as outpatients in the event of a bed shortage,” write the authors.

The tool arrives at a time when hospitals and clinics are facing a triple threat of COVID, flu, and RSV patients. Making decisions quickly about who needs to be considered for hospitalization and who can safely stay home can help curb the burden on stressed health care systems.

This risk score is a gut-check for clinicians, says Ebell, a physician and epidemiologist in UGA’s College of Public Health.

“You have to make a decision as a primary care doctor, as an urgent care doctor, whether you send a patient home and say, you’re going to be fine, or whether you send them home but follow up in a day or two, versus, I’m going to send you over the ER where they can do some further evaluation,” said Ebell. “It helps support that triage decision.”

The researchers aimed to develop rules that used no labs – only age, medical history, symptoms and vital signs. Each has only 3 to 5 predictors, which makes them well-suited for telehealth.

Both are available online: and

– Lauren Baggett

Posted on December 19, 2022.