A customized, institutional early warning score (EWS) for predicting decompensation, defined as transfer to ICU or death, with high specificity and positive predictive value in coronavirus disease 2019 (COVID-19) patients was presented at IDWeek, held virtually from October 21 to 25, 2020.

Investigators from Duke University created their EWS algorithm and applied it to 122 COVID-19 patients at a single medical center. Patients aged 18 years and older with confirmed COVID-19 admitted to a medical or surgical floor were included, while those admitted directly to ICU were excluded. The customized, institutional Duke EWS was compared with a national early warning score (NEWS).

Variables used in Duke’s EWS included demographics such as age, sex, race, and time of admission; vitals such as respiratory rate, blood pressure, level of consciousness, and supplemental oxygen; comorbidities including chronic kidney disease, chronic obstructive pulmonary disease, diabetes mellitus, HIV, malignancy, history of myocardial infarction, stroke, and transplant; and various laboratory values.

Continue Reading

Of the 122 patients, 28 had a decompensation event (event rate, 23%). A total of 8 patients died, 13 patients were transferred to ICU, and 6 patients were transferred to ICU and later died. Patients classified as high-risk for decompensation using Duke’s EWS had a 12-hour positive predictive value of 17.8% and 24-hour positive predictive value of 25.9%. Specificity for high-risk patients was 99.3% and 99.6% at 12- and 24-hour intervals, respectively; specificity for medium-risk patients was 85.0% and 85.4%, respectively. Patients classified as high-risk for decompensation using NEWS had a 12-hour positive predictive value of 7.1% and 24-hour positive predictive value of 12.3%.

According to investigators, the Duke EWS outperformed the national early warning score in COVID-19 patients and predicts deterioration well in this population. The strength of this score is its actionability, the investigators noted. Regularly pooling patient data every 12 hours gives an updated and real time score of the risk of deterioration while other studies only look at risk based on data at admission. Investigators believe that this study demonstrates how large health systems can leverage their data for use in clinical decisions tools.


Cavalier JS, Goldstein B, O’Brien CL, Zhao C, Bedoya A. A modified early warning score predicts decompensation in COVID-19 patients. Presented at: ID Week 2020; October 21-25, 2020. Poster 362.