Background: Early risk stratification scores utilizing data available at the bedside have been used regularly to identify patients at greatest risk of poor outcomes from rapidly evolving conditions like sepsis. Little is known about how point-of-care scores developed for sepsis perform at predicting outcomes in large, heterogeneous cohorts of patients with COVID-19. Our objective was to compare how well point-of-care severity scores, including the Quick Sequential Organ Failure Assessment score (qSOFA), the Shock Index, the National Early Warning Score 2 (NEWS2), and the COVID-specific Quick COVID-19 Severity Index (qCSI) calculated within 24 hours of emergency department (ED) presentation, predicted in-hospital mortality, intensive care unit (ICU) admission, receipt of mechanical ventilation, and receipt of vasopressors during the hospital stay among patients with incident COVID-19.

Methods: Data were collected for adult patients hospitalized with laboratory-confirmed SARS-CoV-2 infection, representing COVID-19 disease, who presented to an ED within a large national system of 186 hospitals until February 11, 2021. Patients who had their first positive COVID-19 test 7 days after hospital admission were excluded. Data on demographics, laboratory studies, vital signs, critical care management, and disposition were collected. Patients were considered at increased risk of poor outcomes based on evidence-based cutoffs if they had a qSOFA score ≥2, a Shock Index value of >0.7, a NEWS2 score ≥5, or a qCSI score ≥4, based on information gathered in the first 24 hours of hospitalization. We calculated test characteristics for each score to predict in-hospital mortality, ICU admission, receipt of mechanical ventilation, and receipt of vasopressors during the hospital stay.

Results: There were 94,895 ED patients admitted with COVID-19 during the study period. Based on information available in the first 24 hours, 3,418 (4%) were categorized as high risk of poor outcomes by qSOFA, 40,732 (43%) by the Shock Index, 45,036 (47%) by NEWS2, and 33,951 (36%) by qCSI. In the full cohort, there were 15,924 (17%) in-hospital deaths, 26,398 (28%) patients admitted to the ICU, 12,615 (13%) received mechanical ventilation, and 12,700 (13%) received vasopressors during the hospital stay. None of the scores predicted outcomes well based on Area Under the Receiver Operating Characteristic Curves (AUROC)(Table). NEWS2 consistently had the highest sensitivity (68.9-76.3%) and AUROC (0.6-0.63) while qSOFA consistently had the highest specificity (97.2-97.7%), lowest sensitivity (7.1-9.6%), and lowest AUROC (0.525-0.535), except for in-hospital mortality. All other scores, including the COVID-specific qCSI, had test characteristics between NEWS2 and qSOFA.

Conclusions: Neither commonly used bedside sepsis mortality prediction scores nor a COVID-specific bedside risk score reliably predicted hospital prognosis in patients with COVID-19 when using data collected at hospital admission. This study highlights the challenges of estimating outcomes in patients presenting with COVID-19. More accurate approaches to risk stratifying illness severity in COVID are needed.