522. The Simple and Novel SAS Score to Predict Mortality at Presentation in 2541 Hospitalized COVID-19 Patients

Abstract

BackgroundThe clinical spectrum of the novel corona virus disease 2019 (COVID-19) ranges from mild to severe disease and death. We aim to construct a simple and novel scoring model that will predict mortality events in hospitalized COVID-19 patients.MethodsWe established a retrospective cohort of 2541 patients admitted with COVID-19 from February 19, 2020 to April 28, 2020 to Henry Ford Health System, MI. Sociodemographic data, comorbidities, and clinical data were collected. Our novel SAS score was constructed using 3 easily available parameters, namely Sex, Age, and Oxygen Saturation at presentation (Table 1 and 2). Primary endpoint was mortality. Multivariate analysis with logistic regression was done and the model was assessed using receiver operating characteristic (ROC) with area under ROC (AUROC) to determine the optimal cutoff for sensitivity, specificity, and positive and negative predictive values. ResultsThe mean age of survivors was 61 compared to 75 years for non-survivors (standard deviation 16 vs 13.8, p< 0.0001), and 1298 (51.1%) were men. Multivariate analysis of the SAS score adjusted for modified SOFA [Sequential organ failure assessment] score (mSOFA) showed that age (odds ratio [OR] 2.4, 95% confidence interval {CI} 2.04–2.72, p< 0.0001) and oxygen saturation (OR 1.6, 95% CI 1.27–1.98) were the most significant predictors of mortality in the model. The SAS score had an AUROC of 0.78 (95% CI 0.77–0.81) (Figure 1). A cutoff score of 3 offered the most sensitivity for predicting mortality while maintaining a negative predictive value of 95% (Table 3). Comparison of AUROC shows that SAS score adjusted to mSOFA has better diagnostic information compared to either SAS score or mSOFA alone (Figure 2). ConclusionThe easy to use SAS score at time of presentation identified hospitalized COVID-19 patients at high risk for mortality. Application of the SAS score in the emergency department may help triage patients to inpatient versus outpatient care.DisclosuresMarcus Zervos, MD, Melinta Therapeutics (Grant/Research Support)