Abstract
BACKGROUND: Accurate risk stratification is essential for guiding hospitalization decisions in COVID-19. We evaluated the performance of CURB-65, CRB-65, NEWS2, qSOFA, and the 4C Mortality Score in predicting 30-day mortality among patients presenting to the emergency department with COVID-19. METHODS: We conducted an external validation study using an ambispective cohort of COVID-19 patients who presented to emergency departments of seven high-complexity hospitals in Colombia between March 2020 and September 2021. We assessed discrimination using the area under the receiver operating characteristic curve (AUC) and calibration using the GiViTI belt, the observed-to-expected (O/E) ratio, and the calibration intercept and slope. Decision curve analysis and net benefit were used to evaluate clinical utility. The 4C model underwent logistic recalibration. RESULTS: Among 7,973 patients included, 30-day mortality was 11.3%. The 4C model showed the highest discrimination (AUC 0.71, 95% CI 0.70-0.73) and clinical utility, but poor calibration. NEWS2, CURB-65, CRB-65, and qSOFA performed poorly across all performance metrics. After recalibration, the 4C model achieved an O/E ratio of 1 and showed a modest improvement in discrimination. Decision curve analysis confirmed its utility for guiding hospitalization decisions at a ≥4% mortality risk threshold. CONCLUSION: The 4C Mortality Score outperformed other models in predicting COVID-19 mortality. Its use in emergency settings alongside clinical judgment can enhance risk stratification, guide hospitalization decisions, and optimize resource allocation. Recalibration and decision analysis are essential for its clinical applicability. Further validation with contemporary data is essential to ensure its transportability across epidemiological settings.