ABSTRACT
Several risk scores were developed during the COVID-19 pandemic to identify patients at risk for critical illness as a basic step to personalizing medicine even in pandemic circumstances. However, the generalizability of these scores with regard to different populations, clinical settings, healthcare systems, and new epidemiological circumstances is unknown. The aim of our study was to compare the predictive validity of qSOFA, CRB65, NEWS, COVID-GRAM, and 4C-Mortality score. In a monocentric retrospective cohort, consecutively hospitalized adults with COVID-19 from February 2020 to June 2021 were included; risk scores at admission were calculated. The area under the receiver operating characteristic curve and the area under the precision-recall curve were compared using DeLong's method and a bootstrapping approach. A total of 347 patients were included; 23.6% were admitted to the ICU, and 9.2% died in a hospital. NEWS and 4C-Score performed best for the outcomes ICU admission and in-hospital mortality. The easy-to-use bedside score NEWS has proven to identify patients at risk for critical illness, whereas the more complex COVID-19-specific scores 4C and COVID-GRAM were not superior. Decreasing mortality and ICU-admission rates affected the discriminatory ability of all scores. A further evaluation of risk assessment is needed in view of new and rapidly changing epidemiological evolution.