Predictive power of early-warning scores used in hospital emergency departments: A systematic review and meta-analysis

Abstract

Objective. To assess the predictive power of scores used in hospital emergency departments (EDs) to give early warning of risk for mortality and hospital ward or intensive care unit (ICU) admission. Methods. Systematic review and meta-analysis. We searched MEDLINE, Embase, the Web of Science, and the Cochrane Library. Observational studies and clinical trials published between January 1, 1950, and June 12, 2020 that used early-warning scores in hospital EDs were included. The main outcomes were mortality (at 24, 48, and more than 72 hours), hospital admission, and ICU admission. Results. Nine studies entered into the systematic review; 4 of them, with 165 580 patients, were included in the meta-analysis. The studies were heterogeneous with respect to the scores used. The one used most often was the National Early Warning Score (NEWS). The meta-analysis of studies using the NEWS scale showed that it had good predictive power for mortality: the area under the curve (AUC) of the receiver operating characteristic was 0.88 (95% CI, 0.87–0.89; P < .001, I2 = 0%) at 24 hours and 0.86 (0.84–0.88; P < .001; I2 = 49.3%) at 48 hours. The AUC for in-hospital mortality was 0.77 (95% CI, 0.74–0.80; P < .001; I2 = 96.2%). The NEWS score had adequate power for predicting risk of hospital ward and ICU admission. Conclusion. Early warning scores used in hospital EDs are able to predict risk of early and in-hospital mortality.