Abstract
AIM: To test National Early Warning Score 2 (NEWS2) versus a single-parameter system to identify critically ill general medical patients in the emergency department (ED), by 1) testing NEWS2s prediction of and association with primary outcome 'mortality' (hospital or 30 day) and secondary outcomes 'intensive care unit (ICU) admission' and 'critical care in ED' and 2) comparing this for different NEWS2 cut-offs and the single-parameter system in use. METHODS: Register-data on adult triage 1 and 2 patients with complete NEWS2 from 2015 and 2016 were retrieved. Prediction was assessed using area under the receiver-operating characteristic curve. Associations were analyzed using multiple logistic regression. RESULTS: 1586 patients were included. NEWS2 showed poor prediction of 'mortality' (AUC 0.686, CI 0.633-0.739) and adequate prediction of 'ICU admission' (AUC 0.716, CI 0.690-0.742) and 'critical care in ED' (AUC 0.756, CI 0.732-0.780). It was strongly associated with all outcomes (all p<0.001). All NEWS2 cut-offs and the single-parameter system showed poor prediction of all outcomes (all AUCs <0.7). The single-parameter system had the strongest association with 'mortality' (OR 1.688, CI 1.052-2.708, p<0.05) and 'critical care in ED' (OR 3.267, CI 2.490-4.286, p<0.001). NEWS2 > 4 had the strongest association with 'ICU admission' (OR 2.339, CI 1.742-3.141, p<0.001). CONCLUSION: For identification in order to trigger a response in the ED, outcomes closest in time seem most clinically relevant. As such, the single-parameter system had acceptable performance. NEWS2 > 4 should be considered as an additional trigger due to its association with ICU admission.