Validation of neonatal early warning score in medical and surgically critically ill newborns in tertiary care center

Abstract

Neonatal Early Warning Scores (NEWS) are widely used to identify children at risk of clinical deterioration, but their performance may vary across clinical settings and patient populations. In this retrospective cohort from a tertiary university hospital, we evaluated the discriminative ability of an institutional NEWS at admission to predict need for intensive care treatment, intubation, and tiered outcomes (ward care, ongoing intensive care requirement, death) in medical and surgical patients. ROC ( Receiver operating characteristic) analyses were performed for the full cohort and for medical and surgical subgroups. AUCs ( Area under the curve) were calculated, and optimal cut-offs determined via the Youden index. Subgroup comparisons used a bootstrap-based DeLong test and Mann–Whitney U test. The study included 229 patients (158 medical, 71 surgical). In the full cohort, NEWS showed acceptable to good discrimination for intubation (AUC 0.75) and for higher-acuity clinical outcomes (intensive treatment vs. death, AUC 0.80; ward care vs. death, AUC 0.85) at an optimal cut-off near 8 points. In medical patients, AUCs ranged 0.72–0.77 depending on the outcome. In surgical patients, AUCs for intubation and mortality-related contrasts ranged from 0.80 to 0.94. Score distributions differed significantly between cohorts ( p = 0.027). Admission NEWS demonstrated acceptable to excellent performance for predicting intubation and higher-acuity clinical outcomes, especially mortality in surgical patients. Its utility for predicting hospitalisation or non-fatal escalation was limited. NEWS should be integrated into structured early-warning pathways and calibrated to local case mix and workflows, with cohort-specific thresholds.