Objectives: The Pediatric Early Warning Score (PEWS) systems were developed to provide a reproducible assessment of a child's clinical status while hospitalized. Most studies investigating the PEWS evaluate its usefulness in the inpatient setting. Limited studies evaluate the effectiveness and integration of PEWS in the pediatric emergency department (ED). The goal of this study was to explore the test characteristics of an ED-assigned PEWS score for intensive care unit (ICU) admission or clinical deterioration in admitted patients.
Methods: This was a prospective 12-month observational study of patients, aged 0 to 21 years, admitted from the ED of an urban, tertiary care children's hospital. ED nurses were instructed in PEWS assignment and electronic medical record (EMR) documentation. Interrater reliability between nurses was evaluated. PEWS scores were measured at initial assessment (P0) and time of admission (P1). Patients were stratified into outcome groups: those admitted to the ICU either from the ED or as transfers from the floor and those admitted to the floor only. Clinical deterioration was defined as transfer to the ICU within 6 hours or within 6 to 24 hours of admission. PEWS scores and receiver operating characteristic (ROC) curves were compared for patients admitted to the floor, ICU, and with clinical deterioration.
Results: The authors evaluated 12,306 consecutively admitted patients, with 99% having a PEWS documented in the EMR. Interrater reliability was excellent (intraclass coefficient = 0.91). A total of 1,300 (10.6%) patients were admitted to the ICU and 11,066 (89.4%) were admitted to the floor. PEWS scores were higher for patients in the ICU group (P0 = 2.8, SD ± 2.4; P1 = 3.2, SD ± 2.4; p < 0.0001) versus floor patients (P0 = 0.7, SD ± 1.2; P1 = 0.5, SD ± 0.9; p < 0.0001). To predict the need for ICU admission, the optimal cutoff points on the ROC are P0 = 1 and P1 = 2, with areas under the ROC curve (AUCs) of 0.79 and 0.86, respectively. The likelihood ratios (LRs) for these optimal cutoff points were as follows: P0 +LR = 2.5 (95% confidence interval [CI] = 2.4 to 2.6, p < 0.05), -LR = 0.32 (95% CI = 0.28 to 0.36, p < 0.05); and P1 +LR = 6.2 (95% CI = 5.8 to 6.6, p < 0.05), -LR = 0.32 (95% CI = 0.29 to 0.35, p < 0.05). For every unit increase in P0 and P1 , the odds of admission to the ICU were 1.9 times greater (95% CI = 1.8 to 1.9, p < 0.0001) and 2.9 times greater (95% CI = 2.7 to 3.1, p < 0.0001) than to the floor. There were 89 patients in the clinical deterioration group, with 36 (0.3%) patients transferred to the ICU within 6 hours of admission and 53 (0.4%) patients transferred within 6 to 24 hours. In this group, an elevated P0 and P1 were statistically associated with an increased risk of transfer with optimal cutoff points similar to above; however, there were poorer AUCs and test characteristics.
Conclusions: A PEWS system was implemented in this pediatric ED with excellent data capture and nurse interrater reliability. The study found that an elevated PEWS is associated with need for ICU admission directly from the ED and as a transfer, but lacks the necessary test characteristics to be used independently in the ED environment.
© 2014 by the Society for Academic Emergency Medicine.