Performance of Pediatric Early Warning Score and physician gestalt in predicting adverse events among critically ill patients in an Emergency Department in Tanzania

Introduction Critically ill pediatric patients presenting to the emergency department (ED) are highly susceptible to adverse events if timely interventions are not initiated. To predict such events, ED physicians may use the Pediatric Early Warning Score (PEWS) and/or clinical gestalt. PEWS has been validated in predicting deterioration in both high and low-middle income countries, but it has not been compared to physician gestalt in either setting. Methods This was a prospective cohort study of children aged 28 days to 14 years who presented to the ED of a tertiary teaching hospital in Tanzania from 1st May to 30th June 2024 and who were triaged as “emergency” (highest acuity). Physicians’ estimates of the probability (0–100%) of an adverse event (admission to the high dependent unit, Intensive care unit, cardiac arrest or death) occurring during the patient’s ED stay were recorded for each patient on arrival. PEWS was calculated by the researchers from the child’s vital signs, and not disclosed to the physician. The discriminatory ability of PEWS and physician gestalt for the prediction of an adverse event in the ED was compared using DeLong’s test. Area under the curve (AUC) sensitivity, specificity, positive (PPV) and negative predictive (NPV) values were calculated, using a cut-off of ≥3 for PEWS and ≥70% for physician gestalt. Results Of the 250 pediatric patients enrolled, 48 had an adverse event. PEWS (AUC = 0.869, 95% CI 0.809 to 0.930), at a cut-off of ≥3 achieved a sensitivity of 75.0% and a specificity of 80.7%, PPV of 48.0%, and NPV of 93.1%. Physician gestalt (AUC = 0.964; 95% CI 0.936 to 0.992), at a cut-off of ≥70% achieved a sensitivity of 77.1% and a specificity of 95.5%, PPV of 80.4%, and NPV of 94.6%. The AUC of physician gestalt was significantly higher than that of PEWS ( p = 0.005). Discussion Physician gestalt demonstrated superior performance to PEWS in this setting, supporting its use as a risk stratification tool in low-resource emergency settings. However, as this study was performed at a tertiary referral and training hospital, further studies across diverse emergency care settings are needed to determine the generalizability of these findings.