External Validation of Empirically Derived Vital Signs in Children and Comparison to Other Vital Signs Classification Criteria

Various vital sign ranges for pediatric patients have differing utility in identifying children with serious illness or injury requiring immediate intervention. While commonly used ranges are derived from samples of healthy children, limited research has explored the utility of those derived from real-world encounters by emergency medical services (EMS). We first sought to externally validate pediatric vital sign ranges empirically derived from the prehospital setting. Second, we compared the proportion of children who received prehospital interventions using current common classification systems versus empirically derived vital sign ranges. We retrospectively reviewed pediatric (99th, >95th, >90th, th, th and st centiles to previously reported centiles derived from EMS encounters in 2019–2020. We compared the deviation of mean Z-score by age between data sources. We identified the proportion of encounters with extreme (defined as th or >90th centile) vital signs who received prehospital interventions for the empirically derived criteria to six other classification criteria. 510,414 encounters were included, of which 66.9% were for medical indications and 70.7% resulted in hospital transport. The study sample had similar proportions of encounters identified at studied cutoffs compared to the previously published derivation sample, with all differences in proportions ≤1.1% between samples. All mean Z-scores were within 0.2 standard deviations of those from the derivation sample for each vital sign. Using empirically derived criteria, 34.2% had at least one extreme vital sign, compared to 69.1% with Pediatric Advanced Life Support criteria. Empirically derived extreme vital signs identified a higher proportion of children requiring most prehospital interventions compared to other vital signs criteria. Previously published empirically derived centiles for pediatric prehospital vital signs were replicated in this large multi-agency dataset. Compared to commonly used vital sign ranges, empirically derived criteria identified a higher proportion of children who received key prehospital interventions. Future steps include evaluating the role of these criteria in predictive models for in-hospital outcomes.