Aims We have previously developed evidence-based centiles for normal heart rate and respiratory rate in children. The predictive value of these centiles has been assessed using independent data collected in an emergency care setting.

Method Predictive value was assessed using data previously gathered from children attending a UK emergency department with medical illness. The ability of various centiles to predict admission for more than 24 hours; and admission for more than 24 hours, administration of oxygen, or the need for IV interventions in respiratory illness were assessed. The performance of evidence-based centiles was compared to that of the existing APLS clinical thresholds.

Results The 3rd and 97th evidence-based centiles produce sensitivities and specificities that are similar to those of the existing APLS thresholds for heart rate. For respiratory rate, similar results to the existing APLS thresholds are observed using the 25th and 75th evidence-based centiles. The c-statistic (estimated area under the ROC curve) shows that individual vital signs have moderate predictive ability for admission and serious respiratory illness in this cohort (ranging from 0.585 for prediction of admission using heart rate alone, to 0.695 for prediction of serious respiratory illness using both heart rate and respiratory rate).

Conclusion Evidence-based centiles and APLS thresholds for heart rate and respiratory rate have similar predictive ability. The choice of centiles allow appropriate trade-offs between sensitivity and specificity to be made in different clinical situations.