Breaths Between the Beeps: Predicting Ventilatory Collapse in Pediatric ICU Admissions Using Integrated Early Warning Systems
Dimitris Alexiou¹, Ioanna Kyriazis², Jing Chen³, Wei Zhang⁴
Keywords:
Pediatric ICU, Early Warning System, Respiratory Failure, Predictive Monitoring, Mechanical VentilationAbstract
Background:
Mechanical ventilation remains a cornerstone of pediatric intensive care, yet early prediction of impending ventilatory failure is often delayed until overt clinical signs appear. Anticipating collapse through real-time data could revolutionise pediatric ICU triage.
Objective:
To develop and evaluate an integrated early warning system (i-EWS) combining continuous physiologic monitoring and laboratory markers to predict imminent respiratory failure in pediatric ICU admissions.
Methods:
A multicenter prospective validation study was conducted in four Chinese pediatric ICUs between April 2022 and February 2024. A total of 614 children aged 1 month to 17 years were enrolled upon admission. The i-EWS integrated heart rate variability, oxygen saturation trends, respiratory rate fluctuations, and lactate levels. Respiratory failure was defined as the need for invasive ventilation within 24 hours of ICU admission.
Results:
Respiratory failure occurred in 137 patients (22.3%). The i-EWS demonstrated an AUC of 0.93 (95% CI: 0.90–0.96), with 87.6% sensitivity and 82.1% specificity for predicting ventilatory collapse. The most predictive variables were a decreasing SpO₂ slope over 4 hours, a rising lactate level>2.5 mmol/L, and heart rate decoupling from respiratory rate (HR-RR disconnect). Compared to standard clinical judgment, the i-EWS predicted deterioration a median of 3.2 hours earlier (p < 0.001). No significant increase in false positives was observed.
Conclusion:
An integrated early warning system utilising physiologic and biochemical parameters can accurately anticipate respiratory failure in critically ill children. Its implementation may enable earlier intervention, reduce intubation-related complications, and improve ICU resource allocation. Future work should explore real-time integration into ICU dashboards.
Downloads
