Derivation and Validation of a Clinical and Endothelial Biomarker Risk Model to Predict Persistent Pediatric Sepsis-Associated Acute Respiratory Dysfunction.

Summary

In prospectively enrolled children with septic shock, machine-learning models that incorporate day-1 endothelial biomarkers and clinical variables predicted persistent sepsis-associated acute respiratory dysfunction on day 3. Children with day-3 dysfunction had higher mortality, longer ventilation, and longer PICU stays; CART models validated in holdout and independent cohorts identified early predictors.

Key Findings

• Day-1 endothelial biomarkers plus clinical variables predicted day-3 persistent sepsis-associated acute respiratory dysfunction using TreeNet and CART.
• Children with day-3 respiratory dysfunction had higher mortality, longer mechanical ventilation, and longer PICU length of stay.
• Models validated in both holdout and an independent test cohort, demonstrating reproducibility.

Clinical Implications

Supports early identification of high-risk children with sepsis for closer monitoring, timely ventilatory strategies, and potential biomarker-guided therapies; facilitates enrichment strategies in interventional trials targeting pediatric respiratory failure.

Limitations

• Single-center external test cohort may limit generalizability
• No prospective implementation trial to assess impact on outcomes

Future Directions

Prospective, multi-center impact studies to test biomarker-guided care pathways; assess integration with ventilatory strategies and anti-endothelial therapies; calibration across varied PICU settings.