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

Summary

Using day-1 clinical variables and endothelial biomarkers, TreeNet and CART models predicted day-3 persistent sepsis-associated acute respiratory dysfunction in critically ill children. The models were validated in a holdout set and an independent test cohort; day-3 SA ARD correlated with higher mortality, longer ventilation, and longer PICU stay.

Key Findings

• Derivation cohort (n=625) and independent test cohort (n=162) confirmed that day-3 SA ARD is associated with higher mortality, longer mechanical ventilation, and longer PICU stay.
• TreeNet and CART models using day-1 clinical variables and endothelial biomarkers achieved comparable predictive performance.
• Final CART model included presence of SA ARD on day 1 and leveraged endothelial biomarkers; performance held in a holdout and independent cohort.

Clinical Implications

Early measurement of endothelial biomarkers alongside clinical variables may help stratify pediatric sepsis patients at risk for persistent respiratory dysfunction and guide trial enrollment and resource prioritization.

Limitations

• Biomarker assays and thresholds may not be widely available or standardized across centers
• Single-center test cohort; prospective impact analysis on decision-making not yet performed

Future Directions

Prospective multicenter impact trials to test biomarker-guided management and evaluate calibration across diverse settings.