Long-Term Mechanical Ventilation in Neonates: A 10-Year Overview and Predictive Model

Area of Science:

• Neonatology
• Pediatric Critical Care
• Respiratory Medicine

Background:

• Neonatal prolonged mechanical ventilation (NPMV) requires significant resources, yet patient outcomes remain poorly understood.
• Understanding long-term respiratory, digestive, and neurological sequelae is crucial for this vulnerable population.

Purpose of the Study:

• To describe the 18-month corrected age outcomes of infants undergoing NPMV.
• To identify early predictors for identifying infants requiring ventilation for ≥125 days.

Main Methods:

• Retrospective cohort study of infants born 2004-2013 requiring ≥21 days of respiratory support.
• Utilized machine learning models (random forests, logistic regression, naive Bayes, XGBoost) for prediction.
• Employed Monte Carlo cross-validation for model assessment.

Main Results:

• 164 infants included; 29% mortality by 18 months corrected age.
• At 18 months, 59% were free of respiratory support, 45% exclusively orally fed.
• Random forests achieved moderate discrimination (AUC 0.65) for predicting ≥125 days ventilation, with PCO2, inspired O2, and gestational age being key predictors.
• Infants ventilated ≥125 days had poorer respiratory weaning, lower oral feeding rates, and higher neurological impairment.

Conclusions:

• NPMV is associated with significant mortality, neurological impairment, and feeding delays at 18 months corrected age.
• Most survivors can be weaned from respiratory support.
• Early identification of high-risk infants for prolonged ventilation is feasible using specific clinical parameters.