Bronchopulmonary dysplasia: risk prediction models for very-low- birth-weight infants

Area of Science:

• Neonatal Medicine
• Pediatric Pulmonology
• Clinical Epidemiology

Background:

• Bronchopulmonary dysplasia (BPD) is a significant complication in very-low-birth-weight infants (VLBWI).
• Accurate prediction of BPD and associated mortality is crucial for timely intervention and improved outcomes.
• Existing predictive models may lack sufficient accuracy or applicability across different postnatal time points.

Purpose of the Study:

• To develop and validate risk prediction models for moderate/severe BPD.
• To develop and validate risk prediction models for the combined outcome of BPD and/or death.
• To assess model performance at multiple postnatal time points: birth, 3, 7, and 14 days.

Main Methods:

• Multicenter study involving 16,407 VLBWI (500-1500g) from the Neocosur Network (2001-2015).
• BPD defined as oxygen dependency at 36 weeks postmenstrual age.
• Forward logistic regression for variable selection and ROC curve analysis for performance evaluation (AUC).

Main Results:

• 15.7% of infants developed BPD, and 37.3% experienced BPD and/or death.
• Area Under the Curve (AUC) values for BPD models ranged from 0.788 to 0.894.
• AUC values for BPD/death models ranged from 0.860 to 0.906, indicating high predictive power.
• Birth weight and gestational age were key predictors at birth; oxygen therapy duration and ventilation were critical later on.

Conclusions:

• High-performance predictive models for moderate/severe BPD and BPD/death were successfully developed.
• Models demonstrated strong predictive accuracy at four distinct postnatal ages.
• These models can aid in identifying VLBWI at risk for BPD and adverse outcomes.