Introduction: Early prediction of survival without bronchopulmonary dysplasia (BPD) at 36 weeks of postmenstrual age remains challenging for infants born extremely preterm. We aimed to provide a new predictive model including variables available only at or soon after birth based on the literature and existing models.
Methods: We conducted a systematic review to identify all variables considered to be significant predictors of BPD and survival at birth in extremely preterm infants. We then assessed the external validity of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Neonatal Research Network BPD estimator on the PREMILOC cohort, a recent French study with a large sample of extremely preterm infants and a vast number of variables at baseline. Finally, we attempted to improve this model by testing the added value of other early predictors reported in previous studies.
Results: Restricted to baseline predictors, the NICHD Neonatal Research Network BPD estimator confirmed its calibration and fair discrimination (area under the receiver operating characteristic [auROC] [95% CI] = 0.73 [0.68-0.77] when used with a published model and auROC [95% CI] = 0.77 [0.73-0.81] when fitted to the PREMILOC dataset). We were able to improve the discriminatory power by adding candidate variables at birth associated with BPD in previous studies. The modified best predicting model included gestational age at birth, birthweight, respiratory support at baseline, gender, center effect, and multiple pregnancy as baseline predictors. This model showed significantly better discrimination (auROC [95% CI] = 0.85 [0.82-0.88]) and better confirmed calibration (Hosmer-Lemeshow test, p = 0.45).
Conclusion: This new model, based on 6 early predictors, appears to improve the prediction soon after birth of BPD-free survival in extremely preterm infants.
Keywords: Bronchopulmonary dysplasia; Extremely preterm infants; Prediction model; Risk.
© 2021 S. Karger AG, Basel.