Development and Internal Validation of Neonatal Outcomes Prediction Score (NOPS) and Model for Neonates of 22-29 weeks' Gestation on Postnatal days 7, 14, and 28.

Abstract

To develop and internally validate models and severity scores at days 7, 14, and 28 to predict death, major morbidity, and bronchopulmonary dysplasia in 22-29 weeks' gestation neonates. This retrospective cohort study included infants admitted to Canadian neonatal intensive care units (2010-2023) and used 16 predefined baseline, intervention, and morbidity variables to develop prediction models and the Neonatal Outcomes Prediction Score (NOPS) at days 7, 14, and 28. Infants transferred to level 2 units before model age were excluded. We studied 26,719 infants of 22-29 weeks' gestation (median GA=27 weeks, median birthweight 980 g). Eleven variables for outcome of death, 8 for death/major morbidity, 7 for death/bronchopulmonary dysplasia, and 8 for death/moderate-severe BPD were associated with respective outcomes in three time-based models. The models were internally valid. The AUC for death were >0.85, death/major morbidity were >0.80, death/bronchopulmonary dysplasia were >0.78 and death/moderate-severe BPD were >0.76 at all timepoints. The NOPS derived from these models showed that a score >30 was significantly associated with all three outcomes, with ORs >8.3 at all timepoints. An internally valid and accurate models for neonatal outcome prediction in the first postnatal weeks were generated including static variables, exposed interventions, and accumulated morbidities. NOPS >30 was significantly associated with all 4 outcomes at all timepoints. The models require external validation.