Early versus late identification of infants at high risk of developing moderate to severe bronchopulmonary dysplasia.

Abstract

We developed a simple method to identify neonates at high risk of bronchopulmonary dysplasia (BPD) and determined whether early (8 hours) and late (14 days) risk assessment is equally useful. A retrospective cohort design was utilized of subjects enrolled in multi-dose surfactant trials to develop each risk identification model. Prospective testing of the late 14-day model was done to determine accuracy. The primary outcome variable (moderate to severe BPD) was defined as the need for oxygen and mechanical ventilation beyond 28 days of life and significant chest X-ray changes. Variables were screened for inclusion in the models by univariate and multiple regression analysis of data available at 14 days or 8 hours of life, converted to yes-no variables by the use of receiver-operator curves; the final model was based on those variables that gave the highest sensitivity and specificity for identifying BPD risk. Thirty-eight out of 116 of the 14-day model subjects developed BPD. The 14-day model (F1O2 > or = 0.30 and ventilation index (defined as 10,000/peak pressure x rate x PCO2) < 0.510 (or < 0.800 if previously septic)] had a sensitivity of 82% and specificity of 89%. It accurately identified 83% of cases (51/61) during at 1-year prospective test. The positive predictive value was 81% and negative predictive value 88%. Forty-four of the 698 early 8-hour model subjects developed BPD. The 8-hour model [gestational age < 31 weeks, 5-minute Apgar < 9, ventilator rate > 23 breaths/min, and ventilation index < 0.895] had a sensitivity of 73%, specificity of 83%, negative predictive value of 98% but positive predictive value of only 22%. These observation indicated that clinical data can create an accurate and simple model to classify infants into high- or low-risk groups for BPD. Using such models very early in life (e.g., at 8 hours) may lead to a high number of false-positive identifications.