WIP 71 - Lung Ultrasound Score for the Prediction of Surfactant Administration in Preterm Infants Diagnosed with Respiratory Failure

Poster Number: WIP 71
Publication Number: WIP 71.811

Background: Respiratory distress syndrome (RDS) is a common cause of respiratory failure in preterm infants admitted to the neonatal intensive care unit (NICU). Guidelines on when to administer rescue exogenous surfactant continues to vary between NICU’s. Preliminary evidence supports bedside lung ultrasound scores (LUS) as a predictive model to identify which infants benefit from exogenous surfactant earlier than current guidelines.

Objective: To determine if bedside LUS can predict the need for exogenous surfactant administration in premature infants with respiratory failure. A secondary objective is to find at what LUS cutoff provides the best sensitivity and specificity for surfactant administration.

Design/Methods: This is a prospective diagnostic accuracy study. We are including infants born < 37 weeks gestational age (GA) admitted to the Connecticut Children’s NICU on continuous positive airway pressure (CPAP). We are excluding infants with lung, cardiac, facial, or airway malformations, infants with APGAR scores ≤ 5 at five minutes of life, infants intubated, infants who already received surfactant, or if the bedside ultrasound was not performed by six hours of life. The bedside lung ultrasounds are performed using a LOGIQ E10 ultrasound machine with a L6-24 linear probe. Scores are calculated by the primary author and a radiologist (considered gold standard) according to the original Brat et al 2015 methods. Data including demographics, bedside vitals and laboratory values are also collected. A logistic regression model will be used to evaluate the relationship between LUS and surfactant administration (yes/no) and predict the probability that surfactant administration is needed based on the value of the LUS obtained. In addition, true positive rates (sensitivity) and true negative rates (specificity) and other classification errors for the fitted logistic regression model will be calculated in order to produce ROC curves and the area under the curve (AUROC).