Risk Factors and Prediction Models for Less Invasive Surfactant Administration Failure in Preterm Infants: A Retrospective Cohort Study in a Low-and-Middle Income Country.

Abstract

Objective: This study aimed to identify risk factors associated with less invasive surfactant administration (LISA) failure in preterm infants with respiratory distress syndrome (RDS) in a low-and-middle income country (LMIC) and develop a prediction model to estimate the risk of LISA failure.

Methods: This retrospective cohort study included 600 preterm infants who received LISA at a tertiary care neonatal unit in eastern India from January 2020 to December 2024.

Results: LISA failure, defined as the need for intubation and mechanical ventilation within 72 h of the procedure, ranged from 30% to 40%. The most important risk factors for LISA failure identified by random forest analysis were higher oxygen saturation index (OSI), higher cord base deficit, lower birth weight, lower admission temperature, and incomplete course of antenatal corticosteroids (ANS). Various prediction models were developed and validated on the testing set, with random forest and support vector machine using radial kernel demonstrating the highest accuracy (92.42% and 90.91%, respectively). In terms of sensitivity, Lasso penalized logistic regression was the best performing model followed by the random forest (96.55% and 93.88% respectively). The incidence of bronchopulmonary dysplasia (BPD), retinopathy of prematurity (ROP) requiring treatment, mortality, median duration of respiratory support, and time to discharge were significantly higher in the LISA failure group compared to the successful LISA group.

Conclusion: This study highlights the need for prediction models to estimate the risk of LISA failure in individual patients, which may inform decisions regarding alternative methods of surfactant administration in patients at high risk of LISA failure.