Objective: To develop a prediction model to identify infants admitted on continuous positive airway pressure (CPAP) requiring intubation within seventy-two hours of life (HOL). Study design: Infants born ≤29 weeks’ gestational age between 2013 and April 2018 were randomly assigned to either a modeling cohort (MC) or a validation cohort (VC) in a 2:1 ratio. Variables available within two HOL were compared between the CPAP failure group (CFG) and the CPAP success group (CSG). Results: Of the 189 infants in the MC, 50% failed CPAP. Compared to CSG, infants in the CFG had lower antenatal steroid exposure, birth weight, higher radiographic severe respiratory distress syndrome (RDS) and fraction of inspired oxygen (FiO2). A forward stepwise logistic regression modeling in both MC and VC showed that FiO2>0.3 and radiographic severe RDS predicted CPAP failure. Conclusion: FiO2>0.3 within two HOL and radiographic severe RDS predicts CPAP failure in preterm infants.