Objective: To determine whether the need for respiratory support can be predicted by oxygen requirement within the first 72 h in term and near-term infants.
Methods: To mimic the population of infants that would often be delivered outside a tertiary centre we studied a retrospective cohort of infants > or = 32 weeks requiring oxygen, divided into three groups: cot oxygen only, nasal continuous positive airway pressure (NCPAP) only, or intermittent positive pressure ventilation (IPPV). We recorded each infant's peak fraction of inspired oxygen (FiO2)--i.e. FiO2 in the first 72 h in the cot oxygen only group or maximum FiO2 prior to commencing the highest level of respiratory support. The peak FiO2 was used as a diagnostic test to predict any respiratory support or IPPV--sensitivity and specificity were calculated and receiver operating characteristic (ROC) curves plotted (FiO2 0.21-1.00) to identify the best balance point for prediction.
Results: The cohort included 592 infants: 516 cot oxygen only, 46 NCPAP only and 30 IPPV. The proportion ventilated increased with increasing peak FiO2--above 0.45 the proportion of infants ventilated exceeded 50%. To predict any respiratory support, the cut-point balancing sensitivity and specificity was a FiO2 > or = 0.35-58/136 required respiratory support (sensitivity = 0.76, specificity = 0.85, positive predictive value (PPV) = 43%, negative predictive value (NPV) = 96%). To predict IPPV the cut-point was a FiO2 > or = 0.5-28/47 treated with IPPV (sensitivity = 0.93, specificity = 0.97, PPV = 60%, NPV = 100%).
Conclusion: The need for respiratory support can be predicted by oxygen requirement within the first 72 h in term and near-term infants with reasonable sensitivity and excellent specificity.