Although extracorporeal membrane oxygenation (ECMO) is known to improve survival in neonates with respiratory failure, there has been a significant decrease in the use of ECMO in recent years. Alternative modalities such as high-frequency oscillatory ventilation (HFOV), inhaled nitric oxide (NO), and surfactant therapy are associated with this decline. The criteria for the initiation of ECMO, developed about 20 years ago, are likely no longer relevant. We examined the predictive significance of the oxygenation index (OI) as a patient entry criterion for ECMO use. We sought a critical OI level predicting death or chronic lung disease (CLD) with and without ECMO use. We also examined whether patients with certain OIs are more likely to have worse outcomes. One hundred and seventy-four term-newborn admissions between 1995 and 2000 requiring mechanical ventilation were enrolled in the study. Receiver operating curve analysis was performed to find a cutoff value of OI for ECMO initiation. Mortality rates and CLD probability were compared to the worst OIs. Our 6-year ECMO administration experience showed that an OI of 33.2 is a suitable cutoff value for ECMO initiation with high sensitivity and specificity as a predictive criterion. The critical OI value associated with the CLD risk when ECMO is not used is in the 40s. OI is a good predictor of CLD; the probability of CLD increases with higher OIs. Our data support the trend toward the use of new interventions over ECMO, especially for patients with OI scores of less than 33.2. Only when the probability of ventilator-associated lung injury becomes significant is it better to consider ECMO than conventional modalities.