OBJECTIVES: To develop a model to predict which newborns >/=34 weeks gestation with respiratory distress will die or will require prolonged (>3 days) assisted ventilation. METHODS: Retrospective cohort study using data from Northern California newborns >/=34 weeks gestation who presented with respiratory distress. We split the cohort into derivation and validation datasets. Bivariate and multivariate data analyses were performed on the derivation dataset. After developing a simple score on the derivation dataset, we applied it to the original as well as to a second validation dataset from Massachusetts. RESULTS: Of 2276 babies who met our initial eligibility criteria, 203 (9.3%) had the primary study outcome (assisted ventilation >3 days or death). A simple score based on gestational age, the lowest PaO 2 /FIO 2 , a variable combining lowest pH and highest PaCO 2 , and the lowest mean arterial blood pressure had excellent performance, with a c-statistic of 0.85 in the derivation dataset, 0.80 in the validation dataset, and 0.80 in the secondary validation dataset. CONCLUSIONS: A simple objective score based on routinely collected physiologic predictors can predict respiratory outcomes in infants >/=34 weeks gestation with respiratory distress.