Late preterm (34-36 weeks gestational age) infants account for 7% of the 3.76 million live births in the United States annually, or over 263,000 infants each year. Compared to term infants, late preterm infants are at increased risk of morbidity from outcomes such as hypoglycemia, temperature instability and hyperbilirubinemia, and often require medical intervention in a neonatal intensive care unit (NICU). Thus, while the vast majority of infants born at term stay with their mothers in a well infant (level I) nursery during the birth hospitalization, many late preterm infants are instead hospitalized in the NICU where they may be separated from their mothers. However, significant variation exists amongst hospitals for NICU admission rates and clinical thresholds for admission in late preterm infants that is not explained by clinical illness. Preliminary data obtained by the PI suggests that institutional criteria for requiring automatic NICU admission in late preterm infants can vary from 34-37 weeks gestational age and 1500-2500 grams birth weight. This represents late preterm infants of varying maturity and size, and likely does not precisely capture infants who are at highest risk of needing NICU level interventions. The goal of this proposal is to identify optimal NICU admission criteria for late preterm infants. A large retrospective cohort of late preterm infants born at a single institution will be assembled, collecting data on admission locations, and occurrence and management of late preterm morbidities. With this, Aim 1 will be addressed: identify the frequency of neonatal morbidities amongst infants born at 34-36 weeks’ gestation, and the frequency of these morbidities requiring medical intervention. Literature on the frequency of morbidities in late preterm infants is limited, and none currently exists delineating the proportion of these morbidities that require clinical intervention. Subsequently, in Aim 2: a prediction model will be developed for which late preterm infants are most likely to benefit from automatic admission to a NICU at the time of birth. The cohort generated in Aim 1 will be utilized to compare clinical parameters of infants who required at least one NICU level intervention to those that did not require any. Training and test data sets will be established. Using cross-validation techniques within the training set, an optimal cut-point for a score derived from the predictive model will be chosen to drive clinical decision-making based on the sensitivity and specificity of the decision rule. The strategy will be evaluated on a test set. The obtained prediction model will be a resource towards informing optimal NICU admission criteria for late preterm infants. The PI will train in study design methodology, data analysis, modeling, and grant writing during this fellowship that will advance her career path towards an independent physician scientist focused on identifying high value care practices that safely promote an...