Data-driven longitudinal characterization of neonatal health and morbidity

Although prematurity is the single largest cause of death in children under 5, the current definition of prematurity — based on gestational age alone — lacks the precision needed for guiding care decisions. This study proposes a longitudinal risk assessment for adverse neonatal outcomes, using a deep learning model over electronic health records to predict a wide range of outcomes across a window starting shortly before conception and ending months after birth.

By linking the EHRs of the Lucile Packard Children's Hospital and the Stanford Healthcare Adult Hospital, the team built a cohort of 22,104 mother–newborn dyads delivered between 2014 and 2018, training a multi-input, multitask deep learning model — featuring a long short-term memory network — to predict 24 different neonatal outcomes. An additional cohort of 10,250 dyads delivered from 2019 to September 2020 validated the model.