Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predicting the time to vaginal delivery of an infant. In one aspect, a method comprises: obtaining patient data characterizing a patient, comprising automatically querying a database storing one or more electronic medical records of the patient; generating a model input to a delivery time machine learning model based on the patient data characterizing the patient; processing the model input using the delivery time machine learning model, in accordance with values of a set of model parameters of the delivery time machine learning model, to generate a prediction for the time to vaginal delivery of the infant; and generating a notification that indicates the prediction for the time to vaginal delivery of the infant.