A system including an inference engine and a query interface. The inference engine predicts a duration for at least one event of a perioperative workflow using one or more models. Each model includes at least one model instance. The query interface queries the at least one model instance, and the inference engine generates the predicted duration responsive to the query.