Distantly Supervised Phenotyping is a phenotype classification method that leverages noisy or approximate labels derived from external knowledge sources, such as electronic health record metadata or structured ontologies, to train predictive models in the absence of manually annotated gold-standard phenotypic labels.