The field of informatics has long recognized the potential to make novel and informative inferences from biomedical data, what today we call the Learning Health System. However, at its root, such inferences are not possible absent that underlying data being comparable and consistent.
The core premise of medical concept representation is the unambiguous representation of a concept, such as diagnoses, procedures, medications, laboratory tests, or even demographics, in a computable context. As such, the discipline of classification, controlled terminology, and formal ontologies applied to biomedical concepts are subsumed under medical concept representation.
Feature extraction from natural language ultimately depends upon such computable formalisms to render computable features. This field has been a core competency of many Johns Hopkins faculty members for decades.