Table 1 Sample of the resulting data structure. The upper table represents a diagnosis table that has a primary unique key (DiagnosisId), a patient and case key, an ICD code and a description of the diagnosis. The diagnosis “cluster headache for two weeks” is annotated with “D0009F4 Bing-Horton syndrome Z000002 two GA000F8 week” and subsequently stored in the annotation table (lower table). The semantic distance is “0” here because the concepts directly represent the narrative description of the diagnosis. Additionally, the parents of all concepts found are stored in the annotation table with the same diagnosis id. Therefore, “cephalea” is a parent of the 4th degree of “cluster headache”. The parent concepts are retrieved with a function call from the terminology server that returns the taxonomy of a given concept. The tables are linked with the relation “DiagnosisId ➔ Id”

From: Integrating terminologies into standard SQL: a new approach for research on routine data