Abstract
Nowadays, most activities on controlled medical vocabularies focus on the provision
of a sufficient atomic-level granularity for representing clinical data. Amongst others,
clinical vocabularies should be concept oriented, compositional and should also reject
“Not Elsewhere Classified” [1]. We strongly share the opinion that there is a need
to deal with serious deficits of existing manually created vocabularies and with new
demands for computer-based advanced processing and exchange of medical language data.
However, we do not share the opinion that methodological requirements like observational
and structural comparability needed for sound statistics should not be included in
desiderata of controlled medical vocabularies. Statistical-oriented classifications
are not developed for representing detailed clinical data but for providing purposedependent
classes where cases of interest are assigned uniquely. Either statistical classifications
are not included into the set of controlled medical vocabularies in the sense of Cimino,
or his desiderata are misleading. We argue that statistical classifications should
be linked to (formal) concept systems, but again this linkage does not change their
different natures. With this article we continue the “classification versus nomenclature”
controversy referring to Cote [2].
Keywords
Medical Informatics - Controlled Vocabulary - Documentation - Data Interpretation/Statistical
