Cimino's desiderata


In 1989, JJ Cimino of New York described seven desiderata for the design of a controlled healthcare vocabulary (the Medical Entities Dictionary - MED - from the Columbia Presbyterian Medical Centre).

Cimino JJ, Hripcsak G, Johnson SB, Clayton PD. Designing an Introspective, Multipurpose, Controlled Medical Vocabulary. In: Kingsland LC (Ed). Proceedings of the Thirteenth Annual Symposium on Computer Applications in Medical Care. New York: IEEE Computer Society Press, 1989: 513-8.


· Domain completeness

The ability to accommodate appropriately all necessary concepts. Schemes should not limit depth or breadth of hierarchies. Compositional approaches allow complex concepts to be represented.

· Unambiguous

Terms should clearly represent only a single concept (see semiotic triangle). Synonyms should be pure.

· Non-redundancy

There must be only one way of representing a concept in the vocabulary, or equivalences between alternative representations should be detectable.

· Synonymy

More than one term (synonym) may describe the same concept.

· Multiple classification

Entities from the vocabulary should be placed in more than one hierarchy location if appropriate. For example, Carcinoma of the colon is both a Malignant disease and a Large intestinal disease.

· Consistency of views

Cimino identified the problem of multiple classification being inconsistent or incomplete and that qualifiers or modifiers might vary between different parts of the hierarchy.

· Explicit relationships

The nature of relationships between concepts in the vocabulary structure should be explicit and usually sub-class (see IS-A).


In 1998, based on experience during the intervening period, Cimino revised the list and added some further items to produce a list of twelve "desiderata for the 21st Century".

Cimino JJ. Desiderata for Controlled Medical Vocabularies in the Twenty-First Century. Methods Inform Med 1998; 37: 394-403.

(see also Jim's presentation to IMIA WG6 in Rome, in 2005)


· Content

To most users"What can be said" is more important than "how it can be said". Omissions are readily noticed and timely, formal and explicit methods for plugging gaps are required.

· Concept orientation

The unit of symbolic processing is the concept and each concept in the vocabulary should have a single, coherent meaning.

· Concept permanence

A concept's meaning cannot change and it cannot be deleted from the vocabulary.

· Meaningless concept identifier

Concepts typically have unique identifiers (codes) and these should be non-hierarchical (see code-dependance) to allow for later relocation and for multiple classification.

· Polyhierarchy

Multiple classification (see above).

· Formal definitions

Semantic definitions of concepts, for example, Streptococcal tonsillitis=Infection of tonsil caused by streptococcus.

· No residual categories

Traditional classifications have rubrics that include NOS, NEC, Unspecified, Other whose meaning may change over time as new concepts are added to the vocabulary. These are not appropriate for recording data in an electronic health record.

· Multiple granularities

Different users require different levels of expressivity. A general (family) practitioner might use myocardial infarction whilst a surgeon may record acute anteroseptal myocardial infarction.

· Multiple consistent views

Although there may be multiple views of the hierarchy required to support different functional requirements and levels of detail, these must be consistent.

· Representing context

There is a crucial relationship between concepts within the vocabulary and the context in which they are used. Cimino defines 3 types of knowledge:

· Graceful evolution

Vocabularies must be designed to allow for evolution and change, to incorporate new advances in healthcare and to correct errors.

· Recognise redundancy

Where the same information can be expressed in different ways, a mechanism for recognising equivalence is required.


© British Association of Clinical Terminology Specialists, 1999.