Knowledge Representation and Clinical Ontologies

Knowledge representation is the part of clinical informatics concerned with encoding medical meaning in a form that computers can store, share, and reason over. Controlled terminologies and ontologies provide the formal vocabularies, concepts, and relationships that let clinical data and decision-support logic refer to the same things unambiguously, supporting interoperability and automated reasoning.

Definition

In clinical informatics, knowledge representation is the use of formal structures, such as controlled vocabularies and ontologies, to capture concepts and the relationships among them so that clinical meaning can be processed by software; an ontology is an explicit, formal specification of a shared conceptualisation of a domain.

Scope

The entry covers controlled vocabularies, terminologies, and ontologies used in health care; the difference between a simple code list and a description-logic ontology with defined concepts and relationships; major resources such as SNOMED CT, the UMLS Metathesaurus, and LOINC; and the role of these artefacts in interoperability and decision support. It is a methodological and infrastructural topic, describing how clinical knowledge is structured rather than offering clinical recommendations.

Key concepts

• Controlled vocabulary and terminology
• Ontology and concept hierarchy
• Description logic and formal semantics
• SNOMED CT, LOINC, and RxNorm
• Unified Medical Language System (UMLS) Metathesaurus
• Semantic interoperability
• Concept mapping and terminology binding
• Post-coordination and compositional expression

Mechanisms

Terminologies assign stable identifiers to clinical concepts and arrange them into hierarchies; ontologies add formally defined relationships (such as is-a and part-of) that support automated classification and reasoning. Integrating resources such as the UMLS links many source vocabularies through a common Metathesaurus and semantic network, so that synonyms and cross-vocabulary equivalences can be resolved (Bodenreider, 2004). The Gene Ontology illustrated, in molecular biology, how a shared structured vocabulary can unify annotation across databases, a model that influenced biomedical ontology practice broadly (Ashburner, 2000). Standards-based interfaces then let coded data and decision logic interoperate across systems (Mandel, 2016).

Clinical relevance

Coded terminologies underlie problem lists, laboratory results, medication records, and the rules that decision-support systems evaluate, so the quality of knowledge representation affects whether data can be aggregated, exchanged, and acted on reliably. This entry describes the structures behind coded clinical data; it does not define clinical meaning for any individual case or provide treatment guidance.

Evidence & guidelines

Knowledge representation is largely a matter of standards and infrastructure rather than clinical-outcome trials. Foundational resources are described in their primary publications: the UMLS as an integrating layer over many vocabularies (Bodenreider, 2004) and the Gene Ontology as a unifying structured vocabulary (Ashburner, 2000). Interoperability standards such as SMART on FHIR define how coded data and apps exchange information across platforms (Mandel, 2016).

History

Controlled medical vocabularies date back decades, from early nomenclatures to MeSH and the development of SNOMED. The 1990s saw the UMLS integrate disparate vocabularies, and around 2000 formal ontologies, exemplified by the Gene Ontology, brought description-logic semantics into biomedical knowledge. Subsequent standards work focused on interoperability so that coded knowledge could move across institutions and systems.

Debates

How expressive should a clinical terminology be?

Highly expressive, post-coordinated ontologies capture nuance but are harder to author, maintain, and use consistently, while simpler enumerated code lists are easier to apply but lose meaning; the trade-off between expressiveness and usability remains contested.

Key figures

• Olivier Bodenreider
• Mark A. Musen
• Christopher G. Chute
• Michael Ashburner

Seminal works

• bodenreider-2004
• ashburner-2000

Frequently asked questions

What is the difference between a terminology and an ontology?

A terminology is primarily a controlled list of named concepts, often arranged hierarchically, whereas an ontology adds formally defined relationships and logical semantics that allow software to classify and reason over concepts automatically.

Why are clinical ontologies important for decision support?

Decision-support rules and patient data must refer to the same concepts to interoperate; shared ontologies and terminologies give them a common, machine-readable vocabulary, which is a prerequisite for portable, reliable support.