SNOMED CT — Clinical Terminology at Scale

A controlled vocabulary the size of a language

SNOMED CT is the most comprehensive clinical terminology in the world — over 350,000 active concepts for findings, disorders, procedures, body structures, organisms and substances. It is governed by SNOMED International, and — crucially — it is licensed: use requires a member-country or affiliate licence, not a free download.

• Three core components — Concepts (each with a numeric SCTID), Descriptions (the human terms / synonyms for a concept), and Relationships (typed links, above all the |is a| subtype hierarchy).
• Post-coordination — beyond the ~350k pre-coordinated concepts, SNOMED's compositional grammar lets you compose new meanings on the fly (e.g. a procedure + a body site + a laterality) without minting a permanent concept. This is its superpower and its complexity.
• Reference sets (refsets) — curated subsets (e.g. a national or specialty value set) that make a giant terminology usable in a specific context.
• An OWL view — SNOMED is distributed in RF2 files, but it also has an OWL 2 EL representation, chosen precisely because EL classifies enormous subsumption hierarchies quickly.

It lives in an ecosystem, not alone

SNOMED CT is the meaning layer. Around it: ICD (billing / epidemiology classification), LOINC (lab and observation codes), and HL7 FHIR (the exchange format that carries codes between systems). A competent health-data engineer knows which job each does.

Use Case Example: A hospital records a diagnosis as a SNOMED CT concept (precise clinical meaning), maps it to an ICD code for reimbursement, attaches LOINC-coded lab results, and ships the whole encounter as an HL7 FHIR bundle — four standards, one patient record, each doing what it's best at.

SNOMED CT is a language, not a list. A list of 350k diagnoses would be a phone book. SNOMED is more like English: a huge vocabulary (concepts), synonyms (descriptions), grammar (post-coordination) that lets you compose phrases nobody pre-wrote, and a thesaurus of broader/narrower terms (the is-a tree). That's why it needs reference sets — like handing a learner a phrasebook for one situation instead of the whole dictionary.

The clinical standards ecosystem

SNOMED CT carries meaning; ICD bills; LOINC codes labs; FHIR moves it all between systems.

Health and life-science licensing field guide

Healthcare standards are where vocabulary rights matter most because real terms can be clinically and commercially sensitive.

The mental model: SNOMED says what the clinician meant, ICD says how it is classified for reporting/billing, LOINC says which observation was measured, FHIR carries the data, IDMP identifies the medicine, and GO/OBO explains biological function with evidence.