Hierarchies & is-a relationships
Every SNOMED concept lives in a tree of meaning. 'Is-a' links are what let one rule catch a thousand diagnoses.
The answer in one paragraph
Every SNOMED CT concept is linked upward to more general concepts by is-a
relationships: Acute myocardial infarction is a Myocardial infarction is a
Ischemic heart disease is a Heart disease — all the way to one of 19 top-level
hierarchies under the root. This is what turns a dictionary into a reasoning tool: a
decision-support rule written against Myocardial infarction automatically applies to
every subtype, including ones added in future releases, without anyone editing the rule.
The 19 top-level hierarchies
Everything descends from the root concept (138875005 |SNOMED CT Concept|) into 19
branches. The ones you'll touch most:
- Clinical finding (~123k) — diseases and symptoms; the biggest by far
- Procedure (~59k) — things done to or for a patient
- Body structure (~41k) — anatomy, including morphologic abnormalities
- Organism (~34k) — bacteria, viruses, everything that infects
- Substance / Pharmaceutical product — chemistry vs. orderable products
- Observable entity — things you can measure ("heart rate" as a question; the answer is usually a value with units — LOINC's overlap territory)
- Qualifier value, Situation with explicit context, Event, Social context… — the supporting cast
The semantic tag in every FSN tells you instantly which branch you're in.
Is-a is a graph, not a ladder
A concept can have multiple parents: Bacterial pneumonia is both an
Infectious pneumonia and a Bacterial lower respiratory infection. So the structure
is a directed acyclic graph — richer than a folder tree, and the reason subtype queries
are so powerful.
Defining relationships: more than is-a
Concepts are also defined by attribute relationships drawn from the concept model:
22298006 |Myocardial infarction|
→ finding site: Myocardial structure
→ associated morphology: Infarct
These definitions are what allow the terminology to be computed: a description-logic classifier checks that every concept sits exactly where its definition implies. When you later write expressions (next entries), you're using this same attribute machinery.
Why this matters to you, concretely
- Analytics: "all patients with any ischemic heart disease" is one subtype query — not a hand-maintained list of 200 codes.
- Decision support: subscribe to a parent; inherit every descendant forever.
- Data quality: if the concept you picked sits in the wrong hierarchy (check that semantic tag), every downstream query silently misses it. Hierarchy awareness is coding quality.
Where to go next
Post-coordination & expressions — what to do when the concept you need doesn't exist as a single code.