SNOMED & LOINC training
The narrowest, deepest, most under-supplied skill in health informatics — how to actually learn the two terminologies, and why so few people bother.
In one line
Terminology is the rarest skill in health informatics, and the one most projects discover they needed six months too late. Learning it well is an unusually high-leverage career move — precisely because almost nobody does.
Why so few people learn it
Be honest about this up front: terminology has a reputation for being boring. Engineers want to build APIs. Analysts want to find insights. Clinicians want to treat people. Nobody grows up wanting to curate a value set.
So teams skip it. They bind a field to "any SNOMED code", ship, and discover a year later that their data is unusable — that they cannot answer "how many diabetics do we have?" because the same patient was recorded eleven different ways.
That is the opportunity. The skill is scarce because it's unglamorous, not because it's hard. The demand is structural: every FHIR project, every national programme, every registry eventually hits a wall that only terminology knowledge gets through. When they hit it, they look around and find nobody in the room can help.
The two, and how they differ
They are not alternatives. They answer different questions.
SNOMED CT — clinical meaning. Diagnoses, findings, procedures, body structures. Poly-hierarchical: a concept has many parents, because reality does. Post-coordination lets you compose meanings that no single code carries.
LOINC — observations. What was measured, and how. Its six-axis structure (component, property, timing, system, scale, method) exists because "serum sodium" is not one thing — it's a component measured a particular way on a particular specimen.
Rule of thumb: SNOMED for what's true about the patient, LOINC for what was measured. The classic error is using LOINC for a diagnosis or SNOMED for a lab code, and it tells an interviewer a lot.
How to actually learn them
Start with the model, not the codes. Memorising codes is pointless — there are hundreds of thousands and you'll look them up forever. What you must own is how the thing is built: hierarchies, relationships, descriptions versus concepts, why an FSN exists.
SNOMED International's e-learning is free and good. Genuinely. Start there rather than with a paid course.
Then query it. This is the step people skip and the step that makes it stick. Learn
ECL and run real queries against a
terminology server — << 73211009 |Diabetes mellitus| and see what comes back. The moment you retrieve a subtype hierarchy yourself,
poly-hierarchy stops being a word and becomes a thing you understand.
Practise mapping, badly, on purpose. Take local lab codes and map them to LOINC. You will hit cases that don't fit. Those cases are the education — they teach you equivalence honesty faster than any lecture.
The terminology labs here exist for this: query, map, get it wrong, see why.
Licensing, and the India angle
LOINC is free. Always, everywhere, from Regenstrief.
SNOMED CT is free at the point of use in member countries — you use it under your country's membership rather than paying individually. India is a member, which means Indian practitioners, students and developers can use SNOMED CT without a personal licence. That is a genuine advantage and a surprising number of people don't know they have it.
Compare with CPT, where the licence is unavoidable. The economics of what you can learn and build differ sharply by code system, and it's worth knowing which doors are already open to you.
The career case, plainly
If you want a skill that is scarce, durable, free to learn, and immediately useful on ABDM work — where SNOMED and LOINC sit underneath the NRCeS profiles — this is it. It will not make you famous. It will make you the person the project cannot finish without.