Apple varieties or apples and apple trees? LOINC, NPU and SNOMED CT in lab tests

As a clinical terminologist, I’ve been involved in or listened to many debates where people compare LOINC, NPU and SNOMED CT terminologies (see references) in the Lab space.  It prompted me to write a few thoughts about what I think of as a misunderstanding in these comparisons: that, essentially, with LOINC, NPU and SNOMED, one is not comparing like with like; this gets in the way of such comparative analyses, although they are still very important.

So, which terminology is best in the Lab domain, LOINC, NPU or SNOMED? I think that’s an ‘Error 400, Bad Request!’  It isn’t a valid question.  LOINC (albeit broadened to physiological measurements etc.) and NPU are dedicated lab standards whereas SNOMED is not.  SNOMED is constructed as a logical set of representations of clinical utterances across the whole of medicine and beyond.  It is a general clinical standard which can then be used as the basis for derived products which can be applied in specific clinical settings (such as Lab).  The SNOMED derived product then is equivalent to LOINC or NPU or indeed any other lab standard, rather than SNOMED itself.  SNOMED can ‘do’ LOINC, it can ‘do’ NPU in the sense that content equivalences can be found.  However, SNOMED isn’t a substitute for these; it is more a formal compositional model, available as an interoperability engine for any Lab product.

When I was working on the PBCL (Pathology Bounded Code List) and the possible successor standard, UTL (Unified Test List) in the UK, the question often asked was: ‘but why did the NHS choose SNOMED rather than LOINC?’  Although the NHS’s general adoption of SNOMED is anyway a given, it is more nuanced than that.  Nowhere did anyone specifically reject LOINC; they simply positively identified SNOMED as the interoperability standard and that for me did not mean ‘not LOINC’ nor indeed ‘not NPU’ nor ‘not anything else’ when it comes to Lab standards.  It may have been perceived as an either/or in the domain but they are not ‘zero sum’.  It is possible to use LOINC expressions built from SNOMED building blocks; they do have mappable content.  In building a PBCL replacement, UTL would create logically defined SNOMED concepts that would support a ‘read across’ at an ontologic level to LOINC and NPU based on SNOMED.  This works because LOINC and NPU both come mainly out of the metrologic models of IFCC/IUPAC, as indeed does SNOMED, at least in the blood sciences domain which forms the largest part of Lab results flow.

So there is a missing ‘product x’ when lining up SNOMED, LOINC and NPU, a SNOMED based Lab product with the same domain scope as the latter two terminologies with SNOMED underpinning them all.  Perhaps this is a nice point but to me, treating them as the same sorts of objects obscures the power of the ontological basis and logical model in SNOMED.

Many scenarios in biomedical research and healthcare need to recognise relationships between different laboratory analyses.  For example, Troponin I and Troponin T are clearly distinct analyses, but high levels of both indicate cardiac damage.  With a terminology that expresses this information by additional semantic relations, applications such as clinical decision support systems can make better use of laboratory results and enhance patient care.  This is precisely where SNOMED shines – SNOMED is able to represent those enzymes and their relation to other substances and their semantic relationships can be reused for things like decision support.

We know that with SNOMED, we can build consistent, defined expressions that are essentially agnostic to the lab terminology product and metrologically sound.  It also enables a holistic representation of care.  In fact, this is why NHS Digital chose SNOMED CT as the preferred terminology, even for Lab Medicine and it’s why Termlex applies SNOMED CT’s strengths to real word informatics problems.

Those who may have watched my SNOMED Expo presentation in October on SNOMED based interoperational forms for Lab will have seen how we harmonise different lab terminologies around a common model.  At Termlex, we call it the ‘harmonised ontology’.  We can use this harmonised ontology and its expressions and terms as an informational nexus for any lab standard product if it obeys foundational ontologic formalisms.  And we can make the terms intelligible to lab scientists and clinicians in wards or clinics. Hence Pathnexus!

The apple counter at the supermarket maybe? Home delivery available.

NPU: Nomenclature for Properties and Units
LOINC: Logical Observation Identifiers Names and Codes
SNOMED CT
NHS Digital Pathology and Diagnostics hosted by NHS Digital
IUPAC: International Union Of Pure And Applied Chemistry