LOINC and SNOMED CT are not the same (and mixing them is expensive)
It is one of the most expensive mix-ups in a laboratory project: using LOINC where SNOMED CT belonged, or the other way round. The catalogue looks coded, passes internal validation, and turns out to be useless for interoperability.
LOINC names the question
LOINC identifies what was measured: "glucose in serum", "haemoglobin in whole blood", "urine culture". It is the name of the test or analyte, with its property, unit, specimen type and method.
SNOMED CT names the answer
SNOMED CT identifies what was found: the isolated organism, the finding, the diagnosis. When the result is a clinical concept rather than a number, that concept is SNOMED, not LOINC.
The boundary, in one line
The test is coded in LOINC. The qualitative result is coded in SNOMED CT.
Where it breaks in practice
- A panel assigned the LOINC code of one of its analytes: the receiver gets a part and believes it has the whole.
- A urine culture whose organism is stored as free text: no epidemiological surveillance is possible.
- The same analyte coded differently per site: historical values stop being comparable.
Why it matters now
In HL7 FHIR, an Observation carries the test in code (LOINC) and the conceptual result in valueCodeableConcept (SNOMED CT). If the mapping is wrong in the catalogue, no integration layer will fix it later.
- LOINC
- SNOMED CT
- interoperability
- HL7 FHIR