LOINC and SNOMED CT overlap on some observations and diverge on others. Deciding which vocabulary owns a specific observation is the small choice that shapes downstream terminology consistency for the life of the deployment. The choice is not arbitrary; both vocabularies have areas of strength and an established convention for the boundary.
Naming the boundary is the discipline. For related walkthroughs, more FHIR implementation patterns collects the surrounding material.
The Rule of Thumb
LOINC names measurements; SNOMED CT names findings. A lab result value belongs in LOINC. A clinical finding, diagnosis, or observation of a state belongs in SNOMED CT.
The rule is loose but useful. Deployments that internalize it converge on consistent terminology; deployments that pick vocabularies ad hoc end up with mixed usage that confuses downstream systems.
Where LOINC Wins
LOINC is the right choice for:
- Lab test results: sodium level, hemoglobin A1c, urine culture.
- Vital sign measurements: systolic blood pressure, heart rate, respiratory rate.
- Panel structures: CBC, basic metabolic panel, comprehensive metabolic panel.
- Assessment scales: PHQ-9 depression screening, Braden pressure ulcer risk.
A pass through the site's LOINC picker surfaces the common LOINC options across these categories.
Where SNOMED Wins
SNOMED CT is the right choice for:
- Clinical findings: hypertension, diabetes, chronic kidney disease.
- Anatomical locations: left kidney, right pulmonary artery.
- Procedures: appendectomy, cardiac catheterization.
- Body structures and morphologies.
Findings and procedures are where SNOMED's compositional grammar earns its keep. LOINC does not attempt to express these categories.
The Overlap Zone
Some observations sit ambiguously between the two vocabularies. A vital-sign measurement like blood pressure has a LOINC code; the finding of "hypertension" has a SNOMED code. Both apply to the same clinical scenario at different points in the workflow.
The pattern is to use LOINC for the measurement record and SNOMED for the clinical interpretation. Both live in the FHIR archive as separate resources. For the category framing, distinguishing lab tests, panels, and observations in LOINC covers the split.
The FHIR Binding Convention
FHIR profiles usually bind specific fields to specific vocabularies. Observation.code is typically bound to LOINC (or a value set that includes LOINC codes). Condition.code is typically bound to SNOMED CT. Following the convention keeps profile compliance easy.
Deployments that deviate from the convention should document the reason. Downstream consumers assume the standard binding until they discover otherwise.
The Cross-Coding Pattern
Some observations carry both codes. A blood pressure measurement might record LOINC for the measurement and SNOMED for the interpretation (elevated, normal, hypotensive). The dual coding preserves the strengths of both vocabularies.
The pattern is not universal. Simple measurements do not need it. Complex assessments benefit from it.
Local Codes as a Fallback
When neither vocabulary carries a suitable code, a local code is the interim answer. For the specific proposal framing, when a local code has no LOINC match and you need to propose one covers the pathway.
The Documentation Habit
Every observation type in the deployment should have a documented vocabulary choice. LOINC for measurements. SNOMED for findings. Local codes for the gaps. Every LOINC integration that lands well maintains this mapping publicly.
Picking LOINC or SNOMED for a specific observation is not the hard part. Documenting the choice so it survives team rotations is.
Sources
- HL7 FHIR terminologies chapter defining LOINC vs SNOMED - HL7 FHIR terminologies chapter defining LOINC vs SNOMED responsibilities for Observation.code and value
