Having a debate with some colleagues about the SNOMED CT binding in a proprietary UK GP system with respect to blood pressure, which uses:
163020007 |On examination - blood pressure reading (finding)|
Which is a finding.
To be terminologically correct my understanding is that values should be observable entities.
So I went to check our openEHR Blood Pressure archetype and it seems it too is using:
163030003 |On examination - Systolic blood pressure reading (finding)|
163031004 |On examination - Diastolic blood pressure reading (finding)|
So I think we should be suggesting bindings to:
271649006 |Systolic blood pressure (observable entity)|
271650006 |Diastolic blood pressure (observable entity)|
as everyone loves purist terminology debates, especially at XMas
What’s the thinking here?
Hi Paul! The SNOMED CT bindings in the current trunk are obviously wrong, including the binding for the archetype concept. The Norwegian archetype governance organisation is working with the Norwegian SNOMED CT NRC on a new set of bindings for all of the vital signs archetypes. You can find a preliminary version in this branch: https://openehr.org/ckm/archetypes/1013.1.4200
Part of this work will be a published guide on which SNOMED subtrees are valid binding options for the concept level (if any), data elements, and value sets in archetypes and other information models.
@Sam will recall the background to those original bindings . As I recall, Sam added them but asked some of the leading UK terminologists (15 years ago!) for advice. There was a lot of mumbling /disagreeing and a delay of a few days before they came back and said they were correct.
Part of the issue was that historically (partly due to READ codes), findings and observables have been a bit mixed up. Daniel Karlsson will know more about this as he worked on the SNOMED observables project.
The other issue to contend with is that LOINC is likely to be the key source of truth for terminology bindings for the ‘name’ of elements, rather than SNOMED CT, so the binding from LOINC to SNOMED may be more important than the ‘proper’ ontological binding in SNOMED terms alone.
The current UK guidance
and SNOMED Int. guidance
And there are some very evil bindings in that SNOMED document e.g Target body weight as a ‘synonym’ for body weight and the same in the UK document (which now looks to be a copy) - That needs corrected - potential risk there.
Nope I got that wrong - misread the SNOMED ECL syntax - these items are excluded - I take it all back, hang my head in shame and will deprive myself of a mince pie as punishment.
However, you can see the challenge of using SNOMED as an accurate binding to an archetype element name because the binding …
"<<271649006|Systolic blood pressure (observable entity)|
MINUS (<<716579001|Baseline systolic blood pressure (observable entity)|
OR <<814101000000107|Systolic blood pressure centile (observable entity)|
OR <<315612005|Target systolic blood pressure (observable entity)|)"
Tells us that ‘Target systolic blood pressure’ in SNOMED is-a kind of Systolic blood pressure, which may be true in an authoring context but is not true in a clinical context.
I’ll stop talking to myself now
Sorry, had a weekend off with ongoing flu!
From my perspective we will use observables where we can and ideally, if needed, value sets derived from these although mostly single concepts will be ok. I agree it can become much more complex and we may need to do that kind of thing downstream.
SNOMED CT is also a bit mixed up with observables as there is not always an observable available when you might expect one. Sure this will be fixed over time.
Am trying to avoid LOINC because brain full already