I don't see that happen easily. Validate your archetypes on content also before using them.
Not an unreasonable point of view, but it sort of implies that there are / will be no well-known / reliable terminology value sets out there - only specific value sets inside specific terminology services.
This problem hads been tackled by IHTSDO. They never allow a concept to disappear, and all members should install the latest updates. There is a lot of thought inside SCT about versioning.
A hospital can run its own SCT service (in member countries, this is free), and a hospital which has a large network problem is in big problems, independent from the fact if it uses SCT.
code subsets are important in data validation or query part, I would
say not so much in semantic label definition
sure thing, that's why we need standard expressions
Exactly Michael, that is what I was pointing at in my blogs a few days ago, making a golden pair of SCT and OpenEHR by combining the full potential of both.
Even go further then FHIR, because FHIR is a message system, while OpenEHR is so much more then that. OpenEHR is the base of an application generator.
sure thing, that's why we need standard expressions
Maybe I don't understand your point. But what I make of it, is that you are referring to the ever lasting discussion about a standardset of archetypes, or every medical institution making its own.
SCT can maybe also solve this point, when every archetype has a post coordinated expression in it which represents the archetype, all information will be machine processable/understandable.
At least, that is the way SCT wants to go, I cannot judge in how far this succeeds, but it succeeds better then OpenEHR does without SCT, I think.
My comments were on the discussion 'expressions vs queries to external
terminology services in archetypes'
does the expansion preserve IS-A relationships (at least optionally)? That's crucial for making any value set of more than about 20 terms usable in a real system.
- thomas
we also still need a standard approach for non-SNOMED CT terminologies, such as ICDx, ICPC, ICF, LOINC and a hundred others... does anyone know of progress on this issue?
- thomas
I think you need to do that by additional refinements, f.e.
< 19829001 |disorder of lung|: 116676008 |associated morphology| = 79654002 |edema|
or
19829001 |disorder of lung|: 116676008 |associated morphology| = 40829002 |acute edema|
or
19829001 |disorder of lung|: 116676008 |associated morphology| = 44132006 |abscess|
Examples from: http://snomed.org/expressionconstraint
There is a detailed answer, I googled on SNOMED to ICD10
https://www.nlm.nih.gov/research/umls/mapping_projects/snomedct_to_icd10cm.html
I think others on this list are better informed to answer this question
Thomar,
I know of the SOLOR project where SNOMED is harmonised with LOINC and RxNorm
http://wiki.hl7.org/index.php?title=CIMI_Quality_Modeling_Collaboration
http://www.healthcare-informatics.com/blogs/david-raths/interoperability/can-solor-snomed-loinc-rxnorm-project-create-terminology
http://www.businesswire.com/news/home/20160719005555/en/Healthcare-Services-Platform-Consortium-Launches-Data-Standard-Integration
Gerard
FHIR terminology servers can (and mostly do) handle all of those terminologies, though I don’t know if anyone has handled ICF in practice.
And expansions can preserve is-a relationships if you want, though… life is complicated and the answer is not automatically ‘yes’
Grahame
Bert,
these are just selectors; what I mean is that in the generated result - the actual value set - that IS-A relationships are returned as well as concept codes. Without IS-A relationships a user can't navigate a value set larger than a few terms in a useful way in a real system.
- thomas
I am aware of the SNOMED / ICD10 mapping project, but it's not directly helpful to all the routine users of ICDx (many variants, not just ICD10), who use it on its own and will do so for years.
- thomas
The best way to resolve this is to make the terminology server part of the local system, and have the resolution a dynamic one between the systems. That allows you to optimise the performance implications of large value sets.
Grahame
indeed there is. But what I am talking about here is not whether Snomed value sets can be versioned but the non-technical question of whether specific, well-known value sets can be established globally, regardless of specific terminology servers etc and whether they can then be relied on globally to exist in the same way a specific concept code is assumed to exist. Specific value sets (generally large ones) have been defined on Snomed CT, what I am thinking about is a fast, efficient mechanism for creating curating and publishing any value set, of which thousands will be needed, or whether each health organisation will end up replicating this work and creating similar / same private value sets with different identifiers.
It may be that this is an argument for sharing value set definitions (i.e. the constraint expressions) within AQL queries and archetypes as being a way to do this.
- thomas
In other words something like a DB cursor to traverse large value sets that reside on the server, in response to client (user) actions on the screen? Has that been implemented in FHIR-land?
- thomas
yes. In our terminology, this is paging through an expansion
Grahame