FHIR Chat · Migration of data between IG versions · implementers

Stream: implementers

Topic: Migration of data between IG versions


view this post on Zulip Morten Ernebjerg (May 29 2020 at 13:26):

Hi :wave: I was wondering if there are guidelines and examples for using ConceptMap and/or StructureMap to define the relation between different versions of a single IG. The use case I have in mind is where an IG is in productive use but still being extended or modified in (primarily) non-breaking ways (i.e. MINOR changes in the SemVer lingo) - as one might have it in a very dynamic situation like COVID-19. Typical change examples would be changes in individual codes or value sets. In such a scenario, it would be ideal to equip each update with a "migration script" to allows user to bring their already collected data to the latest IG stage (or researcher to do it at a later point to pool data). Are there examples of that?

view this post on Zulip Morten Ernebjerg (Jun 23 2020 at 09:18):

Trying to revive this as there were no answers so far:point_up: - would be very grateful for any experiences, lessons learned, ideas... :slight_smile:

view this post on Zulip René Spronk (Jun 23 2020 at 09:36):

It probably means nobody ahs done this before, and that all of us are waiting for your best practices :-) Please do tell us!

view this post on Zulip Lloyd McKenzie (Jun 23 2020 at 14:32):

It's not a bad idea, it's just a degree of work that not a lot of IG implementers have been willing to take on.

view this post on Zulip Lloyd McKenzie (Jun 23 2020 at 14:33):

It's done for the core spec, but not as yet for any HL7 International-published IGs.

view this post on Zulip Grahame Grieve (Jun 23 2020 at 21:46):

how would you migrate existing data for new codes?

view this post on Zulip Morten Ernebjerg (Jun 24 2020 at 16:15):

@Grahame Grieve The one concrete case I had in mind was that of changing the value set specifying the allowed answers to a specific question in a Questionnaire. Specifically, I was thinking of the case where there is a 1-to-1 relation between the codes in the old and the new value set so a precise mapping is possible (in QuestionnaireResponse instances). For instance, both SNOMED and LOINC have code codes for "yes" and "no", so one could switch from one to the other and still map the data.

Background: In Germany, there is an ongoing effort to align the (FHIR) data collected on COVID-19, the COCOS initiative. We put out an IG for a COVID questionnaire app while this initiative was just getting started. However, in some instances the codes we used differ from those COCOS is now converging on. Hence, I am considering how to align our IG while still providing a "bridge" for existing data. COCOS hat tip to @dr Kai U. Heitmann - maybe this is interesting to you as well.

view this post on Zulip Grahame Grieve (Jun 24 2020 at 20:22):

ok sure that's a situation where you could. So if you're IG has 2 versions, you can publish a structure map that identifies the 2 versions and maps between them

view this post on Zulip Grahame Grieve (Jun 24 2020 at 20:22):

there's no IG infrastructure to support doing anything with that inside the IG itself


Last updated: Apr 12 2022 at 19:14 UTC