FHIR Chat · Conceptmap · implementers

I'm working on an implementation of concept mapping that tries to conform to FHIR's ConceptMap as much as possible. we have a use case which involves translating an entire set of concepts in one set of systems to another set of systems. I was looking at the $translate and $closure operations for ConceptMap, and $translate seems to match this operation, but only for a single input concept (when in our use case, we have many)

Andrew Ross (May 06 2016 at 15:00):

would it make sense to either update the $translate operation to accept multiple concepts, or to add another operation (maybe as a POST) to allow for the $translateMany case?

Grahame Grieve (May 06 2016 at 15:00):

http://hl7.org/fhir/terminology-service.html#2.5.1.9

Andrew Ross (May 06 2016 at 15:00):

ah, nice!

Andrew Ross (May 06 2016 at 15:01):

i think that answers my question, thanks!

Grahame Grieve (May 06 2016 at 15:02):

Andrew Ross (May 06 2016 at 15:06):

actually... this might not fully match our use case. i'm looking at the docs for batch operations, and it seems like they require that there are no interdependencies between the individual elements, but here there are. For example, if coding system 1 has concepts for "male" and "breast-cancer", while coding system 2 has a concept for "male-breast-cancer", when I ask to translate "male" and "breast-cancer" in coding system 1 to coding system 2, i want to receive "male-breast-cancer" back

Andrew Ross (May 06 2016 at 15:08):

for a single request, maybe the right way to structure it would be to request a translation for one of the concepts with the other added to the "dependency" section

Grahame Grieve (May 06 2016 at 15:08):

ahh. then batch isn't what you want

Andrew Ross (May 06 2016 at 15:08):

but for a batch request, i want each concept's "dependency" to be every other concept

Grahame Grieve (May 06 2016 at 15:09):

look here:

Grahame Grieve (May 06 2016 at 15:09):

http://hl7.org/fhir/conceptmap-operations.html

Grahame Grieve (May 06 2016 at 15:09):

at dependency and product

Andrew Ross (May 06 2016 at 15:09):

yep

Andrew Ross (May 06 2016 at 15:10):

i guess we could just have it be a single request, where we pick one concept at random to be the main one and put the rest into "dependency", then get back some number of matches/products

Andrew Ross (May 06 2016 at 15:11):

but it can't really be a GET request, because there might be quite a lot of inputs

Andrew Ross (May 06 2016 at 15:12):

this is all happening internally for right now, as part of how we interpret data we're getting in different coding systems from different EHRs

Andrew Ross (May 06 2016 at 15:12):

but ultimately we might expose our own mapping logic as a service

Grahame Grieve (May 06 2016 at 15:12):

sure, post

Grahame Grieve (May 06 2016 at 15:13):

the question of which is primary - the underlying ConceptMap requires that you identify a primary one, and so it would match that.

Andrew Ross (May 06 2016 at 15:13):

right

Grahame Grieve (May 06 2016 at 15:13):

Agree that in some cases there isn't a 'primary' but saying that you identify a primary simplifies the model

Vadim Peretokin (Jul 01 2016 at 03:20):

ConceptMap has a target ValueSet - can I only map to codes that are a part of the target valueset? What if I map to a codesystem that's not included in the target valueset? This is given the May2016 snapshot, I know group changed things slightly in latest CI version