Stream: CCDA / FHIR mapping stream
Topic: Progress
Grahame Grieve (May 20 2018 at 21:58):
@Jean Duteau I'm home and recovering from the WGM, and actually working now.... where are we with this?
Jean Duteau (May 20 2018 at 22:13):
The most important things are:
1) CDA.Narrative to/from FHIR.xhtml
2) ConceptMap translations - could be my syntax
3) getting my source where condition clauses working right
For #3, see above where I give you my mapping that isn't working
For #2, you think I'm using them the "wrong way around", so I'd need to know what I'm doing wrong
And #1 needs to be done but I'm now not sure that it should be done in code.
Grahame Grieve (May 20 2018 at 22:17):
#1 - ok why not in code?
Grahame Grieve (May 20 2018 at 22:17):
#2 - the translations are unidirectional - from FHIR to CDA. but you're invoking them as if they are from CDA to FHIR
Grahame Grieve (May 20 2018 at 22:18):
#3 - the conditions work for me but they have to start with a variable name
Jean Duteau (May 20 2018 at 22:23):
okay, for #2 - I guess we (and when I say we, I probably mean my team) need to create the CDA->FHIR mappings?
Jean Duteau (May 20 2018 at 22:23):
#3 - I do have variable names. see above to see what I've done
Grahame Grieve (May 20 2018 at 22:24):
#2 yes, for now. guess we should normalise this into the spec?
Jean Duteau (May 20 2018 at 22:26):
#1 - the mapping language and the maps are, right now, tool agnostic. anyone can potentially write a mapping engine and then execute any maps. that seems to me to argue that all maps should be done via map files and not be built-in to any tool. basically, if you write the narrative text transform into the validator code, then anyone else who wants to build a mapping engine is going to have to replicate that code in their engine.
Jean Duteau (May 20 2018 at 22:27):
thus whatever you write in code needs to be totally documented so that others can replicate it. wouldn't the best way to document that be via a mapping file?
Grahame Grieve (May 21 2018 at 21:12):
well, it would sure be the most comprehensive mapping. but is is the most effective? That I'm not sure about. Here's what we have for mapping at the moment:
Grahame Grieve (May 21 2018 at 21:13):
http://wiki.hl7.org/index.php?title=CDA_Narrative_to_html_mapping
Jean Duteau (May 21 2018 at 22:55):
i would hope that others would chime in on this.
Grahame Grieve (May 21 2018 at 22:56):
@Rick Geimer volunteered to look at it.
Jean Duteau (May 21 2018 at 22:59):
i looked at it and updated it a bit. it's not the mapping itself but rather whether this should be done in code or external that I'd love more than you and I to decide.
Grahame Grieve (May 22 2018 at 12:14):
thinking more about it... the problem with doing it in the mapping is that then I have to work up models for both xhtml and Struc.TExt
Grahame Grieve (May 22 2018 at 12:39):
that sounds like a heavy lift and is outside what I expected to do in the short term
Grahame Grieve (May 22 2018 at 16:46):
https://build.fhir.org/ig/HL7/fhir-cda/
Jean Duteau (May 22 2018 at 16:56):
that's nice.
Grahame Grieve (May 22 2018 at 17:38):
could be made very nice if someone wanted to. e.g. could replace the existing CDA guide - start with the RMIM and all. Just a matter of how much someone wants to invest
Jean Duteau (May 30 2018 at 04:38):
And any progress on Narrative <-> XHTML?
Grahame Grieve (Jun 04 2018 at 08:13):
so looking at the wiki page...
Grahame Grieve (Jun 04 2018 at 08:13):
what
Grahame Grieve (Jun 04 2018 at 08:13):
what's the right thing to happen with referencedObject?
Grahame Grieve (Jun 05 2018 at 00:47):
hey Jean - one of your test documents has this in it:
Grahame Grieve (Jun 05 2018 at 00:47):
<substanceAdministration classCode="SBADM" moodCode="INT">
<templateId root="2.16.840.1.113883.10.20.22.4.16"/>
<id root="a0e312da-2007-4f4b-b8ed-3973e93b948d"/>
<text>
<reference value="#medication920000875889"/>
</text>
<comments>
<reference value="#medcommentsmedication920000875889"/>
</comments>
Grahame Grieve (Jun 05 2018 at 00:47):
I don't recognise the comments element?
Grahame Grieve (Jun 05 2018 at 00:50):
.. I'm doing round trip testing on the CDA documents...
Grahame Grieve (Jun 05 2018 at 03:51):
ok. committed. Round-tripping CDA -> logical model -> CDA looks good to me (attributes get re-ordered, of course, but everything is preserved that is valid CDA). That means that CDA narrative will not map correctly.
Grahame Grieve (Jun 05 2018 at 03:52):
the way the FHIR framework works, doing the mapping by the mapping language is actually very difficult; it's not about the mapping language, but about the way StructureDefinition has been set up to handle CDA narrative - we effectively already decided that it would be converted by magic.
Grahame Grieve (Jun 05 2018 at 03:53):
but the code is actually pretty simple, since there's no context to carry around (yet!):
Grahame Grieve (Jun 05 2018 at 03:58):
Grahame Grieve (Jun 05 2018 at 04:48):
by product of testing out the CDA round-tripping: CDA in Json....
Grahame Grieve (Jun 05 2018 at 04:49):
Grahame Grieve (Jun 05 2018 at 04:55):
what is this?
<doseQuantity/> <doseQuantity unit="Tablet(s)" value="1"/> <doseQuantity unit="MG" value="300-30"/>
Grahame Grieve (Jun 05 2018 at 04:56):
in what world is that valid?
Grahame Grieve (Jun 05 2018 at 05:03):
we can have RDF CDA documents too ;-)
Grahame Grieve (Jun 05 2018 at 05:03):
Grahame Grieve (Jun 05 2018 at 05:04):
though the only thing that shows is that RDf is pretty much the worst way to say anything useful
Grahame Grieve (Jun 06 2018 at 05:01):
btw, the CDA Implementation guide - as opposed to the CCDA on FHIR implementation guide... needs some love. We kind of need to figure out how far we want to take it... if we really wanted to, it could replace the CDA spec (R2.1, I guess....) . We probably don't want to... but we need to base artifact, so we need to decided how much to do with it
Last updated: Apr 12 2022 at 19:14 UTC
