FHIR Profiles

eric: One goal with ShEx is to do an elegant job of modeling FHIR Profiles.
... Inelegant would be to have ShEx for profiles, but you cannot use the original shex after constraining it.
... But an elegant way would be to use the original shex and simply add shex constraints to make a FHIR profile.
... Now trying to use shex in a project, and eventually we'll get to profiles. I want the extensions to do the Right Thing.

harold: About 3 years ago we went through the FHIR constraint language and found that a significant chunk was co-occurrences.

eric: Eg, if you have a timing duration, then you need a timing duration units.

harold: Nice thing about those is that they can be modeled using sets.
... Another example: A hematacrit result requires a numeric result.
... But there were some constraints that seemed to need a path language to resolve correctly.

<scribe> ACTION: David to search the minutes for discussion relevant to FHIR profile constraints

eric: Who understands slicing?
... After asking Grahame some questions, i seemed to get contradictory answers.

harold: Maybe Claude or Bryn.

eric: Currently looking for hard cases.
... What fraction of the FHIR constraints does shex already cover?
... What fraction of valuesets are covered, and what of those would you want to do in shex, rather than referencing something defined externally?
... Slicing is another challenge
... And what wrinkles show up if you extend another profile in shex
... Folks I'm working for are using FHIR/RDF and shex, and we're approaching a release candidate, so they all need to work together.

<hsolbrig> https://github.com/smart-on-fhir/fhir-parser -- generator for FHIR python

<hsolbrig> http://build.fhir.org/

<hsolbrig> https://github.com/HL7/fhir

<hsolbrig> http://build.fhir.org/fhir.schema.json.zip

<hsolbrig> definitions.json.zip

ADJOURNED