Meeting minutes
Issue 94 - Concept URIs
gaurav: Modified UTG repo to include prefixes. Not all of the info in the XML files in there get put into the node package.
https://
gaurav: I have links to SNOMED CT. Was able to put in prefixes for all but the last three.
… MESH, PUbmed, and NCBI Nucleotide
… Also working on FHIR JSON-LD repo, that EricP and James have been working on.
… Package assumes that you have HL7 terminology package install. Then it goes through all the JSON files in the package collecting all the prefixes. Given a coding it will give you a concept URL by attaching a prefix.
… It will also translate back by naively chopping off the code from the end of the URL.
… Now working on other prefixes, and I'll add tests for them also.
… Next need to write code to go into all of the FHIR examples, get a list of all systems and codes, and try converting all of them.
… But also check them for prefixes that we cannot handle with a simple prefix.
… Want to be added to FHIR JSON-LD repo.
https://
gaurav: My test file: https://
brad: I think there's a more canonical SNOMED prefix, I saw in Harold's tutorial.
eric: Tempted not to bother with versions.
david: Agree.
eric: Importance of versioning diminished after protest about changes.
david: How prevalent is versioning in the system / code pairs?
eric: A lot of what people care about in versioning are about which choices were available at time t.
… So you need versions on valuesets but not concepts.
David: Do we have the same issue with valueset URIs?
eric: We don't talk about the URIs for valuesets.
gaurav: Next I need to pull system and code from FHIR examples, so that we know what's covered and not covered.
RDF lists and OWL
jim: Looking at Harold's tutorial, querying for cancer diagnoses.
… Talked about looking at use cases for FHIR+OWL. What do we want to accomplish?
… Discussed problems that RDF lists cause for OWL. What's our goal?
… It would throw a big wrench to OWL if RDF lists were used.
eric: Should go back to the community. What would OWL apologists recommend we do?
david: Our goal is to make FHIR RDF easier to use.
jim: You'd have to introduce your own list.
eric: Could say a codeableconcept has a list of codings.
jim: That should work intuitively, but it's prohibited by the spec.
… The OWL WG looked at, thought about it, and decided not to do it.
eric: Does it actually work in common reasoners, even if it isn't technically legal?
ACTION: Jim to see if RDF lists work in common reasoners (even if they're not technically legal)
jim: Deeper problem: You have to decide if lists hold either objects or literals -- not both.
ShEx work for FHIR RDF
eric: Working on "extends" capability for ShEx.
FHIR Knowledge graph
dagmar: Might scare off some people who work with KGs, because it takes a lot of nodes to get anywhere.
… Issues brought up by Harold in 2019 for improvement, such as fhir:value
… Makes the KG too complicated.
david: That's issue 77: https://
eric: I'll be prototyping that in the playground.
dagmar: Could also transform the data before working with it.
… For data science.
… Got videos from last year's KG conference.
… Could also transform to another RDF format to focus on the main knowledge of interest.
eric: Stan Huff's CIMI group put a lot of effort into this. Clinical models more of interest, rather than EMR models.
… They'll have a blood pressure with systolic and diastolic directly.
eric: Working on grant involving transformation of data, clinical models to/from FHIR.
http://
eric: lots of times people have tried to do this. OpenEHR also tried to do this.
ben: OpenEHR is still going. Some traction in some pockets.
… Australia, South America.
… Lots of connection with FHIR work.
EBMonFHIR and COKA
ben: Wanted to raise awareness about this: This is the group I was referring to the met the other week https://
… would raise awareness of initiative
ADJOURNED