Semantic Web Health Care and Life Sciences Interest Group Teleconference -- 21 Jun 2016

Remaining items that need to be implemented in FHIR RDF

<inserted> Harold's slides: https://lists.w3.org/Archives/Public/www-archive/2016Jun/att-0003/FHIR_RDF_Extensions_v2.pdf

harold: Pattern, minValue and maxValue are now implemented.

<hsolbrig> file:///Users/mrf7578/Development/fhir/build/publish/patient.shex.html

harold: Also put comments in the shex.

<hsolbrig> http://hl7-fhir.github.io/medication.shex.html

harold: Eric wondered if we could drop the turtle template language. Eric changed commas to semicolons.

<hsolbrig> http://hl7-fhir.github.io/fhir.shex

harold: Also created fhir.shex, which has all the shex definitions of everything.
... I want to remove the comments, because they make it much bigger.
... When we do something like Patient.shex, all of these shex definitions are CLOSED.
... When we're pulling definitions into a bundle it's wide open -- it must have at least one type arc.
... In the fhir.shex it makes more sense to change that wide open def to one that explicitly lists all of the possible resources.
... But that would only be in the fhir.shex.

dbooth: CLOSED means a closed content model?

harold: Yes. If you have something else it's an error.

dbooth: What does "<Resource> {a .+;" mean?

harold: the dot is a wildcard, meaning anything.

eric: regarding CLOSED, on one hand is the RDF ideal of allowing other content and ignore it you don't care about it.
... Another is, for clinical exchange, where you know the content model and you don't want anything else. Even the extensions fit into the closed content model, which means that the typos get caught.

harold: whenever I generate a new version of shex I test all of the FHIR examples, and this works against them.

dbooth: The passing examples?

harold: yes. No failing examples yet.

james: Suggest enumerating as little as possible.
... To reduce maintenance difficulties.
... Having an abstract class may be better.

harold: grahame wanted a case with completely closed system.

james: Then you could never subclass it.

harold: To extend a core resource, you extend it with tagged value pairs.
... I'm not allowed to add a Price element to the Medication element.
... But I can create an extension by adding an extension element.

eric: it is a closed content model, you cannot add to it, but one thing that is in there is an extension mechanism with the FHIR keyword and grammar.

dbooth: So if you want to subclass in the RDF world, you can do that until you want to send your content as FHIR RDF to someone else. Then you have to limit what you send to ONLY the FHIR RDF portion.
... We could consider standardizing a way to put arbitrary RDF in the FHIR extension mechanism.

harold: I need two different kinds of code, to handing FHIR RDF extensions: one for basic FHIR RDF, the other for the extensions.
... Our ideal approach would be to have the extensions and core FHIR resources be similar, though with an extension we're not guaranteed to have the schema available.

ISSUE: Should we try to unify the treatment of extensions with regular resources?

<trackbot> Created ISSUE-37 - Should we try to unify the treatment of extensions with regular resources?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/37/edit>.

shahram: This works if extensions are not modifying extensions.

harold: Correct. Extensions can theoretically override the base semantics.
... I vaguely remember hearing someone say that you can change the types is the core elements.
... But I'm trying to address Well Behaved extensions.

eric: At least in the past there were modifier and non-modifier extensions. Non-modifier extensions could be safely ignored; modifier extensions could not.
... That meant that you have to check for modifier extensions before you process.
... They need to be hidden from someone who will interpret it monotonically.

harold: This discussion is for Well Behaved non-nodifier (monotonic) extensions
... One option is the ideal translation. An alternative is to use the FHIR extension wrapper on the outside, but also use the preferred representation inside.
... This requires two kinds of typing, but gives consistent typing.
... Another alternative 2, is to have a first-class predicate as the tag.
... My hope is that we can come out with the ideal, for non-monotonic extensions.
... But this leads to an interesting condition when discussing open/closed validation.
... If this form is used, then none of the shex forms can be closed.
... But a validator could close it to the extent of only allowing unrecognized predicates whose object is of type fhir:Extension.
... We'd have to do some work to make this work in shex.
... Also, for every fhir type, we need to allow an optional type arc.
... That seems reasonable.
... We could declare the range of fhir;Extension.valueIdentifier to be Identifier, but we're not doing that now.
... Should we stick with literal RDF translation? It requires less explaining. Disadvantage is that it is much harder in SPARQL to use.
... Migrating from extension to core is non-trivial this way.
... Doing it the suggested way allows a more uniform way to treat extensions and core resources.

michael: Trying to figure out how this affects the generation of instances

harold: This is an instance from a FHIR example.

<hsolbrig> diagnosticreport-examples-general.ttl

harold: slides 10-13 discuss modifiers
... The fhir:Quantity.comparator [ fhir:value ",>" ]; non-monotonically changes the meaning of the value.

ISSUE: How to handle modifiers?

<trackbot> Created ISSUE-38 - How to handle modifiers?. Please complete additional details at <http://www.w3.org/2014/HCLS/track/issues/38/edit>.

Next steps

Harold: One possibility would be to post these questions on zulip

dbooth: Hoping to schedule a discussion of FHIR vocabulary and schema.org next week, 5pm ET.

ADJOURNED

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

21 Jun 2016

Attendees

Contents

Remaining items that need to be implemented in FHIR RDF

Next steps

Summary of Action Items

Summary of Resolutions

Scribe.perl diagnostic output