Peter G. Williams

Rob Hausam

Recording, notes & attendance.

Note: No meeting on Tuesday 24 July. Was due to be TS, postponed to 31 July ie skipping a week, but keeping turn-about fair.

SNOMED on FHIR meeting planned during the business meeting - closed session (observers space dependent).

• Jane Millar to arrange session and room with events team.

Jane Millar

Call for abstract - presentation on the work of this group (2 presenters max).

• Rob Hausam to work with Jane.
• Jeremy Rogers also planning to present.
  

Peter G. Williams

Rob Hausam

• 2018-06-19 - SNOMED on FHIR Meeting (TB)

Dion McMurtrie

Summary: Expressions filter on CodeSystem Resource - Dion asked Graham about origin of these two items. GG clarified true = permit PCE eg for use in validate-code and similarly for expand. Default = true also. Suggested that expand call should then return every possible post coordinated expression (!!) which is a) hard and b) probably not useful. Such expressions could be available if an expression library had been implemented. However, validate code should handle arbitrary PCEs since this will be a finite set. Note that people do post coordination for other code systems eg UCUM and MIME.

Update: Difference between two positions - when expanding value sets defined intensionally, expectation that any existing pre-coordinated concept or PCE would be included. Graham expects server to return "Too Costly" as logical behaviour would be to return every possible PCE.

Suggestions:

• Possible valid use case in calculating lateralized expressions (even at runtime).
• Principle could be applied that in general SNOMED Pre and Post coordinated expressions should be interchangeable.
• Incomplete response should be labelled as such.
• That we should return "published" content, which might include post coordinated content in some cases. ML suggested "at least" this content.
• We could make a statement in the "Using SNOMED with FHIR" page. Clarify with the HL7 Vocabulary group.

Questions / Discussion

• In the "membership" query, does a PCE that is equivalent to a pre-coordinated that is part of the value set match? RH: If the value set says that PCE is included, then yes it should handle that.

Update 26 June

• PCE validation would be expected to take into account MRCM rules.
• PCE would not necessarily have a term associated with it - arguably useless in an EHR.
• No know use cases for expression libraries (JC)

PJ: Could change excludePostCoordinated to includePostCoordinated (and default to false) in operation-valueset-expand.html to better reflect the current capabilities of 99% of systems. Option for finer grained enumeration ("PostCoordination" / "Composition Behaviour"?)for varying efforts in PCE generation. RH: Remember these changes would apply to all code systems.

Suggested Enumeration:

• None (returning a valuset in this case would indicate that it's a subset)
• Pre-defined (eg library) PCEs - this would be the default setting.
• Generated PCEs eg Laterality

• Dion McMurtrie to capture this discussion into a page, for review by the rest of the group followed by progression into a tracker item. Target 10 July for hand over to Rob.

Dion McMurtrie

Dion McMurtrie

Dion McMurtrie

• SNOMED CT "Universal" Edition
  • Definitely useful to have a catalogue of all concepts ever issued if not a proper edition which may be much harder?
  • Is this work for this group, or should this be handed on to a more appropriate SI group?
• Check and fix ECL in the FHIR spec
  • Is there a way to systematically identify all ECL expressions in the FHIR spec?
  • If we run these through validation is there appetite/volunteers to review and fix them as feedback to the spec?
  • Is there a way to integrate ECL validation into the FHIR spec tooling to prevent recurrence?

• Rob Hausam to progress Michael's tracker item on this issue.
  

• Normal form and normal form terse definition

• Rob Hausam to progress returning NormalFormTerse to the page.

• Linda Bird to progress defining Terse Normal Form with the Family of Languages group (Rob Hausam confirms needed for July)
  
  • SLPG agreed that the Languages group is not the appropriate forum for this definition. Instead, we believe this should be defined in the DL subgroup of the Modelling Advisory Group.
  • Also agreed that the terms that need defining are "Canonical Normal Form" (which is terse by definition) and "Necessary Long Normal Form"

Dion McMurtrie

All participants are invited to review this local copy of that page.

• Section 4.2.1.0.5 suggestion that we terse Normal Form properties (Peter J disagreed). Clarity needed on which Normal Form is being represented (eg breaking sufficiently defined concepts down to their primitive components, unlike what is supplied in the browser). Further discussion needed on what these properties are being used for. Perhaps only 1 is necessary since terms can be added/removed as required
  
• Supplements are a possible way to allow language reference set type functionality.
  

Dion McMurtrie

Next Meeting: Tuesday 10 July

Questions for next week: