Semantic Web Health Care and Life Sciences Interest Group Teleconference -- 23 Jun 2015

<trackbot> Date: 23 June 2015

<scribe> scribenick: ericP

HL7 Project Scope Statement (PSS) for computable semantic links from FHIR to RIM

<dbooth> https://lists.w3.org/Archives/Public/public-semweb-lifesci/2015Jun/0027.html

<dbooth> The PSS: http://gforge.hl7.org/gf/project/arb_fhir_rim/scmsvn/?action=browse&path=%2F%2Acheckout%2A%2Ftrunk%2FArB-computable%2520semantic%2520links%2520from%2520FHIR%2520-PSS.docx

dbooth: per Charlie Mead: we should:
... .. take a well-written text-based description of a link from a FHIR-spec/profile and map it to RDF.
... .. validate using RDF defn's of the O-RIM
... .. want to show how sem techs can aid this
... .. we want to demo cross-spec consistency, but we should start with pharmacy

Lloyd: and the intention is that we update all FHIR/RIM mappings for FHIR 2.1
... it's fine to lead with pharmacy, but we need to make this part of FHIR governance

paul: we just need to define the scripting

Tony: is this an ITS scope?

Lloyd: this isn't really about the representation of information, but ITS has expertise in this space
... i'd expect it to be a joint project with M&M

Tony: so it's not the mapping of current FHIR into RDF

Lloyd: it's been proposed to ARB but I'd like to move it out because it's not really architecture

paul: ARB started it because they felt it was important

Lloyd: I'd like FHIR/infrastructure to own the project, cosponsored by M&M and ITS

<Zakim> dbooth, you wanted to ask what these mappings would look like? How would they be written?

dbooth: what would these look like?

Lloyd: there are two potential mappings:
... .. structural elements (class code, mood code, negation indicator). not fussing with code.
... .. when defining the semantics of a resource, we can say that AllergyIntollerance corresponds to OBS EVT
... capturing that it's related to an allergy is harder
... i think it's reasonable to map at the resource property level as well
... ideally expressed in RDF, if we can make it comprehensible to the WGs

<Zakim> ericP, you wanted to talk about shex stuff (as usual)

dbooth: so this is an ontology mapping exercise

<dbooth> eric: I've been working w Claude on this front.

<dbooth> ... If we have detailed clinical models that clinicians care about, we want an ont that is intuitive to clinicians, they way they would phrase questions or rules.

<dbooth> ... If we have a mapping between these logical clinical models and FHIR, e.g., BP in FHIR will look a certain way.

<dbooth> ... It will look another way in RIM.

<dbooth> ... If we're thinking about the part of the mapping that is computationally useful for people who are trying to work with this data, because they're trying to map data from one form to another, or mapping defs from one form to another, there's a pattern of observations, pattern of things that happened

<dbooth> ... Those are the two places where people will care the most about the mappings.

<dbooth> ... The semantics of RIM are mostly ACTS and ACT relatinoships.

<dbooth> ... FHIR is more along the lines of body site, and diagnosis. So we have an opportunity to map between FHIR and RIM, or we could map between FHIR and QUICK or CIMI.

<dbooth> ... I'm trying to figure out the degree to which the work I'm doing with Claude would apply here.

<dbooth> lloyd: The objective here is not converting instances -- slim to no hope of doing that.

<dbooth> ... The purpose here is to bind semantics, doing it legitimately, and better understanding FHIR and validating FHIR, because FHIR

<dbooth> ... FHIR resources are being created by a lot of groups.

<dbooth> ... (validation)

<dbooth> ... Quality checking. Mappings of likely instances would be in the CDA and CCDA to FHIR process.

<dbooth> ... Not likely elsewhere.

<dbooth> ... Likelihood of v3 messaging to FHIR is uncommon.

<dbooth> ... So the driver of this work is to better understand the FHIR models -- not instance conversion.

-> http://www.w3.org/wiki/HCLS/ClinicalObservationsInteroperability/C-CDA#Mapping_to_FHIR mapping C-CDA to FHIR

<dbooth> ... A huge number of people will want to convert CCDA to FHIR, but few of them would understand.

<dbooth> ... So primary conversion process would be that.

Lloyd: conversion of CDA to FHIR using a jar or an XSLT is reasonable for implementors to understand

<dbooth> ... Conversion of CCDA to FHIR using a java jar, or XSLT, are reasonable tech for people to use. Whereas RDF would involve three levels of transformation (CCDA, RDF, XML)

paul: the mapping is for testing completeness
... also serves as a testing langauge, which e.g. jars and XSLT can test against

dbooth: we expect that factoring out format translations from semantic translations will be beneficial, but that remains to be proven

paul: the content of this project scope will be useful no matter what group sponsors it

dbooth: i haven't seen confirmation from cecil that he can take on the roles of project and vocabulary facilitator

Lloyd: i think this is because it's from the ARB. i'd expect graham

paul: why don't we add what [who] we'd propose

<dbooth> ACTION: David to check with Grahame about being project facilitator on PSS [recorded in http://www.w3.org/2015/06/23-hcls-minutes.html#action01]

<trackbot> 'David' is an ambiguous username. Please try a different identifier, such as family name or username (e.g., dbooth, dderour, dhansen2, dnewman, dshotton).

Lloyd: there are several resources that won't map to rim:
... .. value set
... .. conformance
... .. structure definition
... .. some extensions

<Zakim> ericP, you wanted to ask why the RIM/ITS mapping isn't sufficient

<dbooth> eric: Why were RIM ITS mappings not sufficient?

<dbooth> ... Eric a FHIR status, e.g., is quite complex. Are we expecting to do better than the XPATH that lloyd used to navigate RIM ITS?

<dbooth> paul: RIM ITS isn't in here.

<dbooth> ... Better to view it as a rootless graph of the RIM

<dbooth> eric: I understood them to be RIM ITS mapping, in that the names of the path you're traversing are the elements defined by the RIM ITS

<dbooth> ... Want to be sure that we don't overmarket.

<dbooth> ... Want to have a reasonable chance of satisfying this goal. Need to understand the alternatives, and why one that was started is not continuing.

<dbooth> paul: the initial goal was to put together some kind of mapping back to the RIM.

<dbooth> ... Decided in an eveningn to use XPATH, and started that way.

<dbooth> ... Different groups have different levels of understanding, but we don't have a way of computing them, or verifying whtehr they point to nowhere or whether they overlap.

<dbooth> ... The hope is that by doing them in RDF there may be utilities that would allow us to check that a thing is pointing to a thing, and we can check for duplicates.

paul: whether we sit with pharmacists and explain RDF or XPath or whatever, it will be obscure
... the pharmacy WG has been maintaining their "text links".
... we can use the success of the pharmacy mappings to evaluate and market the approach

<rhausam> lost my connection

<dbooth> ACTION: David to send back the revised PSS and ask whether Jan 2016 end date is realistic [recorded in http://www.w3.org/2015/06/23-hcls-minutes.html#action02]

<trackbot> 'David' is an ambiguous username. Please try a different identifier, such as family name or username (e.g., dbooth, dderour, dhansen2, dnewman, dshotton).

<dbooth> ADJOURNED

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

23 Jun 2015

Attendees

Contents

HL7 Project Scope Statement (PSS) for computable semantic links from FHIR to RIM

Summary of Action Items

Scribe.perl diagnostic output