15 Jul 2013


EricP, Gökçe, Mustafa, Kerstin_Forsberg, +1.206.___.aaaa, +1.530.___.aabb, +1.858.___.aacc, michael-ISB, StuartTurner, EmoryFry


terminology mappings

ericP: In the SALUS project, Gökçe and Mustafa are opportunistically using existing terminology mappings in order to execute drug safety surveillance queries over diverse data sets.
... what other mappings are folks using?

EmoryFry: There are three projcets that I know of within DoD [US Department of Defense], including one that Conor Dowling and I are doing.

mmiller: related terms tend to get mapped to the most expensive term. hard to back out

ericP: What's the ROI for having an index of terminology mappings with corrections and fine-grained provenance (and mayby profiles of use)?
... Gökçe, how far are you through SALUS?

Gökçe: ½ (18 of 36 months)

ericP: would a mapping index have accelerated your work? have you spent 1 month on finding and coding mappings, or 17.5 months on finding and coding mappings?

Gökçe: more the latter

ericP: how different are the models used in SALUS and Dod? I noticed that SALUS uses an RDF representation of 11179

EmoryFry: we're using SKOS

Gökçe: we're also using SKOS

ACTION: EmoryFry to work with Gökçe and Mustafa to compare expressivity and model of mappings used in DoD to those used in SALUS (or get Connor to do it).

StuartTurner: many variables in quality of mappings
... so more information would be valuable

Information Model mappings

ericP: on the other side of the coin from terminology is the Information Model, a graph which connects actions like observations which use these terminology terms.
... e.g. HL7 RIM, Indivo (was XML, now RDF), OpenEHR (used in CIMI), FHIR, C-CDA
... we need to find the mappings that were most important and map between them
... those mappings are very useful in themselves, and maybe in the mapping process, we'll learn which one true ontology we'd like to use or what we'd want to invent
... HCLS invented a bit of a clinical information model before to add to TMO, but it wasn't motivated by a study of existing models and a comparison of their strenghts and weaknesses.
... Lloyd McKenzie XSLT'd MIF to OWL for ontology management purposes. that ontology also defines *the* RDF graph structure for RIM-based clinical artifacts
... [for instance, C-CDA is defined in terms of RIM so any C-CDA document has a defined RDF representation, e.g. http://www.w3.org/2013/C-CDA/IJ.xml (note sources is XML, rendering is Turtle)]
... i picked C-CDA because the patient portability requirements in ACA indicate that it will be widely available.
... another candidate is FHIR, which seems to have a lot of energy behind it now
... Josh Mandel and I have been tinkering with RDF representations of FHIR
... the values properties within a FHIR information model are defined in terms of RIM so mapping the information models permits us to map e.g. C-CDA to FHIR
... we should also look at OpenEHR. Josh and I looked at it a while ago and tried to see what RDF a clever algorithm could produce given OpenEHR instance data

Summary of Action Items

[NEW] ACTION: EmoryFry to work with Gökçe and Mustafa to compare expressivity and model of mappings used in DoD to those used in SALUS (or get Connor to do it).
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2013-07-15 12:38:23 $