Stream: terminology
Topic: SNOMED CT to ICD-10 extended map
François Macary (Oct 23 2019 at 13:16):
Has someone already implemented the SNOMED CT to ICD-10 extended map refset into ConceptMap resource?
We don't understand which elements of ConceptMap to use in order to represent the attributes mapGroup and mapPriority of the refset.
Thanks for your advices.
Grahame Grieve (Oct 23 2019 at 17:31):
where is the source for this? What is mapGroup and mapPriority?
Michael Lawley (Oct 23 2019 at 22:35):
The ConceptMap resource has very simple semantics. There is nothing in it that corresponds to the complexity of SNOMED's ICD-10 map.
If you wanted to operationalise this map in the context of FHIR, then I think you'd need to look to the FHIR Mapping Language
François Macary (Oct 24 2019 at 13:12):
More details to my question for @Grahame Grieve : SNOMED International specifies a particular kind of refset called "extended map", which is used by the SNOMED CT > ICD map. For a given source concept, candidate targets are organized in groups, with priorities defined within each group. The specification is there: https://confluence.ihtsdotools.org/display/DOCRELFMT/5.2.10+Complex+and+Extended+Map+Reference+Sets
Thanks @Michael Lawley for the reference to the FHIR mapping language. We'll look into it. Our initial intent was to implement this map as an instance of ConceptMap resource. Another possibility would be to build the extensions we need within this resource, and propose to complement this resource in R5. The resource type does not seem so far from the Snomed extended map refset specification. Any thoughts against that?
Grahame Grieve (Oct 24 2019 at 17:34):
well, after reading the documentation, this is my take:
- mapGroup = a set of ConceptMap.group.element.target
- mapPriority = implicit in the order of ConceptMap.group.element.target, but good to make it explicit with an extension
- mapRule = an extension on ConceptMap.group.element.target of type Expression (though there's no details about this is made computable)
- mapAdvice = ConceptMap.group.element.target.comment
I would not think of this as a suitable thing for StructureMap
François Macary (Oct 28 2019 at 12:15):
Thanks @Grahame Grieve . We'll follow this path. Our main issue of understanding was about mapGroup.
A concepts that maps systematically to two ICD-10 codes will have two map groups, each one containing a single target.
Example: 385513001|Firework burn of skin (disorder)| maps to T30.0 |Burn of unspecified body region, unspecified degree
)| and W39 |Exposure to inanimate mechanical forces, discharge of firework|.
Therefore, 2 occurrence of element with code 385513001, the first one with target T30.0 and the second one with target W39.
FYI, the grammar for map rules is in the technical guide for the map to ICD-10:
https://confluence.ihtsdotools.org/display/DOCICD10/Appendix+A%3A+mapRule+Grammar+and+Formatting
Michael Lawley (Oct 29 2019 at 01:28):
Note my response was mostly wrt the issue of mapRule and making it computable. ConceptMap has no mechanism for this and I'd be interested in how this might be done natively in FHIR
Grahame Grieve (Oct 29 2019 at 01:35):
well in this case you'd just reproduce the snomed language and meaning. I don't know if that would suit us.... I expect it to be too bound to snomed semantics. The grammar doesn't tell me how it works
Michael Lawley (Oct 29 2019 at 02:35):
exactly - syntax but no operational semantics, and for it to work, you need to know how it binds to the information model (ie the FHIR resources)
Grahame Grieve (Oct 29 2019 at 03:40):
yes so maybe we want to do something about, but the depends/produces probably accounts for most of the actual functionality
Last updated: Apr 12 2022 at 19:14 UTC
