Defining implementers versus authors of maps

[DaveraGabriel]

In N3C, the OMOP Vocab maps provided by OHDSI are implemented.

Details: Stephanie Hong has been asked to provide the N3C maps (content) to various platforms (a FHIR server, possibly also as a component of the CDMH2 project being catalogued on the caDSR). In representing the N3C map set, there is a need to reflect the OMOP Vocabulary maps from source coding systems onto the OMOP Vocabulary. These maps are intrinsic in implementing OMOP, and albeit N3C would provide these as components of an over-arching set of maps reflecting the N3C transformations, N3C is not the author of these maps.

When applying SSSOM to this maps set, we have run into an issue re: the value set when assigning the "narrower / broader" SKOS-like attributes to the relationship when the OMOP Vocabulary transforms are the source of the maps within the larger set of N3C maps. In some cases, the OMOP vocabulary intrinsic map will be semantically equivalent (esp for source vocabularies they map directly to "Standard" such as SOMED CT & LOINC) and in other cases there is a wide variance of narrow / broader (especially for classifications such as ICD*)

The team's suggestion to Stephanie was to indicate "undefined." This solution, albeit the most correct in our opinion, lacks the true expression of: "OHDSI made this map, but N3C is using it"

[matentzn]

This seems like two issues to me, correct me if I am wrong:

• Expressing provenance "OHDSI made this map, but N3C is using it", which would require a new role, "user" to accurately reflect (you could, for example, ask for a new SSSOM element user_id to reflect that use case).
• The problem that some source data (like OMOP) does not distinguish accurately between exact and broad (narrow really is a bug - that should be reported and fixed by OMOP!). In this case, if you agree, I would suggest the use of a new mapping relation: "broadOrExactMatch". However, I want to note here that this is strategically (while perhaps necessary) wrong. The main point of SSSOM is to ensure that mappings are semantically well defined. There is next to no point publishing a mapping set where the mapping relation is not well defined. I would back you up defining this relationship, but against my better judgement and against 90% of the other SSSOM developers (I assume).

Do I understand your issue correctly?