OHDSI Home | Forums | Wiki | Github

Fact_relationship domain_concept_id

CDM Builders
#1

Per spec domain_concept_id are integers, and are FK to the “Concept” table. I see that the example uses “Person” as values, which it isn’t really the case.
In other post, it says "These concepts have domain_id=“Metadata” (because they basically refer to CDM’s own structure) and vocabulary_id=“Domain” "
(The result should be this, I leave it for future reference)
My issue with this, is that Person (with id 56) is non-standard code, which seems strange for a _concept_id code.
Then we have codes for the omop tables themselves such as Person (with id 1147314) which is standard

Should the former codes be used even if they are not standard? Should documentation be improved to point to the actual values?

FACT_RELATIONSHIP Table - Domain Mappings
#2

I am also confused about domain concepts. When the “other post” was written they were still standard, but apparently became non-standard sometime early 2020.
Thanks for raising the question, would be great to have this clarified.

1 Like
#3

Hi all!
I am also wondering if we can still use the non-standard Domain concepts (vocabulary_id = ‘Domain’, domain_id = ‘Metadata’). ?

As mentioned by @rookie_crewkie , these concepts were standard before. The only alternative that I see, it would be standard concepts at the field level (vocabulary_id = ‘CDM’, concept_class_id = ‘Field’ ). Since, if I need to create a relationship between procedure and multiple modifiers, it would allow to indicate that the 2nd fact is a modifier: domain_concept_id_1 = procedure_occurrence.procedure_concept_id and domain_concept_id_2 = procedure_occurrence.modifier_concept_id.

#4

Domain_concept_id_1 and domain_concept_id_2 fields are the indications on the OMOP tables, where you’re gonna find the respective fact_id_1 and fact_id_2 records.
Nobody cares about the Domains and they’re not strictly assigned to the tables.

So yes, these are the correct concepts.

And yes, we need to improve the documentation in the specs.

@clairblacketer @Christian_Reich Can we do it for both 5.3.1 and 5.4?

#5

BTW, are we ready to do such a thing for 5.4?

#6

Good point. Let’s show @clairblacketer, since she has her new R (why R???) auto DDL constructor.

t