| OHDSI Home | Forums | Wiki | Github |
Hey CDM pros ( don’t know how this topic appeared in Vocabulary Users 
):
I peacefully live my life thinking that in OBSERVATION table we can only store concepts with domain Observation. But a couple of days ago I was pointed out that CDM spec says: “Valid Observation Concepts are not enforced to be from any domain.”
How does it make sense and why wouldn’t I store, say, conditions there?
In Colorado we have two different reasons for storing data in the Observation table with a domain_id not equal to observation.
#1 A Person’s past medical history codes may have a domain_id = Condition, but are stored in the value_as_concept_id field with the observation_concept_id equal to past medical history since these are “Observations” of something that occurred in the past and the observation data doesn’t have an exact date & associated attributes.
#2 Tobacco concepts (and other “social history” data) have domain_ids = to Device, Measurement, Condition, and Observation. In order to keep things consistent for our query writers and the recipients of our data sets, we ETL all source tobacco concepts to the Observation domain. I brought this up as a Themis issue last year, I should check up on it…
We should probably tighten this rule to only allow “Observation” concepts in observation_concept_id. That does not apply to value_as_concept_id, @MPhilofsky. We need to fix smoking, pathology procedures, all those guys.
I agree that the current rule may be way too broad and tightening would make sense. I guess you need to file an issue to the CDM WG about this to get this convention reworded.
Yes, @Christian_Reich, let’'s re-domain (is that even a word?) smoking, pathology, pain, etc.
@Vojtech_Huser - Should this go to Themis first so we can bat it around and create a strong proposal before bringing it to the CDM WG?
Yes, ma’am. And I think it is in Themis, just sitting idle for somebody to pitch in. I know that @aostropolets has a strawman for smoking. Will need to close it out.