Open ericaVoss opened 6 years ago
This is Themis #3.
In fact, that's how we should fix this problem: Precoordinate concepts if we need to say more than one thing ("Outpatient detail", and "2nd position"), even though in this case I think we don't need the Outpatient detail thing. But let's discuss.
@cgreich , there's a request to add a a feature to CIRCE related to excluding condition_type_concepts, and one idea was proposed to leverage concept set expressions (which assumes that there would be concepts in concept_ancestor for the concepts of interest).
Question: do you plan on having condition_type, drug_type, etc concepts in concept_ancestor so that we will have a hierarchy of {domain}_type concepts?
At the moment, we assume that there is no hierarchy (and infact, you wouldn't find any _type concepts in concept_ancestor), Is this assumption wrong, or is this assumption goign to change in the future if you support concept types in a hierarchy?
@cgreich @chrisknoll {domain}_type_concepts have hierarchy
I'm not seeing any of that in my version of the cdm vocabulary v5.0 01-DEC-17
.
@cgreich : is the figure presented by @gowthamrao based on concept_ancestor, or is there something else that is driving this view?
Update, we got the v5.0 09-JUN-18
version installed, and it does appear we have type concepts in concept ancestor:
SELECT c.domain_id, c.vocabulary_id, COUNT(*)
FROM dbo.concept_ancestor ca
JOIN dbo.concept c ON ca.ancestor_concept_id = c.concept_id
WHERE c.domain_id = 'Type Concept'
GROUP BY c.domain_id, c.vocabulary_id
Yields:
domain_id vocabulary_id n
Type Concept Condition Type 102
Type Concept Death Type 14
Type Concept Device Type 3
Type Concept Drug Type 15
Type Concept Meas Type 6
Type Concept Note Type 10
Type Concept Obs Period Type 6
Type Concept Observation Type 13
Type Concept Procedure Type 95
Type Concept Visit Type 48
Interesting to see so many condition types, here's the hierarchy query:
SELECT c.concept_name, count(*) AS children
FROM dbo.concept_ancestor ca
JOIN dbo.concept c ON ca.ancestor_concept_id = c.concept_id
WHERE c.domain_id = 'Type Concept' AND c.vocabulary_id = 'Condition Type'
GROUP BY c.concept_id, c.concept_name
But all these rows say there is only 1 child (indicating that there isn't actually a hierarchy to any of these condition types). So, it's a little mysterious about how we use these values from concept ancestor as of this version of the vocabulary.
Visits do seem to have some level of hierarchy in it:
concept_name children
Visit derived from encounter on claim 14
Visit derived from encounter on medical claim 10
Visit derived from encounter on medical facility claim 4
Visit derived from encounter on medical professional claim 4
Visit derived from EHR record 3
Visit derived from encounter on medical facility claim denied 1
Visit derived from encounter on medical professional claim denied 1
Visit derived from encounter on medical professional claim deferred 1
Visit derived from encounter on vision claim 1
Visit derived from encounter on dental claim 1
Visit derived from encounter on pharmacy claim 1
Visit derived from encounter on medical facility claim deferred 1
Visit derived from encounter on claim authorization 1
Visit derived from encounter on medical facility claim paid 1
Visit derived from encounter on medical professional claim paid 1
Visit derived from EHR encounter record 1
Clinical Study visit 1
Visit derived from EHR billing record 1
@chrisknoll:
Thanks for bringing this up. I was thinking "If we introduce hierarchies to Types we need to tell Chris, so he can offer it in the tools". So, yes, I think this is where this goes. Will be done before the Symposium when we release Themis conventions and changes, but not much earlier.
@gowthamrao and I would like to see a simple classification system created for the TYPE fields. Basically if I want to know all the "patient reported" types or "claims" types I should be able to use the Vocab the find those.
Maybe the CONCEPT_CLASS_ID might be good enough for this?
I see there a whole bunch of SNOMED Type Concepts, I don't know how this idea would affect them. I know for the OMOP generated concepts, this would probably work. But this is just an idea, implementation could be as the Vocab team see fits.