OHDSI Home | Forums | Wiki | Github

OMOP standardized vocabularies and ONC

Do the OMOP standardized vocabularies conform to “Section I: Best Available Vocabulary/Code Set/Terminology Standards and Implementation Specifications”? 2015 Interoperability Standards

@MPhilofsky:

Comes close. It’s the same except:

  • Diagnoses: SNOMED instead of ICD10CM
  • RxNorm for Immunization instead of CVX (but are planning on bringing it in)
  • Medication allergies: SNOMED instead of RxNorm

Why?

@Christian_Reich

Curiosity. I’m trying to wrap my head around the transformation from source to OMOP standardized vocabularies.

Another question: Does OMOP highlight when the crosswalks aren’t direct, 1:1 relationships? Not during the ETL, but when a researcher queries the database are they notified that there is the possible loss of specificity or granularity?

@MPhilofsky:

That’s a cool idea. Currently, in vocabularies that are not strictly stratified (like drugs where the concept_class_id tells you excactly what level of granularity you got), the mapping doesn’t know whether it is “uphill” or 1:1. We should add that to the long to-do list.

But what it does do is to allow utilizing the hierarchical relationships. That’s much more powerful than what folks generally do like in ICD9CM where they cut off a character or two.

I don’t quite understand this:

“drugs where the concept_class_id tells you excactly what level of granularity you got”. Does the concept_class_id describe granularity for drug mappings?

The CDM describes concept_class_id as “The attribute or concept class of the Concept. Examples are “Clinical Drug”, “Ingredient”, “Clinical Finding” etc.”. I don’t see where the data on granularity is stored.

I’m new to this, so please excuse my confusion.

@Christian_Reich

Yes. They do. For example, Branded Drug has a lot more granularity than Ingredient. See here.

t