OHDSI Home | Forums | Wiki | Github

CONCEPT_SYNONYM has Invalid and nonstandard concepts present


(Chetan) #1

Hi @Christian_Reich, @Chris_Knoll,
These questions are related to CONCEPT and CONCEPT_SYNONYM.

First Question -
As per the PDF for OMOP CDM v5.3 , one of the conventions for CONCEPT_SYNONYM is as below :

But we did find that invalid and non-standard concepts are present in CONCEPT_SYNONYM table.

Below screenshots have few such examples -


Let’s check for concept_id - 45066624.
This is present in CONCEPT_SYNONYM-


This is also present in CONCEPT, but as a non-standard concept -

Second Question -

Standard and valid Concept_ID is missing in CONCEPT_SYNONYM table even though it is present in CONCEPT table.
Below screenshot of CONCEPT table -

The same is missing in CONCEPT_SYNONYM table -


P.S.- All data for vocabularies is downloaded from Athena.

Thanks in advance.

(Christian Reich) #2

We are not purging the CONCEPT_SYNONYM table. If a concept has synonyms, and then becomes invalid - so be it. Doesn’t hurt.

Don’t mix up non-standard with invalid. A concept can be valid, but non-standard (and hence have all the synonyms). Non-standard means that it won’t be in a *_concept_id field of a domain table. Instead, it’s equivalent standard concept one does. The CONCEPT_RELATIONSHIP table will give you the mapping through relationship_id=‘Maps to’.

Not all concepts have synonyms. If they don’t there won’t be a record in the CONCEPT_SYNONYM table.

Bottom line: This table is for convenience and for helping search engines. It is also useful for localization of tools to other languages. But for OHDSI analytic use cases what counts is the semantic content in the Concept, no matter whether you call it “Myocardial Infarction” or “Heart attack”.

(Gerald Pulver) #3

So for every non-standard but valid Concept_ID there is there is a corresponding row in CONCEPT_RELATIONSHIP with a “Maps to” relationship to a valid standard concept_ID?

(Dmytry Dymshyts) #4

Not every non-standard concept has “Maps to” to a standard one.
You see, standard concept means that you can use it in the CDM.
But some concepts are valid, but can’t be used in the CDM, for example Dose Forms of Drugs. For example, a patient can’t have presciption of “Oral Table” only. There should be exact drug concept, let’s say, “Aspirin Oral tablet”, which is standard.
But if the concept describes entity used in CDM, then it should have “Maps to” relationship. Sometimes they don’t because of the vocabulary lifecycle, for example you can notice such NDC drugs that don’t have mappings to RxNorm, but this is being fixed now.

(Christian Reich) #5


As @Dymshyts said.

Bottom line: We don’t have a mapping for all source codes you find in your data yet, but it is close. What his team has put together here is pretty awesome, both in coverage as well as quality. Nobody else has that.

But we need help from the community to find the holes and mistakes.

(Gerald Pulver) #6

Thank you @Dymshyts and @Christian_Reich!
As I am now able focus my attention on OMOP, much is becoming clear to me.

If you have a chance, could you please take a look at a topic that I posted on the Vocabulary group?

I will be sure to inquire should I encounter any errors or holes. Actually the question in the topic I mention above might suggest such. Though I would imagine an apparent ‘hole’ just means that no one has had time to address a particular need, given the prioritization needed.