OHDSI Home | Forums | Wiki | Github

Could local concepts be standards?

Hi all,

I’m reviewing an OMOP database and have a question about local concepts. I found costum concepts (2 billions) designated as “standards” (indicated by ‘S’ in the standard_concept column).

I’m a bit confused by this. Could someone clarify whether it’s correct to classify custom vocabularies as “standard” concepts within the OMOP framework?

For context, these local “standard” concepts are linked to other versions of the concept through the CONCEPT_RELATIONSHIP table using “Mapped from/Maps to” relationships.

Any insights on this would be greatly appreciated!

Best regards,

One can create standard 2billion codes and, if created properly, they can be useful (eg can be seen in Atlas). The problem is that other databases won’t know what they are (which may be a challenge if you conduct studies with other ohdsi-ers).

Those concepts should be mapped to themselves in concept_relationships (not to other concepts).

3 Likes

Hi @MeriemGrs, just another point that may help you decide on such concepts - usually it is good to thoroughly check the use case for making custom concepts standard; it may not be necessary at all. Also, often having good hierarchical relationships between them is important, so if there is none, usability may suffer.

3 Likes

Thank you for the clear explanations!
To confirm my understanding, would a correct mapping for a standard custom concept look like this?

2000000001 Concept Name Local vocabulary S Mapped from 2000000001 Concept Name Local vocabulary S
2000000001 Concept Name Local vocabulary S Maps to 2000000001 Concept Name Local vocabulary S

I’ll investigate our data further with this in mind and circle back if I have any additional questions or findings.

Right, this + an inverse (Mapped from). Seeing examples would be good - it might be the case that you don’t need custom S concepts to begin with.

1 Like

Hi,

Following up on our previous conversation, we’ve summarized the relationships between concepts in our database. You can find the table below:

To clarify, our study uses surveys to gather data, and we track different questionnaire versions in the Survey Conduct table. We create a standard custom concept for each survey question and map it across all versions throughout the longitudinal study.

Importantly, different concept_id values indicate variations in question wording across versions.

I’d love to get your feedback on this mapping approach. Does it seem logical and comprehensive?

Thanks, Meriem

Hi @MeriemGrs,

It makes sense and looks logical if I got it right and you map Non-Standard to concept to Standard:

concept_id_1 concept_name_1 standard_concept_1 relationship_id concept_id_2 concept_name_2 standard_concept_2
2000000002 Please indicate how bad your acne is now null Maps to 2000000001 Please indicate how bad your acne is now S

Please, remember that your S concepts should also be mapped to themselves as discussed above:

concept_id_1 concept_name_1 standard_concept_1 relationship_id concept_id_2 concept_name_2 standard_concept_2
2000000001 Please indicate how bad your acne is now S Maps to 2000000001 Please indicate how bad your acne is now S

An additional idea for storing your data could be the population of concept_synonym table by adding the names of the questions from the previous versions to the actual one. In this case, you will not need to create relationships and multiply the same entities in the concept table from different versions if the only difference is the wording.

t