OHDSI Home | Forums | Wiki | Github

Dealing with unmapped investigational drugs

Vocabulary Users
#1

Hi all,

I searched the forum but didn’t see the answer to this exact question.

I have been asked to explore the possibility of using a particular OMOP (EHR) database as a source of external controls. I know a lot about the criteria for selecting external controls, which in and of itself is challenging, but my problem is not with the methodology but with how the drugs were mapped. I did not perform the ETL on this data set so I am trying to untangle it.

Let’s say I want to exclude anyone who was already enrolled in a study. I noticed that many (but not all) of the unmapped drugs, where concept id = 0, have words like “IRB” or “investigational” or “placebo” in the source value, making it somewhat easier to identify these folks with a flag and then manual review.

However, I also noticed that a lot of people with the same words in their source values are mapped to a variety of codes like this: Athena (ohdsi.org)

I want to provide feedback to the data owner on how to map this in the future. Or perhaps there is nothing to be done for this particular use case?

Thank you in advance!

#2

Hello @DanielleBoyce, as you’ve correctly identified, the way of dealing with such drugs may vary between database owners depending on the use case. One suggestion here may be to default to mapping such drugs to 0. This way, the event domain is not changed (as in the example with the code you attached); also, this approach will probably prove more consistent, since there will be no need to choose an arbitrary target.
I hope that helped with your question.

1 Like
#3

Hello @DanielleBoyce,

Assuming this EHR database utilized the vocabulary tables found in Athena, particularly Concept and Concept Relationship, the HCPCS concept you linked comes from structured, billing data. And most likely the source values mapped to concept_id = 0 come from medication tables and are not coded because there isn’t a NDC or RxNorm attached to the record. And by the source values you cite:

They can’t be mapped. And assuming my assumptions about these data are true, there isn’t anything more to be done.

Reading this:

I don’t think searching drug source values would be the best way to identify persons already enrolled in a study. What does “IRB” mean in the context of a source drug record?

#4

Thank you very much!

#5

Hi, Thanks for responding. This is very helpful. Going into this, my assumption was that the “IRB” source values in question would be mapped to zero because there is no corresponding code in RxNorm or NDC, which makes perfect sense. But then I found some of the same types of sources mapped to codes other than 0 and didn’t know what to think.

The “IRB” is source typically includes an IRB number and then some indication that there was an investigational drug administered, like “Investigational drug, IRB#12345, XYZ company.” I noticed the same source format in the concept =0 as concept = something else, but often not a specific drug, which made me think that the site enters investigational drugs in a standard format into their EHR but maps the concept differently for some reason.

t