OHDSI Home | Forums | Wiki | Github

Mapping ATC and MedDRA to OMOP vocabularies


(Oskar Gauffin) #1

Dear all,

I’ve been watching the OHDSI-tutorials on the OMOP vocabulary (https://www.ohdsi.org/past-events/2018-tutorials-omop-common-data-model-and-standardized-vocabularies/), and have a few questions on how to proceed. We are working on an R-package that is supposed to run some scripts on longitudinal OMOP-data, and return aggregated data for a long list (~1000) of drug-reaction-combination-pairs (coded in ATC and MedDRA). What we would like is an automated procedure to translate each of these terms, into some vocabulary we can use in OMOP-databases, i.e. standard concepts.

Drugs are coded in whodrug at source, which we translate to ATC with an in-house dictionary. As an example, I try “Influenza pandemic vaccine”, which has an ATC of J07BB. I query the concept table for the concept_id of this ATC, and find 21601333. Following the tutorial, I would now try to find the “Maps to” in the concept-relationship-table. Even though I find 40 000 matches, I only find relations of the relationship_id “ATC - RxNorm”. Is this what I want?

If I take the mappings from “ATC-RxNorm” and ask for them in the concepts-table, I find around 17 000 standard concepts. Is this the (/at least one reasonable?) way to translate the ATC of “influenza pandemic vaccine” into OMOP vocabulary, or should I include the children of these 17 000 standard concepts as well, if I want to capture all pandemic influenza vaccines?

Second part of my question relates to MedDRA. From reading other forum posts, I understand that MedDRA is “on top” of SNOMED in some way, but I don’t understand what this means in practice. Can I proceed with some possibly defect mapping or is mapping not possible at current state? I found a thread (Relation between MedDRA and SNOMED) that claims that only 39 % of the PT-terms in MedDRA are mapped to SNOMED.

When I try with a medDRA-PT-term, I don’t find “Maps to”, instead I sometimes find “MedDRA - SNOMED eq”, but also “Is a”/“Subsumes”. Which one should I use?

For the (if still accurate) 61 % of the PT:s which don’t have a mapping to a standard concept, what is recommended, should I for instance move upwards in the medDRA-hierarchy before searching for a standard concept for the non-mapped PT:s?

And if all/any part of this is documented somewhere, I’d be happy to read up on it, it’s just that I don’t know where to look for it.


(Christian Reich) #2

Welcome to the community, @OskarGauffin

Uh-oh! Bad boy. :slight_smile: A good community member would submit the need to have whodrug included in the Standardized Vocabulary and provided the mapping table, so other folks can use it. Any chance you can do that?

No. You would follow the CONCEPT_ANCESTOR table down. ATC is a drug classification which sits on top of RxNorm. You need to be in RxNorm. You may as well map whodrug to RxNorm. That would probably be a “Maps to” job. The Vocab Team would help with that.

9201 MedDRA PT have a descendant out of a total of 23,722. That’s about 40%. The ones that are not mapped fall into the categories:

  1. They are not Conditions (e.g. 788115 “Ureterolithiasis”)
  2. They are junk and cannot be mapped (i.e. 788422 “Manufacturing facilities issue”)
  3. They need mapping

The reason we have 3) is this: Refreshing the MedDRA-SNOMED mapping is on the backlog, but SNOMED and MedDRA decided to do that officially. Now we are waiting for them to produce. Looks like we shouldn’t hold our breath, really. At some point we’ll pull the plug and do it ourselves.

Again, CONCEPT_ANCESTOR. Don’t try to re-engineer what the Ancestor Constructor does.

Well, many of them you don’t need. For the rest - bring them on here. We may do a quick and dirty fix, if it is not too many.

NO WAY. Then you will sweep in all sorts of other concepts.

Lagging behind.


t