OHDSI Home | Forums | Wiki | Github

Mapping in Usagui: What vocabularies (with standard codes) to donwload from Athena?

We are mapping national vocabularies to standard OMOP-codes using usagi.
We first downloaded all the possible voabularies from Athena.
However, there are some vocabularies where the large majority of the codes are non-standard and few are standard.

Is there any recomendation to just download a handful of vocabularies from Athena?
If so, which ones ?

More specifically:
Given a local code, is it better to map it to a vocabulary with-more-standard-codes but less preccision on the match OR a vocabulary with-few-standard-codes but where the match is more precise ?

What are the guide lines/prefferences in the community ??
(may be @MPhilofsky, @Christian_Reich )

(this was a similar question but deviated to much on the specific example of procedures
Mapping a new vocabulary: is better to map to snomed or to other standard vocabularies? )

Hello @Javier,

Yes, since we map local codes to standard concept_ids, you only need the vocabularies containing standard concept_ids. @Christian_Reich or @Alexdavv can probably point you to a list of vocabularies which contain standard concept_ids.

Map to the closest match to maintain the most accurate representation.

The guidelines are as follows:

  1. Map to a standard concept_id
  2. Map to the correct domain
    a. morphine has a standard concept_id in the drug domain
    b. morphine also has a standard concept_id in the meas value domain

It’s been a while since I used Usagi, @Christian_Reich Did I miss anything?

On the March 15th OHDSI community call, I will be presenting on Vocabularies: Mapping, Usagi, and a general overview of the OMOP Vocabs. Please join, @Javier

To add to Melanie’s great answer, in your case you probably only need RxNorm and the RxNorm extension as that is where all standard drug concepts will be from.

-edit- there are a few exceptions, as you can see in Athena by filtering on the Drug domain and Concept Standard.


Are you all good?

We don’t have “Standard Vocabularies”. We only have standard concepts. Some vocabularies donate more, others less (e.g. the proprietary donate none). What’s the problem with feeding Usagi?

Yes, @Christian_Reich , I understand that. I just didnt know the words to express it correctly.

Let me rephrase:

A concrete example for mapping one local code:

I have a local code I want to match to an OMOP standard code.
On one hand, there is a perfect match in vocabulary X, but vocabulary X donates very few standard codes. Lets says it has 4 standar codes and 2000 non-standard codes.
On the other hand, there is less precise match in vocabulary Y, but vocabulary Y donates most of its codes to standard codes. Lets says it has 4000 standar codes and 2 non-standard codes.

Is it worth to donwload the whole vocabulary X ?

Same question generalised to map N local codes:

I use usagi to automaticaly match N local codes to a standard concepts.
Is it worth to download all the vocabularies that contain at least one standar code ? or is it more clever to download only few vocabularies that have a proportion of standar codes.

If I understand right the answer is:

Yes, it is better to download all the vocabularies that contain at least one standar code and map the local codes to the most precise possible standard code.

1 Like

Yes, you want to represent your source data as accurately as possible. One exception, if you don’t have Cancer data, then you don’t need to download one of the “cancer only” vocabularies such as NAACCR