OHDSI Home | Forums | Wiki | Github

ICD10CM duplicates

Hi @Dymshyts , @Christian_Reich,
there is an issue with ICD10CM vocabulary. There are the same codes, but in lower case and upper case. Lower case codes are invalid, but for source_concept_id we don’t check invalid_reason, so there may be problems with mapping of source codes.

Some examples of such codes:
H60.8X1 - H60.8x1 (Other otitis externa, right ear)
H60.8X2 - H60.8x2 (Other otitis externa, left ear)

Thank you,
Irina

Well, the source gave us these concepts with lower case and in the next release with the upper case.
So technically that’s not a problem: we add “Maps to” relationship from deprecated concept (H60.8x1),
you use case sensitive matching (like you do for Read-codes), and anyway, get the proper mappings.

@Christian_Reich, @IYabbarova, what do you think?

Makes sense. We also may want to fix the case in the ICD10CM input, otherwise we get the entire vocabulary in there twice with Maps to between the halves.

I know the rule is to never delete concepts, but it would be nice to make an exception in this case. This problem will continue to show up in the forum every few months and has the potential to mess up an ETL, especially with databases that are not case sensitive.

1 Like

@DTorok:

We don’t have to do that. We can invalidate and remove the concept_code. There are such examples when you and Mark did a rogue release a couple years ago. :smile: LIke this one: 44823821

But I agree, the vocab ETL from the source needs to not let that happen. The team already put a Jira in.

Well, @Christian_Reich,
so you want us to do the same with those duplicates as we did for

“Duplicate of ICD9CM Concept, do not use, use replacement from CONCEPT_RELATIONSHIP table instead” ?

Another my thought was to introduce the new invalid_reason entity displaying Error in OHDSI vocabularies,
because now “D” means both - our error and concept outdated by the source

@Dymshyts:

Don’t change the model. Just kick these guys out, without losing the concept_id.

Ok, will do in the same way as was done before.

@Christian_Reich, @aostropolets
so how do we work with the out of the market drugs?
They disappear from the source, so they became deprecated and lose their relationships to RxNorm Extension. Do they?
but we need to keep this relationship as we keep “Maps to” relationship for ICD concepts.

@Dymshyts:

No, they don’t. Maps to relationship don’t get deprecated in the generic_update process when the source drops them. They keep their last mapping.

1 Like
t