Duplicate standard concept across SNOMED and Nebraska Lexicon

[MaximMoinat]

Description
There are standard concepts in both Nebraska Lexicon and SNOMED with the same name. This makes it ambiguous what concept to use.

How to find it
One example is the concept of 'Graft versus host disease of liver', which has two standard concepts with the same name

Concept ID	Concept Name	Standard Concept	Vocabulary ID	Valid Start Date
3173966	Graft versus host disease of liver	Nebraska Lexicon	S	28-Jan-2016
3655102	Graft versus host disease of liver	SNOMED	S	31-Jul-2020

We expect here that only the SNOMED concept is standard.

Expected adjustments
Upon refresh of SNOMED, also check if a concept with the same name already exists as a standard concept. If so, make that existing concept non-standard.

Additional context
Luckily, both concepts are descendant of SNOMED 'Graft versus host disease', so at least the Nebraska concept is placed correctly in the hierarchy, mitigating issues for studies.

[aostropolets]

Thanks, Maxim. There are indeed quite a few duplicates in Nebraska Lexicon (which we also report in unique standard concept report we release with each release). We are looking for a steward to own the vocabulary and refresh it as right now it is not on the roadmap and doesn't have a community steward.

[cgreich]

We should de-standardize Nebraska Lexington entirely. It is incorporated into SNOMED and therefore no longer relevant. And stale.

[mgurley]

Echoing @cgreich

The Nebraska Lexicon was a POC worked on by Scott Campbell. This work has been rolled into SNOMED international itself, in collaboration with the College of American Pathologists (CAP). CAP/Scott Campbell have developed maps from synoptic CAP-encoded ckey data to SNOMED proper, enabling the first ability to map real world discrete cancer pathology data to an international standard. So I think all of the Nebraska Lexicon should be de-standardized. A separate question arises if that now that the Nebraska Lexicon is part of the SNOMED itself, the idea of a separate Cancer Modifier vocabulary makes sense. Should OHDSI embrace SNOMED for tumor characteristics and not try to keep rolling its own?

[aostropolets]

I'm fully on board with that. We just need a steward to run the load stage and do the mapping. It may be as straightforward as full string match, but I'm looking at the load stage and am sensing that there is more to it...

On CA vs SNOMED, I think we have a meeting with SNOMED in place to align our strategies

[MaximMoinat]

Before finding this, I was actually under the impression that Nebraska Lexicon was all non-standard. So fully supportive of the de-standardization proposed by Christian.

Btw, I quickly did a count, and of the Nebraska Lexicon (n=465,801) there are 4,187 standard concepts of which 130 have a same-named concept in SNOMED.
For reference, the query I used:

select * 
from concept as nl 
join concept as snomed
  on snomed.vocabulary_id = 'SNOMED' 
  and nl.vocabulary_id = 'Nebraska Lexicon'              
  and snomed.standard_concept = 'S' 
  and nl.standard_concept = 'S' 
  and nl.concept_name = snomed.concept_name;

[TinyRickC137]

How about we do it in February release? Should not be a big lift. I hope it does not affect a hierarchy in any way.

@MKhitrun thoughts?

[m-khitrun]

We have finalized the scope of the Feb release already. Also, I agree with @cgreich that we should strive for comprehensive solutions instead of applying a series of small band-aid fixes.

If this issue is a high priority, we can discuss including it in the scope of the summer release.