OHDSI Home | Forums | Wiki | Github

Error in ICD10 mapping (Vertebrobasilar artery syndrome)

In source data there is data in ICD9 and ICD10. In particular, two different codes are used for Vertebrobasilar artery syndrome
ICD9 435.3 Vertebrobasilar artery syndrome
ICD10 G45.0 Vertebro-basilar artery syndrom

ICD9 (435.3) mapping is correct, as it maps to Snomed Vertebrobasilar artery syndrome, however, ICD10 mapping (G45.0) seems to be wrong, as it maps to Basilar artery syndrome. UMLS mapping from ICD10 to snomed seems to be correct (for example, ebi website shows correct mapping)

When detecting something like this (apart from reporting it), it is ok to create a source_to_concept_map with the correct mapping to be used in the ETL while a new version of the vocabulary is released or just the nonspecific mapping should be used assuming that it will be corrected in the foreseable future?

@divakhnenko please take a look and include this mapping fix in the upcoming ICD10 (ICD10CM) release.

@Diego_Bosca_Tomas do you normally use source_to_concept_map or 2 billion concepts+concept_relationship to store custom mappings?
If you use source_to_concept_map, you can put correct relationship there, but also you need to delete the erroneus mapping pair from the CONCEPT_RELATIONSHIP table, otherwise you end up with 2 mappings and 2 rows in CONDITION_OCCURRENCE.

I was thinking in a simple mappings and ‚Äėnew‚Äô vocabulary, prioritizing that over the existing one
(something like 'if code is in new vocabulary apply mapping, else use existing icd10 mapping)

Ah, indeed, you can just put this rule in your ETL that Source_to_concept_map beats CONCEPT_RELATIONSHIP if there‚Äôs the same source_concept and relatiosnship_id =‚ÄėMaps to‚Äô.

1 Like

Hello @Diego_Bosca_Tomas!
The reason why the same-named concepts from ICD9 and ICD10 have different mapping is that mapping for ICD9 was chosen based on the full name similarity and for ICD10 - based on Approximate Synonym Basilar artery syndrome. However, neither SNOMED’s Vertebrobasilar artery syndrome nor Basilar artery syndrome can be the best choice for mapping due to the weak SNOMED hierarchy. We suggest mapping Vertebrobasilar artery syndrome to Vertebrobasilar territory transient ischemic attack.
This mapping will be fixed in the upcoming ICD9/ICD10 release.

1 Like

But this is a UMLS mapping, is UMLS mapping wrong?

@Diego_Bosca_Tomas, we cannot say that UMLS mapping is wrong :slight_smile:, but, as Daryna said, it is not the best choice, especially for researchers. In the figure below you may see that ‚ÄúVertebrobasilar territory transient ischemic attack‚ÄĚ 1) includes ‚ÄúBasilar artery syndrome‚ÄĚ, which is too granular semantically, and 2) has a more clinically significant hierarchy than ‚ÄúVertebrobasilar artery syndrome‚ÄĚ. In an ideal SNOMED world, th–Ķ latter should be a child of ‚ÄúVertebrobasilar territory transient ischemic attack‚ÄĚ. But we are where we are.

1 Like

I see, thank you both!

I am pleased to inform you that this problem is solved for ICD10 and ICD10CM vocabs.
See the latest release