OHDSI Home | Forums | Wiki | Github

Mapping to ICD10AM by using and existing mapping of ICD10 to SNOMED

I need to create a vocabulary mapping between ICD10AM and SNOMED. I have a downloaded a mapping table from this link https://www.ihacpa.gov.au/resources/icd-10-am-and-achi-mapping-tables. I used the twelfth edition. I currently have in the concept table and concept_relationship tables in our database a mapping between ICD10 and SNOMED. I am trying to use the ICD10 to map the SNOMED codes to the ICD10AM using the mapping tables I downloaded. The issues I found was that some terms have one to many or many to many relationships in the downloaded tables and when I use SQL code to link the downloaded mapping tables to the existing ICD10 to SNOMED duplicated entries are created in the concept table and duplicated relationships. So for instance Z84.8 ICD10AM code with description “Family history of other specified conditions” Maps to Z84.8 ICD10 Code and also maps to Z81.1, Z81.2, Z81.3, Z81.4 and in this mapping has a slightly different code_name -“Family history of disorder of other specified conditions”. The result is that there are 2 entries in the concept table ICD10AM for Z84.8 with 2 different names and in the concept_relationship table there are several mappings created for each of the concept_ids. So when we look to map those in the conditions_occurrence table there will be several condition records created for one condition. How would you suggest to do this mapping? There is also complications when it comes to mapping additive and partial codes how would you suggest to map those as well? I know some of you have experience in mapping ICD10CN to SNOMED using ICD10 how did you deal with those issues when you created that mapping.

Hi, @gruber

I think your workflow has some severe limitations.

First, as I understood from your post, you are trying to map ICD10AM codes to SNOMED. For this task, you are not supposed to use any mapping tables from SNOMED to ICD10 (any version). There are reasons why all authoring organizations provide mappings from ontology A to ontology B, and from ontology B to ontology A separately. So you need to pick the right direction and go from the ICD site to SNOMED.

Second, the longer the way between ontologies (in your case ICD10AM to ICD10 to SNOMED), the more mapping discrepancies you collect on the way. So please make sure that you review everything and I propose using the mappings you get this way only as mapping candidates - to review and accept them if you are lucky, or to fix them if you are not.

Third, you don’t need to put the whole mapping chain into the concept_relationship table. You only need your source and target concepts. This approach should remove duplicates.

This is up to you to fix. So you need to pick the correct targets for your concepts.

Also, we plan to overhaul ICD10 family in the upcoming release, please follow Upcoming changes, release notes and OHDSI forum to stay updated.

Hi @zhuk,

I am a researcher from Australia. I notice for the upcoming releases you mentioned above back in Sep last year, the listed ICM code does not include ICM10-AM, which is used in Australia. We are in the middle of mapping our EHR into OMOP. Is there an official way to request for the inclusion? If yes, please kindly share with us so our team here can kickoff the process. If not, any suggestion what we can do? Any help would be greatly appreciated.

Best Regards.

Hi, @clau
Welcome to OHDSI!

Unfortunately, ICD10AM is not on the Roadmap at the moment. That’s the bad news. The reason for this is the limited resources of the OHDSI vocabulary team. However, we would appreciate your help, which leads to the good news.

The good news is that you can become a “Vocabulary steward” yourself through the mechanism of Community contribution. Your case is adding and refreshing complex non-drug vocabularies. It requires certain knowledge, skills and dedication and is also prioritised by the Committee (the current agreement is one complex contribution per release, and one for the August release has already been chosen).

So in your case, it is possible to use custom mapping tables and create links between ICD10AM and SNOMED internally, as I suggested in the original reply. I believe you can utilise some mappings in UMLS / NCI Methathesauruses. Please check all the links, since some mappings are not the best matches in our opinion.

You can also create an issue here with the vocabulary request so we know what the needs of the community are, but realistically it is not happening soon unless we get a boost in resources.

On top of what Oleg said we have Australian chapter collaborators who may be willing to share their mappings with the community (tagging @nicolepratt or @Nicole_Pratt, @gkenno; Jitendra Jonnagaddala also has been super helpful in past but can’t find her on Forums).

I’d start with asking them. If no luck, you can reuse mappings, just make sure the names of the terms in ICD match - sometimes extensions have different (significantly different!) meanings for codes. That will give you something but not the full coverage.

Oh, right, and I wonder if @gruber would consider benefiting the community by sharing their mappings :slight_smile:

And the effective way of doing this is to add this vocabulary to the whole ICD10 family machinery where semantics and mapping reuse/conflicts are handled properly by the common data environment (CDE). The vocabulary team drives the machine but we’re limited in resources and aimed to the roadmap, so let’s talk and see how this can be supported elsewhere.

1 Like

Thanks! @zhuk! I shall talk to my research team and go from there!

Thanks @aostropolets, I will get in touch with the Australian chapter collaborators! Sorry, I was super tied up with academic activities.