Ethnicity - UK SNOMED Categories

[brianmackenna]

I have drafted a new SNOMED UK ethnicity category codelist here

Why do we need another ethnicity codelist?

The current gold standard OpenSAFELY codelist for ethnicity is here. It is a CTV3 codelist with categories for broad (6) and detailed (16) groupings.

More recently for federated analysis across EMIS and TPP we have been using the PRIMIS ethnicity codes. These are set up to foreground 2001 in the main codelist here. This is a definite choice as the NHS standard for ethnicity in the data dictionary remains 2001 (we are unsure why). Therefore anyone using this codelist misses a fair chunk of codes e.g. any related to the 2011 standard.

PRMIS have another codelist Any other ethnicity code that does include the 2011 standards but this has two issues

1. It requires someone to know that you need multiple codelists to construct ethnicty in snomed unlike in ctv3
2. It does not have categories meaning people with a 2011 code are unhelpfully lumped into the "other" category

If people agree this is enough justification I will need to add categorisation to the draft list above. @inglesp I would be grateful on advice on how to do this technically for two reasons a) you seem to have done it on the main primis list so maybe you have code to do it? b) if I do it it will be downloaded to excel/sheets but risk is that SNOMED codes will be truncated before uploading again. (related to this issue in OpenCodelists-core )

[alexwalkerepi]

Definitely sounds like a more complete list is needed.

Is this new codelist a combination of the Primis 2001 and 2011 lists, or created anew? Is there any sort of cross checking we could do with the CTV3 list?

[brianmackenna]

I created this using the OpenCodelists builder and hitting the hierarchy at the appropriate point which I worked out from the visual representation on the shrimp browser.

Cross checking will definitely be required and I will do when I work out a plan with @inglesp on best way to categorise technically

[inglesp]

@brianmackenna have you seen https://www.opencodelists.org/codelist/primis-covid19-vacc-uptake/eth2001/v1/? It was built by combining the separate PRIMIS ethnicity codelists.

[brianmackenna]

Yes - it only has 2001 codes in there so we miss all 2011 census codes and any others, it's one of the linked codelists above.

[inglesp]

Ah yes sorry.

So to categorise, I'd suggest downloading the codelist as a CSV, importing into a Google Sheet, and categorising manually. We can then (a) validate the 2001 categorisations against the PRIMIS eth2001 codelist, and (b) manually check them.

Have you seen there are other ethnicity codes, not under Ethnic group finding (397731000)? For instance, there's a whole bunch under Ethnic group, and there are 9996 occurrences of Afro-Caribbean in EMIS.

[brianmackenna]

I hadn't! Right so we need a two step process.

1. Agreed list of codes
2. categorise the agreed list

[brianmackenna]

On 1. This is quite a big job and because SNOMED is comprehensive there are codes classed as ethnic groupings that verge into nationality. I will convene short discussion with @rohinimathur and others to decide best way of tackling this!

[rohinimathur]

Hello all,
I have made a trimmed down SNOMED Ethnicity codelist- this essentially includes all ethnicity codes which ever appear in TPP. SNOMED codes with a zero count are excluded from the codelist.

As suggested by @inglesp I have manually categorized the codes into 16 categories and am requesting review of these categorisations from @brianmackenna and others before final sign off.

The codelist is here
https://www.opencodelists.org/codelist/opensafely/snomed-ethnicity/4942d89f/

The categorizations are in the attached CSV. There is also a column which shows the number of times the code appears in TPP.
snomed_ethnicity_counts_categorized.csv

[brianmackenna]

Great work! I will review the categorisation later this week but just two immediate thoughts

1. D/w @rohinimathur and decided to be over inclusive and include all terms even if they verge on to nationality.
2. I don't think it is a good idea to exclude codes with zero counts in TPP from the codelists. We know from our previous work that these snomed codelists will be used in other backends/systems and we also know that small tweaks in how they are implemented on the clinical front end can result in quite different coding behaviour. Therefore for completeness we should include (obviously low number results should be suppressed)

[rohinimathur]

Thanks @brianmackenna - i've updated the codelist to include all TPP codes with zero counts - this should help with future proofing the codelist.
https://www.opencodelists.org/codelist/user/rohini-mathur/ethnicity-2021/7282cb98/#full-list

[rc16]

Hi - just coming back to this issue as I'd like to use the SNOMED ethnicity categories.

I have a csv file with the categories from Rohini but I'm not able to add it to opencodelists. When I upload it the categories are not included. The solution I'd seen here opensafely/documentation#442 from Peter Inglesby doesn't work for me. When I go to the page at the start, ending with add, it just comes to an error page.

[rc16]

Hi,

Caroline has added the categories to the ethnicity codelist: https://www.opencodelists.org/codelist/opensafely/ethnicity-with-categories-snomed/5eb3345b/

It just needs signing off @rohinimathur

[andrewscolm]

Working with @rohinimathur and @alexwalkerepi we expanded the SNOMED ethnicity codelist , created by @brianmackenna and @rohinimathur, identifying 657 ethnicity SNOMED codes. By comparing the codelist to the ctv3 codelist, the primis codelist, and a codelist from GDPPR we grouped the codelist according to the 2001 census.

https://www.opencodelists.org/codelist/user/candrews/full_ethnicity_coded/6281e42c/

[andrewscolm]

The NHS classification is based on 2001 Census classifications and does not capture information on any of the GRT groups separately. This issue is discussed in https://web.archive.org/web/20220213182343/https://www.ethnicity-facts-figures.service.gov.uk/summaries/gypsy-roma-irish-traveller. Based on the gov.uk summary above we have decided to categorise Irish Travellers as White Other.

[andrewscolm]

Quite a few of the codes in the SNOMED list have descriptions that are nationalities like "English", "Czech", "Poles", "Egyptian". After discussion it was decided to take a pragmatic / consistent approach and categorise these according to region (e.g. White British, Other White, Other White, and African respectively for the above).

[andrewscolm]

Published version with categories that don't have ethnicity information (e.g. "Ethnic category - 2001 census") removed.

https://www.opencodelists.org/codelist/opensafely/ethnicity-snomed-0removed/2e641f61/