*TESTS* Total Cholesterol tests

[richiecroker]

https://codelists.opensafely.org/codelist/user/richard-croker/cholesterol-tests/09896c09/ For review/discusson: Do we want to include level only, and remove measurement?

[brianmackenna]

Brief discussion between @inglesp @brianmackenna and @richiecroker . We are unsure what to include in a codelist for cholesterol e.g.; will a cholesterol test be recorded using a finding or an observable entity

We need to further work to understand a) what is the expected deployment of these codes as per NHS data dictionary and other standards. b) what is the actual deployment and usage of these codes

@inglesp is going to produce an analysis of these codes in OpenSAFELY-EMIS to help inform best measure of service restoration. Useful things we thought about including

• Prevalence of each code
• Amount of codes with a value attached e.g. cholesterol 5.5
• Range of values observed
• unit fields?

[HelenCEBM]

Counts for EMIS and TPP here

[richiecroker]

Thanks @HelenCEBM this is very helpful - we should do with the rest of the proposed sentinel tests. The question for me is whether the 32k plasma cholesterols is all from one lab/area or spread around? Is there any way we can find out? Just looking at openpath data (for slightly different timescale), and NDHT did 42k total cholesterol tests in 12 months, so it's not out of the realms of possibility that it's a single lab using odd codes (or actually doing plasma).

[richiecroker]

I also wonder if cholesterol level tests are from patients self-reported tests.

[richiecroker]

(I mean, it's all <1%, but still interesting)

[HelenCEBM]

The question for me is whether the 32k plasma cholesterols is all from one lab/area or spread around? Is there any way we can find out?

This would be easier to do in the next step via study definition

[richiecroker]

@brianmackenna a complementary codelist for numerical values https://codelists.opensafely.org/codelist/user/richard-croker/cholesterol-tests-numerical-value/7dee4fa4/

[brianmackenna]

Included codes look good

850981000000101 : I'm still not sure about excluding this code as yet - Although low in comparitively there are 1000s of blood tests being processed through this way with numerical results.

At a minimum I think it needs to be explcitly documented in the methodology section that it has been omitted and we can maybe revisit for specific use cases?

[richiecroker]

that's true, but only 27% of them (3000) have numerical results, which suggests that it's not used consistently for numerical purposes

[brianmackenna]

I think it will depend on the use case. I would favour including and explicitly documenting this but happy to have it left out but we must document explicitly in the methodology so it's obvious an a judgement call can be made for specific use case

(happy with the first list)