OHDSI Home | Forums | Wiki | Github

Mapping transfusion to procedures

General
#1

Hi,

Our team is attempting to map transfusions. We tried to follow the best practices in the OMOP 5.4 documentation and map to devices; however, we found that we were unable to map a large share of our tests. For instance, we were unable to map the “Transfusion of Albumin to a device”. Does anyone have any suggestions regarding mapping to procedures? Should we consider this option?

Many thanks,

Kieran

#2

Hello @Kieran ,

What is the source of your data? US data? Do you have Epic? Do you have ISBT codes for your transfusion data?

Cheers,
Melanie

#3

Hi @MPhilofsky ,

This is Canadian data, and more specifically Ontario data. Each hospital has its own electronic medical record system. Some are Epic, but others are different systems entirely. There is no ISBT codes for the data.

Thanks so much!

Kieran

#4

That’s seems to be a problem.
Users of US data might have a licence for the ISBT vocabulary and use it as a target (and often it is a source as well), but international users don’t have such a target vocabulary.
We probably need to create a standard vocabulary for transfusions (kind of ‘OMOP transfusions’) free of licence.
We can start collecting concepts from different sources, and based on this and some standard vocabularies (SNOMED, dm+d,etc) create our ‘OMOP Transfusions’.
@Kieran can you please share your mappings and what you can’t map. This can be a good start.

Tagging the other leads of the vocabulary team @mik @Alexdavv @Christian_Reich

1 Like
#5

Without ISBT codes it’s a little trickier. @Dymshyts & the vocabulary team will take good care of you!

#6

Hi @Dymshyts and @MPhilofsky ,

Thanks for reaching out. I’ve attached an xlsx sheet. We originally mapped our raw blood products to procedures. I left that in. The final column in the xlsx sheet is concept_description_device. This column has a list of mapped raw codes to domain devices with SNOMED vocabulary.

Does the vocabulary team strongly discourage mapping transfusion to procedures?

Thank you so much for your advice and help,

Kieran

concept_map_v2_device_for_ohdsi.xlsx (13.4 KB)

#7

You are right. We actually should have a Github, maybe not. Proprietary concepts should never be standard. @mik?

#8

concept_map_v2_device_for_ohdsi.xlsx (14.3 KB)
I briefly reviewed the file and added my comment, it’s either can be mapped to a device (like a Red Blood cells or Platelets) or is probably a drug classification (like Immunoglobulin, coagulation factors, etc), so we are not that bad with a representation of blood products as devices actually, but we need to think about device vs drugs

#9

We have had that discussion. We just need to finish the work. @mik?

#10

Hi @Dymshyts ,

Thanks so much for all of your help. This is incredibly helpful.

Kieran

#11

That discussion precedes my engagement a little, @Christian_Reich , however, in the linked discussion, the final sentence is more or less: have we settled the convention? => No!..
So, what are you asking? Finish the discussion or invent a new vocabulary? When I look at the vocabulary, we have a long list of entries for whole blood, lots of entries for red blood cells (this doesn’t look like we have a good consolidated approach), a better collection for platelets and again a long list for fresh frozen plasma.
And to your question about proprietary concepts not being standard: I think today this is common practice for various reasons, right? You cannot necessarily use them in network studies everywhere, but maybe in your particular network? Yes, it would be preferable if we would map all of these to a common standard concept (in a model like RxNorm Extension? => this would become huge), but if we aim at mapping them less specifically to; let’s say SNOMED, maybe more granular analysis within those proprietary ones would be made harder? Please provide some advice on how to finish the work!

#12

Yeah. I don’t have a solution. Let’s sit down and figure out solutions.

t