OHDSI Home | Forums | Wiki | Github

Mapping a new vocabulary: is better to map to snomed or to other standard vocabularies?


(Javier Gracia-Tabuenca) #1

We are mapping a procedures vocabulary (NOMESCO finnish version) to the OMOP.

We were mapping to SNOMED, but we now noticed that thera are other standard vocabulaires (selecting in Athena: “Standard” & domain=“procedure” & “valid” -> vocabularies = ICD10PCS, OPCS4, …).

How to porceed ?
a) Map code to the closest SNOMED even if loosing some info on the concept
b) try to map to SNOMED unless there is a more specific match in any of the other vocabularies (in that case which one to look first)
c) map to the best matching concept, it does not matther what vocabulary it is

thanks


(Eduard Korchmar (Terminologist @ Odysseus Vocabulary Team)) #2

Hello! Both b) and c) will work similarly.

Although SNOMED is preferable for procedures, there are no hard restrictions on target vocabulary. We gradually align other “standard” vocabularies like LOINC, ICD10PCS or OPCS4 to SNOMED by creating unified hierarchy based on latter. It’s a slowly ongoing process ca. since 2018, so concepts from other vocabularies can lose standard status and be remapped to SNOMED. Software in OMOP can handle these changes automatically, so if concept A is mapped to B, and B gets a mapping to C, A will also get a mapping to C.

You can use any standard concepts as targets for mappings. In case there are duplicates between standard SNOMED concepts and other vocabularies, SNOMED concepts should have preference.


(Vojtech Huser) #3

See this thread: Standardizing procedures (call for non-US based collaborators)

SNOMED CT will not introduce certain granularity (by policy). And this is a big problem for analysis. So SNOMED Extension for procedures may at some point be in the cards (or use national SNOMED CT realms).

In addition to b) and c) - my project (linked above) aims to collect all ‘too granular’ procedures that lack good target at the moment.

So please submit ‘unmappable procedures’ (ideally with titles translated to English; a problem often) to the community. (By posting on the forum). Quantify them with some usage rank so that we know which are more frequent than others. Can you please contact me by email. (for that procedure study). (vojtech.huser at nih.gov)


(Leilei Zhu (Clinical Data Standards Lead, UCLH, UK)) #4

I think it is worth clarifying that it is not “SNOMED CT doesn’t introduce granuality”, it is a balance between pre and post coordination. If all systems can handle post-coordination, it shouldnt be an issue with “granularity”.

In the UK, as we are member country for SNOMED CT, we have various extensions. We are also allowed to have Namespace for local extension. When you say “We were mapping to SNOMED” @ Javier, which “SNOMED CT” package you are referring to?


(Christian Reich) #5

@LeileifromUK:

We like pre-coordination. Because it’s clean. Only things are pre-coordinated that fit together. Otherwise, you have to put all sorts of policing together to avoid things like “Malignant neoplasm of the hair” or “Bipolar disorder of the liver”. With respect to procedures, the problem is that the atoms or attributes are not well defined, and SNOMED has a different system than ICD10, and most of the local systems have no defined atoms at all, really.

How do you guys do that?


(Javier Gracia-Tabuenca) #6

Thanks @Eduard_Korchmar, very usefull to know about the future comaptibility


(Javier Gracia-Tabuenca) #7

@Vojtech_Huser. in Finland procedure (and many other codes) are publiclaly available in often with english traslations here https://koodistopalvelu.kanta.fi/

For procedures :
https://koodistopalvelu.kanta.fi/codeserver/pages/classification-view-page.xhtml?classificationKey=57&versionKey=119

I will send you this table with the appened frequency in our data bases once I clear that there are no isues on that.


(Javier Gracia-Tabuenca) #8

Hey, @LeileifromUK.
Sorry, I meand SNOMED CT.
We are mapping to SNOMED CT as it is the standard used in OMOP.


(Eduard Korchmar (Terminologist @ Odysseus Vocabulary Team)) #9

You should be careful when mapping directly to SNOMED because of two reasons.

SNOMED Vocabulary you can find in OMOP in Athena is not strictly equivalent to SNOMED international. It also contains codes from SNOMED UK and SNOMED US editions, so if you can’t find needed mapping target in SNOMED International, it may still be available in one of local editions. Like this term that is specific to SNOMED UK Edition but is still usable as Standard concept in OMOP.

Another problem with SNOMED is that it should not be used as standard hierarchy for some concepts that should belong to Drug Domain. For instance, source concept “Typhoid vaccination” ideally should be mapped to Drug concept from CVX vocabulary and not SNOMED Procedure concept to ensure interoperability.


(Leilei Zhu (Clinical Data Standards Lead, UCLH, UK)) #10

hi Javier, SNOMED CT has two packages, international release and local subset. If something is affecting at a national level, we normally ask UK to add it into its subset.


t