OHDSI Home | Forums | Wiki | Github

Route 'standard' concepts not standard anymore?

i am not understanding the problem with the rxnorm dose form concepts.
they are very specific, and sometimes you want to group them, but they
seem to cover the space we need and are already contained in the
vocabulary. are we talking about adding a set of classification concepts
(from hl7) that map to the rxnorm dose form concepts?

I suppose we need the “Route” concepts to fill route_concept_id in DRUG_EXPOSURE table.
And RxNorm Dose Form isn’t good for that, because that represent actual drug form, like Lotion, Cream, Oil, etc.
But we need only “Oral” as a Route. This way we aren’t satisfied with RxNorm Dose Form concepts only.
And then, @Patrick_Ryan, you are right, it would be nice to introduce these HL7 Routes and build relationships to existing Dose Forms

You might take a look at RxTerms if you have not done so. It rolls up dose forms. I used it to derive “oral”. It is also an NLM product. Apologies if this has already been investigated.

We should not be inferring the value for ‘route_concept_id’ in the
DRUG_EXPOSURE table, that should only be used if a source has verbatim data
to indicate the route. If its derived from the RxNorm concept, then
there’s no reason to store it there, because it’d be a direct look-up in
the vocabulary. Also, while sometimes you may want to aggregate to
‘Oral’, other times, you may want to specificity of the type of
administration, so it would seem reasonable to offer a classification atop
RxNorm Dose Forms, but not as a replacement.

Ok, let’s take a look on the use case.
@rtmill, can you please provide us with the use case you need Route concepts for?

I ran into the problem while trying to map HL7 Immunization records and found no standard concept for intramuscular. An example from HL7 can be found here.

In @Patrick_Ryan’s words, the source has verbatim data to indicate the route.

I had the same problem with non-standard routes (that were discussed in this topic) while making route custom mappings, so I think that we need more reliable standard concept set.

I reviewed different approaches for Route definition, and I like this:
@rtmill gave us as example, but it can be used as a separate vocabulary for Routes.
Then we will map existing routes to this list, and in the future there will be no problems.
@IYabbarova, please take a look on this list. Does it satisfy your needs?


yes, it looks good)

@Christian_Reich, need your approval for a new vocabulary implementation

What’s the proposal? Change Standard Route Concepts to another vocabulary? Don’t want to create bureaucracy, but can you summarize the change?

and also make mappings from all existing a little bit messy routes (SNOMED derived) to these new

And the old ones are mapped to the new ones?


Hi all!
I made some analysis and discovered that SNOMED, actually gives us pretty nice Routes list.
Here it is
snomed_routes.xlsx (17.2 KB)

Please, review and tell me if it satisfies your needs.

Standard Route concepts from SNOMED added, looks that they cover every existing Route.
Note, some concepts that were the standard Route changed their domain now, so you need use the new concepts:

Old concept_id concept_name -> new concept_id concept_name
45956873 Obsolete-Oromucosal other -> 4186839 Oromucosal
45956876 Obsolete-Intraventricular -> 4222259 Intraventricular cardiac
4112421 Intravenous -> 4171047 Intravenous
4115462 Rectal -> 4290759 Rectal
4120036 Inhaling -> 45956874 Inhalation
4128792 Nasal -> 4262914 Nasal
4128794 Oral -> 4132161 Oral
4136280 Intravaginal -> 4057765 Vaginal
4139962 Subcutaneous -> 4142048 Subcutaneous
4231622 Topical -> 4263689 Topical route

This is really helpful. I somehow missed this discussion. @Dymshyts is there a mapping from RxNorm Dose Forms to these standard routes? I like how SNOMED has the routes, but our data is mainly in RxNorm.


You cannot really do that. The relationship between form and route is fuzzy. So, Oral Tablet is Oral, but Solution for injection? Could be subcutaneous or intravenous or intraperitoneal.

@aostropolets probably has a pragmatic many to many (few to few really) crosswalk for her drug mapping. Just go down the hallway and ask her.

Hello! There are cases in the source data when the drug is administered as an injection and this cannot be reflected in the mapping to the clinical drug form or more detailed classes. It seems to me, this is more suitable for route. Don’t we want to make a super type of different injectable route administrations? @Christian_Reich @Dymshyts


The RxNorm Extension machine (the “boiler”) will figure this out. If you have to map your input information to Drug Concepts the input format allows you to do a fuzzy matching: “Injectable Solution”, “Pen Injector”, “Injection”, “Injectable Foam”, “Injectable Suspension” etc. You can even give an order of preference. The boiler will then find the best match in the existing list of drugs, and if it can’t it will create a new one with the Dose Form you gave the highest priority.