Stream: implementers
Topic: Code system for a formula that generates codes
Jean Duteau (Mar 18 2019 at 15:28):
NCPDP has a question about their product identification code system. They use an 11-level "NDC" which takes the three parts of FDA's NDC and uses the larger of the numbers, 5-4-2 for labeler code, product code, package code. I'm pretty sure that they can't just use the FDA NDC code system URL since the codes are different, but since they are different algorithmically, I could be wrong on this. My take is that someone (probably the NCPDP) would need to define a code system for them. Does this make sense?
Grahame Grieve (Mar 18 2019 at 19:29):
so they're just picking a particular NDC code?
Rob Hausam (Mar 18 2019 at 19:37):
I believe this is the same unofficial "normalized" 11-digit NDC format that we've discussed previously - it's not published by FDA, but a lot of people use and rely on it, like apparently NCPDP and drug database vendors like Multum (and I believe FDB and others, as well). I think we do need to get something in place for it. We have some older threads on this, too, that maybe we should dig up. We just haven't landed on anything in particular yet, as I recall.
Jean Duteau (Mar 18 2019 at 20:12):
@Rob Hausam yes, that is exactly it.
@Grahame Grieve NDC's can be 4-4-2, 5-3-2, or 5-4-1 (labeler, product, package respectively) and the NCPDP formula takes these three parts and inserts a '0' (I think) where needed so that it becomes 5-4-2.
Grahame Grieve (Mar 18 2019 at 20:35):
yes we definitely need to land this. @Rob Hausam can we accelerate this?
Jean Duteau (Mar 18 2019 at 20:42):
@Rob Hausam @Grahame Grieve I believe that it is actually NCPDP that has created the formula and the US formularies/catalog have adopted it. At least that was the impression that I got from my recent NCPDP call.
Rob Hausam (Mar 18 2019 at 20:44):
Sure. I don't see why not. I think we just need to decide where the "unofficial source of truth" for the 11 digit codes is going to be hosted. I was just looking earlier today about what to load to tx.fhir.org for NDC, since we need it for IPS. So far we used the official format with dashes in our IPS example, but it makes sense to go ahead and deal with the 11 digit format, too. I have the full current NDC download as of today. But it's released daily, so ongoing maintenance has to be considered and could be a significant challenge. To be reasonably reliable and useful, I think we would need a plan to update it at least monthly?
Grahame Grieve (Mar 18 2019 at 20:45):
how many codes in NDC?
Jean Duteau (Mar 18 2019 at 20:47):
~126,260 as of 3/18/2019
Rob Hausam (Mar 18 2019 at 20:47):
yep
Grahame Grieve (Mar 18 2019 at 20:49):
a little big for a code system resource then
Jean Duteau (Mar 18 2019 at 20:49):
hmm, that's the number of product codes, there are ~241,000 package codes and I think that's the actual NDCs (still learning my way around the code system)
Jean Duteau (Mar 18 2019 at 20:50):
yep
Rob Hausam (Mar 18 2019 at 20:52):
yes - the "official" NDCs are the package codes
I've never seen the product codes used independently, but they are distributed that way now and those may be useful for organizing, searching, etc.
Grahame Grieve (Mar 18 2019 at 20:52):
I'll have to look into it. How is it released?
Rob Hausam (Mar 18 2019 at 20:52):
tab delimited text, or Excel
Grahame Grieve (Mar 18 2019 at 20:53):
I'll add it to my list
Grahame Grieve (Mar 18 2019 at 21:13):
is there any release of the 11 digit codes?
Jean Duteau (Mar 18 2019 at 21:25):
Not an official release that I've seen. All of the drug database vendors have them but they are generated by a formula so I haven't seen anyone release them officially
Grahame Grieve (Mar 18 2019 at 21:50):
do we think that the product ids have a use without the package code?
Rob Hausam (Mar 18 2019 at 22:52):
I think they may be useful for organizing and searching (and probably picklists, etc.) as I mentioned. I don't think there is any "official" use so far that I am aware of for the product codes.
Grahame Grieve (Mar 18 2019 at 23:02):
is it an is-a relationship?
Rob Hausam (Mar 18 2019 at 23:12):
I'm not sure that I would necessarily consider it to be a true is-a relationship - but if you decided to represent it that way I don't think anyone would have trouble understanding or using it.
Rob Hausam (Mar 18 2019 at 23:19):
I think I would consider it to be more like an "axis" model as in LOINC (e.g. it might be analogous to a LOINC axis like COMPONENT as it's related to the entire LOINC code).
Jim Steel (Mar 20 2019 at 04:51):
Why is that too big for a CodeSystem resource?
Rob Hausam (Mar 21 2019 at 00:41):
Good question. What actually is "too big"? I would be able to generate a CodeSystem resource for the full NDC (for a particular daily release) - if it's useful to do that. In the meantime I have a CodeSystem resource with a few example NDC codes about ready to push to the packages repo.
Grahame Grieve (Mar 21 2019 at 00:57):
I'm not interested in trying generate a code system resource with 360k codes
Rob Hausam (Mar 21 2019 at 01:02):
What are you thinking about for NDC now - both for the official release with 10-digit codes and for the unofficial 11-digit codes that we've been discussing?
Grahame Grieve (Mar 21 2019 at 01:12):
so the 11 digit code for 0002-0800-01 would be 00002080001 ?
Rob Hausam (Mar 21 2019 at 01:13):
yes
Grahame Grieve (Mar 21 2019 at 01:13):
and there's no 11 digit code for just 0002-0800 ?
Rob Hausam (Mar 21 2019 at 01:13):
right
Rob Hausam (Mar 21 2019 at 01:15):
The 11 digit codes are for the package
theoretically you could have normalized 9 digit product codes, but I haven't seen that
Grahame Grieve (Mar 21 2019 at 01:22):
ok so I think they should have the same code system, and we should define a property so that you can separate them
Rob Hausam (Mar 21 2019 at 01:25):
I expect the 11 digit codes probably should be a code system supplement, since they don't come from the official publisher - FDA
Grahame Grieve (Mar 21 2019 at 01:25):
probably.
Jean Duteau (Mar 21 2019 at 01:25):
how do we identify it as a supplement?
Rob Hausam (Mar 21 2019 at 01:27):
using CodeSystem.supplements (in R4)
Grahame Grieve (Mar 21 2019 at 01:27):
ok I'll have NDC support up on tx.fhir.org in the next couple of hours
Jean Duteau (Mar 21 2019 at 01:28):
the 11-digit codes are basically equivalences for the "real" code so could we make the 11-digit code a property of the "real" code?
Grahame Grieve (Mar 21 2019 at 01:29):
that doesn't make the 11 digit codes real codes that can be used in Coding.code, which is the goal here
Jean Duteau (Mar 21 2019 at 01:29):
true, okay
Rob Hausam (Mar 21 2019 at 01:32):
we've said the supplements can't define new codes
so that might argue for a separate code system
it would be sort of like the legacy SNOMED identifiers - we decided on a different system for them
Grahame Grieve (Mar 21 2019 at 01:33):
yes we have said that. Do we want to do that though? It would make it easier to say 'just 11 digit codes' but they're purely equivalent codes, yes?
Rob Hausam (Mar 21 2019 at 01:36):
yes, they are purely equivalent - each 10 digit code (in all three formats) generates a unique 11 digit code
you can't deterministically go from the 11 digit code back to the original 10 digit code, as you don't know which format was used, but that shouldn't matter
Grahame Grieve (Mar 21 2019 at 02:03):
well, for the moment, tx.fhir.org is using the same code system
Grahame Grieve (Mar 21 2019 at 02:12):
http://tx.fhir.org/r4/CodeSystem/$lookup?system=http://hl7.org/fhir/sid/ndc&code=00002080001
Grahame Grieve (Mar 21 2019 at 02:12):
http://tx.fhir.org/r4/CodeSystem/$lookup?system=http://hl7.org/fhir/sid/ndc&code=0002-0800-01
Lloyd McKenzie (Mar 21 2019 at 02:29):
Code system supplements CANNOT define new codes - only new properties and relationships for existing codes. If you have distinct codes, you have a distinct code system. No way around that.
Grahame Grieve (Mar 21 2019 at 02:38):
yeah so? we said that but it doesn't close out this discussion
Lloyd McKenzie (Mar 21 2019 at 02:44):
We either declare that the codes are implicitly part of the NDC code system and in the usage description allow them to be treated as such, or we define a separate code system. So the question is really whether the maintainer of SDC will let us get away with doing the former.
Grahame Grieve (Mar 21 2019 at 02:51):
right those are the choices. I don't think that NDC publisher will understand or care
Lloyd McKenzie (Mar 21 2019 at 02:53):
Right, but will they not understand and care and say "no" or "yes". (Neither understanding nor caring are required to provide an answer :))
Rob Hausam (Mar 21 2019 at 02:53):
that may be true (what Grahame said) - but I think we need to find out
Grahame Grieve (Mar 21 2019 at 03:24):
fda.. I guess I'll ask @Mary Ann Slack
Last updated: Apr 12 2022 at 19:14 UTC
