Stream: implementers
Topic: ICD Code Formatting
Karl M. Davis (Feb 24 2018 at 02:19):
Evening all! Quick question about ICD 9 CM code formatting conventions: is it expected for us to include the category separator in the ICD codes that our FHIR API returns, e.g. "123.4" vs. "1234"?
Lloyd McKenzie (Feb 24 2018 at 02:21):
Based on the guidance here: http://build.fhir.org/icd.html
the periods are included
Karl M. Davis (Feb 24 2018 at 02:22):
Lloyd, where on that page are you seeing such guidance?
Lloyd McKenzie (Feb 24 2018 at 02:25):
I'm looking at the example near the bottom
Lloyd McKenzie (Feb 24 2018 at 02:25):
However, if you'd like an explicit statement, you can certainly submit a change request
Karl M. Davis (Feb 24 2018 at 02:30):
Might be for the best, since this statement at the top makes it sound optional to me: "Plus, if period ('.') characters are disregarded (as occurs in some systems), a few of the codes are overlapping between ICD-9 and ICD-10."
Karl M. Davis (Feb 24 2018 at 02:31):
As an example, it appears that literally all of the Medicare claims data doesn't bother including periods. I'm happy to fake it in our API, but that's a pretty big precedent.
Karl M. Davis (Feb 24 2018 at 02:35):
Didn't have a Forge account, so the change request will have to wait until that's activated. Thanks, Lloyd!
Lloyd McKenzie (Feb 24 2018 at 02:36):
Should now be active, so go for it :)
Rob Hausam (Feb 24 2018 at 02:37):
It shouldn't matter if the codes overlap between ICD 9 and 10 without the periods, as long as the system is specified correctly. I do think that many or most systems will have to handle both variants, and at least the format without the periods should be acceptable in most cases.
Lloyd McKenzie (Feb 24 2018 at 02:38):
We need to nail down the expected format of the codes - because most software (and many terminology servers) will do a straight string match.
Rob Hausam (Feb 24 2018 at 02:41):
I'm not at all sure that we can for this, Lloyd, except in profiles.
Rob Hausam (Feb 24 2018 at 02:43):
As Karl said, there is a lot of precedent - and it's both ways.
Rob Hausam (Feb 24 2018 at 02:46):
And a similar situation is going to be true regarding the ICD dual coding conventions. Our guidance right now is to separate the codes using a space. But some jurisdictions, particularly Germany, have requirements that make it necessary to include the symbols, and others have a preference for doing that.
Lloyd McKenzie (Feb 24 2018 at 02:48):
If we can force systems to map OIDs to URIs and can force systems to remove separators from social security numbers, I don't understand why - for transport - we can't force the inclusion or removal of periods. How they're persisted and rendered can be different
Rob Hausam (Feb 24 2018 at 02:56):
There will be similar variations in handling the NDC codes for medications, too - both 10 digit with dashes and normalized 11 digit without dashes formats, in particular. And to make it even worse, the 11 digit format, while unique itself, can't always be algorithmically transformed back to the correct 10 digit with dashes format.
In both cases, I think the reason that you can't force the inclusion of the periods (or the dashes) is that some systems simply won't have them available in their data, and there may not always be an algorithmic way to add them back (at least in NDC that's true). The only absolute answer in those cases is to mandate removing them, but in both of those cases the formats without the non-alphanumeric characters are not the official formats that are published by the code system authorities.
Lloyd McKenzie (Feb 24 2018 at 02:58):
If there's no algorithmic way to do the conversion, then I can accept that - though I'd argue for distinct system URIs for the different serialization approaches. If there is an algorithmic way to do the conversion, I don't understand the issue.
Rob Hausam (Feb 24 2018 at 03:06):
I agree that if there is an algorithmic way to do it we could try to require that - but even if that's possible, people will likely want to send the data the way that they have it. If we hold firm on the requirement and if conformance actually has enough teeth, then that might change. And we may be able to use distinct system URIs in the other cases, but I'm not quite as sure if we will be able to make that work.
Lloyd McKenzie (Feb 24 2018 at 03:09):
Well, people might want to send the OID for SNOMED-CT or include dashes in their SSN - but if they do that, we declare them non-conformant
Lloyd McKenzie (Feb 24 2018 at 03:09):
I don't see how this situation differs
Lloyd McKenzie (Feb 24 2018 at 03:10):
The effort isn't really different and the rationale is identical
Rob Hausam (Feb 24 2018 at 03:20):
I agree. But if we don't state in the core spec that the codes SHALL always be in the officially published format (unless a separate non-official system is created and maintained by "someone" to handle the variants), then we won't be able to hold that line, and in some cases (like NDC in particular) I think we can expect to have pushback.
Lloyd McKenzie (Feb 24 2018 at 03:37):
Right. I'm asserting that "SHALL" language is appropriate for ICD. For NDC, we may have no choice but to support both - but distinct URIs might be an option there.
Lloyd McKenzie (Feb 24 2018 at 03:37):
(And ideally the FDA could maintain a concept-map to support bi-directional conversion)
Rob Hausam (Feb 24 2018 at 03:38):
I don't mind trying that approach with ICD. And you can talk to the FDA. :)
Rob Hausam (Feb 24 2018 at 03:58):
For ICD, we also absolutely have to decide whether multiple approaches can be allowed for the "dual coding" situations. That's particularly true because often there is a fairly widely accepted convention that is used, which varies between jurisdictions and is required in some and optional and/or different in others. And GF#13210 has not yet been applied to the Dual Coding section (4.2.11.3) on the "Using ICD-[X] Codes with FHIR" page, but it needs to be soon. Germany, in particular, is concerned about their required use of the non-alphanumeric (+*!) characters for indicating the "role" of each code being recognized as a valid syntax for ICD dual (or actually multiple) coding.
Grahame Grieve (Feb 24 2018 at 05:26):
well, whatever we decide we have to be very clear about this. I would obviously prefer that we pick a single choice and use that, but if it's not possible, it's not possible
Joseph Athman (Apr 20 2018 at 13:17):
Hi everyone, I realize this is an old conversation, but I'm wondering if there has been any progress related to NDC codes? At a minimum I'm hoping we can have a distinct URI to be able to indicate that the NDC is the 11 character variety. It is unfortunate that there is no systematic way to go from 11 digits to 10, but that's what we have. If you are a system which only has the 11 digit variety then you really have no option right now.
Grahame Grieve (Apr 20 2018 at 21:13):
please ask on the terminology stream. this is an active discussion this week.
Brendan Keeler (Jul 19 2021 at 17:50):
deleting, as this was a dumb question
Last updated: Apr 12 2022 at 19:14 UTC
