FHIR Chat · extension patterns · implementers

Stream: implementers

Topic: extension patterns


view this post on Zulip Jay Lyle (Nov 07 2019 at 16:10):

We have two different classifications for drugs - a "national" and a "VA" classification.
We are considering two designs to support populating FHIR.

  1. Two extensions, one for each legacy property
  2. One extension, specified twice in profile with different terminology bindings

What are the pros and cons others have encountered between these patterns?

view this post on Zulip Lloyd McKenzie (Nov 07 2019 at 16:20):

Why are you using extension at all? Why aren't they just additional codings?

view this post on Zulip Jay Lyle (Nov 12 2019 at 19:22):

OK, that makes sense. Thanks.
But there are other cases of the question: we have, e.g., classifications of med "type" (similar to doseForm, e.g., SYRINGE, ADMIXTURE, CHEMOTHERAPY, PIGGYBACK, TPN, UNIT DOSE, but on the order not the med) that share some values: the pending inpatient has 10 values; finished IV has a subset; a third case has a different subset. What design issues would help choose between more extensions or deferred bindings?
This, too, could use core properties: med.product.form, assuming we can instantiate another resource. I'll look for another example, because there are some where a pattern would help.

view this post on Zulip Jose Costa Teixeira (Nov 12 2019 at 21:28):

I have a concern about using a product code for a classification. That creates confusion. A product cannot be designated or identified by its classification(s).

view this post on Zulip Jose Costa Teixeira (Nov 12 2019 at 21:29):

Medication code is there to identify a product. A classification does not identify a product

view this post on Zulip Jay Lyle (Dec 09 2019 at 22:51):

Well, the product code will be there. The classifications will be additional, broader encodings, with different systems that should scope their use. Is there a problem with that?


Last updated: Apr 12 2022 at 19:14 UTC