FHIR Chat · Identifying a slice by pattern · IG creation

Stream: IG creation

Topic: Identifying a slice by pattern

Richard Townley-O'Neill (Feb 26 2020 at 04:33):

I have a Fetal Movement profile that slices (STU3) Observation.code.coding by value:system and pattern:code.
When I identify a slice FetalMovementsSNOMED-CT using fixed values in the discriminator it works fine.

      {
        "id": "Observation.code.coding:FetalMovementsSNOMED-CT",
        "path": "Observation.code.coding",
        "sliceName": "FetalMovementsSNOMED-CT",
        "min": 1,
        "max": "1"
      },
      {
        "id": "Observation.code.coding:FetalMovementsSNOMED-CT.system",
        "path": "Observation.code.coding.system",
        "min": 1,
        "fixedUri": "http://snomed.info/sct"
      },
      {
        "id": "Observation.code.coding:FetalMovementsSNOMED-CT.code",
        "path": "Observation.code.coding.code",
        "min": 1,
        "fixedCode": "364617005"
      },

Richard Townley-O'Neill (Feb 26 2020 at 04:35):

We would like to identify slices using patternCoding.

      {
        "id": "Observation.code.coding:FetalMovementsSNOMED-CT",
        "path": "Observation.code.coding",
        "sliceName": "FetalMovementsSNOMED-CT",
        "min": 1,
        "max": "1",
        "patternCoding": {
          "system": "http://snomed.info/sct",
          "code": "364617005"
        }
      },

But the validator does not recognise the slice.

Could not match any discriminators ([system, code]) for slice Observation.code.coding:FetalMovementsSNOMED-CT in profile http://hl7.org.au/fhir/ch/v1/StructureDefinition/ncdhc-observation-fetal-movement - None of the discriminator [system, code] have fixed value, binding or existence assertions

Richard Townley-O'Neill (Feb 26 2020 at 04:38):

Though the IGpublisher produces a differential table that recognises system and code as 1..1 with fixed values.
image.png

Lloyd McKenzie (Feb 26 2020 at 04:39):

If you're slicing by pattern, the discriminator should just be coding - no need for both system and code as both would be covered by the pattern

Richard Townley-O'Neill (Feb 26 2020 at 04:47):

(deleted)

Richard Townley-O'Neill (Feb 26 2020 at 04:49):

We are slicing Observation.code.coding, not Observation.code

Richard Townley-O'Neill (Feb 26 2020 at 04:51):

      {
        "id": "Observation.code.coding",
        "path": "Observation.code.coding",
        "slicing": {
          "discriminator": [
            {
              "type": "value",
              "path": "system"
            },
            {
              "type": "pattern",
              "path": "code"
            }
          ],
          "rules": "closed"
        }
      },

Richard Townley-O'Neill (Feb 26 2020 at 04:53):

Observation.code is 1..1, so it cannot be sliced by coding.

Lloyd McKenzie (Feb 26 2020 at 04:55):

But why are you declaring two discriminators? Why not:
"slicing": {
"discriminator": [
{
"type": "pattern"
"path": "$this"
}
]
}

Lloyd McKenzie (Feb 26 2020 at 04:56):

And then just declare patterns on the codings for your slices?

Richard Townley-O'Neill (Feb 26 2020 at 04:56):

I did not know that was an option.

Richard Townley-O'Neill (Feb 26 2020 at 04:58):

Thanks. I'll investigate that option.

Richard Townley-O'Neill (Feb 26 2020 at 05:07):

That replaces the old error (Could not match any discriminators ...) with

Unsupported fixed pattern type for discriminator($this) for slice Observation.code.coding:FetalMovementsSNOMED-CT: org.hl7.fhir.r5.model.Coding

Lloyd McKenzie (Feb 26 2020 at 05:10):

That's very weird. @Grahame Grieve ?

Rob Hausam (Feb 26 2020 at 20:44):

I ran into exactly the same thing a few days ago. Apparently pattern is not currently supported with the Coding data type (fixed value constraints are also not supported for Coding, but pattern is what I was looking for). I had to constrain Coding.system and Coding.code in order to get it to work.

Shovan Roy (Feb 26 2020 at 23:14):

I did the same thing as @Rob Hausam all of our profiles (except vital sings) are following pattern and we are living with this work around. http://build.fhir.org/ig/hl7au/au-fhir-childhealth/profiles.html

Grahame Grieve (Feb 26 2020 at 23:53):

pattern is not currently supported with the Coding data type

what happens if you try?

Rob Hausam (Feb 26 2020 at 23:55):

You get the Unsupported fixed pattern type for discriminator error that @Richard Townley-O'Neill reported.

Grahame Grieve (Feb 26 2020 at 23:58):

Does someone want to contribute a test case?that would be:

• a profile based of R4 base resource that defines a pattern with a Coding based discriminator
• an instance that conforms to the profile
• an instance that doesn't conform to the profile

Rob Hausam (Feb 26 2020 at 23:59):

or a similar error message (I could try to reproduce it again on my end if needed)

Grahame Grieve (Feb 27 2020 at 00:00):

that's the error. I found it and will fix it for next release if someone provides test cases

Shovan Roy (Feb 27 2020 at 00:01):

I or @Richard Townley-O'Neill will get you the test cases

Rob Hausam (Feb 27 2020 at 00:01):

sure - I'll try to find some time tonight to do it (unless someone else is able to do it first)

Rob Hausam (Feb 27 2020 at 00:01):

one of us should be able to

Shovan Roy (Feb 27 2020 at 00:37):

Hi @Grahame Grieve ,

Please see the two Profiles and the examples as provided below:

--Failing with Pattern--
For http://build.fhir.org/ig/hl7au/au-fhir-childhealth/branches/master/StructureDefinition-ncdhc-observation-urinalysis-protein.html , we have slice discriminator on pattern = code the sample is failing with error:
"Could not match discriminator ([code]) for slice Observation.code.coding:Urinalysis-SNOMED-CT in profile http://hl7.org.au/fhir/ch/v1/StructureDefinition/ncdhc-observation-urinalysis-protein - the discriminator [code] does not have fixed value, binding or existence assertions"
here is the link for the sample: http://build.fhir.org/ig/hl7au/au-fhir-childhealth/branches/master/ncdhc-observation-urinalysis-protein-example.html

--Working with Pattern (with workaround)--
As @Richard Townley-O'Neill mentioned in this discussion thread, For http://build.fhir.org/ig/hl7au/au-fhir-childhealth/branches/master/StructureDefinition-ncdhc-observation-fetal-movement.html we have slice discriminator on pattern = code and value=system, the sample is working.

here is the sample: http://build.fhir.org/ig/hl7au/au-fhir-childhealth/branches/master/ncdhc-observation-fetal-movement-example.html

Here is the link for QA page : http://build.fhir.org/ig/hl7au/au-fhir-childhealth/branches/master/qa.html

@Rob Hausam, fyi.. on profile/samples we are using

Grahame Grieve (Feb 27 2020 at 12:17):

ok.. fixed next release

Grahame Grieve (Feb 27 2020 at 12:17):

thanks.

Grahame Grieve (Feb 27 2020 at 13:00):

btw i had to hack the profile - check the test cases at https://github.com/FHIR/fhir-test-cases/commit/1d95524a886c2704ebdc6dd4361634a35a4626f8

Shovan Roy (Feb 27 2020 at 20:32):

Thanks @Grahame Grieve I can see that you are using pattern = $this as the discriminator, we will apply this to our profiles once the fix is published. Did I miss anything else?

Richard Townley-O'Neill (Feb 27 2020 at 23:33):

In commit 1d95524a886c2704ebdc6dd4361634a35a4626f8
validator/profile-slicing-coding-bad.json and validator/profile-slicing-coding-good.json have the same value for Observation.id

Grahame Grieve (Feb 28 2020 at 00:20):

it does. doesn't really matter but I'll fix it

Richard Townley-O'Neill (Feb 28 2020 at 00:22):

:+1:
I feared that unique ids were important to validation.

Shovan Roy (Mar 01 2020 at 22:17):

I took the latest IG Publisher, I can see the slicing with 'patern=$this' is coming out nicely. thanks @Grahame Grieve
image.png

Last updated: Apr 12 2022 at 19:14 UTC