Stream: implementers
Topic: Patient identifier
Lauri Karppinen (Aug 16 2018 at 10:28):
Hi! I'm trying to figure out patient resource. At the moment I'm struggling patient identifier. I need right system for values of identifiers. Patient ID itself is going to be used as UUID coming from program. Other one needs to be social security number and other person code from our program. I can't figure out right codes for systems or do I need to add more code to my patient to get it right?
´{
"resourceType": "Patient",
"text": {
"status": "generated",
"div": "<div xmlns='http://www.w3.org/1999/xhtml'><a name='mm'/></div>"
},
"id": "3fd5242b-4b31-4bfd-bbf6-aaa8724a68f7",
"identifier": [
{
"value": "1234",
"system": "?"
},
{
"value": "1234",
"system": "?"
}
],
"name": [
{
"given": [
"Esimerkki",
"Erkki"
],
"family": "Savolainen"
}
],
"birthDate": "2001-12-19",
"gender": "male"
}´
Grahame Grieve (Aug 16 2018 at 10:43):
which countries social security number? The other, the id from your program, make up some URL that no one else will use, that quickly identifies which identifier. typically [your website]/fhir/Identifiers/[something], but the details don't matter - just that your the only one who'll use it, and you know what it is
Lauri Karppinen (Aug 16 2018 at 11:10):
which countries social security number? The other, the id from your program, make up some URL that no one else will use, that quickly identifies which identifier. typically [your website]/fhir/Identifiers/[something], but the details don't matter - just that your the only one who'll use it, and you know what it is
We need finnish, swedish and unknown.
Grahame Grieve (Aug 16 2018 at 11:11):
you'll need to ask HL7 sweden and finland for URIs to use for their social security numbers. If they want, we can add them to the main FHIR spec, but we cannot choose them
Grahame Grieve (Aug 16 2018 at 11:12):
for unknown, leave system blank, and use type with this code: http://build.fhir.org/v2/0203/index.html#v2-0203-SB
Grahame Grieve (Aug 16 2018 at 11:12):
... or... http://build.fhir.org/v2/0203/index.html#v2-0203-SS
Grahame Grieve (Aug 16 2018 at 11:13):
@Rob Hausam I think these should get some definitions to differentiate them
Rob Hausam (Aug 16 2018 at 11:34):
yes, it looks like those and quite a few others in 0203 are missing definitions - sounds like a harmonization proposal for next cycle
Brian Reinhold (Aug 27 2018 at 17:40):
How does one specify an unknown patient in FHIR? There is a way to do it in PCD-01 PID segment but I can't find any info on how this might be accomplished in FHIR. For the moment I am mapping as best as I can the unknown status of the V2/0004 unknown code in the identifier.
Lloyd McKenzie (Aug 27 2018 at 17:41):
Unknown patient in what sense? A patient whose name is not known?
David Hay (Aug 27 2018 at 17:43):
The classic 'unconscious patient in ED'?
Brian Reinhold (Aug 27 2018 at 17:44):
The classic 'unconscious patient in ED'?
Yes. That is one of the most common use cases. Also unable to communicate due to language, perhaps a child, etc.
Lloyd McKenzie (Aug 27 2018 at 17:49):
You can simply omit the patient's name and generate a temporary MRN for them with the intention to possibly merge it once they become known.
Brian Reinhold (Aug 27 2018 at 18:05):
You can simply omit the patient's name and generate a temporary MRN for them with the intention to possibly merge it once they become known.
Would following the PCD-01 example and encoding an identifier with a system of
http://terminology.hl7.org/CodeSystem/v2/0004
and a value of "U" (Unknown)?
Grahame Grieve (Aug 27 2018 at 19:24):
use a code as the identifier? Sounds like a bad idea to me. Should be a data absent reason.
Alexander Henket (Aug 27 2018 at 19:31):
We use the good old NullFlavor (extension) all over the place when applicable. E.g. MSK for when the identifier is masked, or UNK when the identifier is not known (yet).
Alexander Henket (Aug 27 2018 at 19:32):
NullFlavor served us well in V3 and through the extension, continues to do so in FHIR
Kristof Taveirne (May 25 2020 at 14:22):
I'm working on a client application where we need to find a patient based on an identifier.
The user can choose the type (MRN, SSN, ...) , and then can provide a value. I see however that the FHIR endpoint we're using may return multiple identifiers of the same type, but where the system is different (when searching by name).
In other words, we might have to let the user choose between the 2 to really be sure he/she finds the right patient. The problem is of course that these typically are not really human-readable. What would be the right approach for this?
Is it correct to say that "identifier#assigner" can be interpreted as being the human readable representation of the system? Can I make a 1-to-1 mapping between the 2 concepts?
Thank you
Lloyd McKenzie (May 25 2020 at 14:34):
The Identifier.type.text is generally used to provide the human-readable designation. For example, if a Practitioner has multiple license numbers, the type.coding would be the same (indicating that it's a professional license number), but the type.text would indicate the specific jurisdiction and kind of license
Michele Mottini (May 25 2020 at 15:09):
You need to map (out of band) what the user chose in the user interface to identifier systems - type is unevenly supported and (as you notice) not unique - eg identifier.type can be entirely missing, or not informative (identifier.type.text = 'MRN' for multiple identifiers
Ramandeep Dhanoa (May 25 2020 at 21:58):
Hey @Lloyd McKenzie we are working on identifying codes for the patient identifier, which is internally assigned (as in patient primary key). Is that okay to map it to 'PI' - 'Patient internal identifier' ?
https://terminology.hl7.org/1.0.0/CodeSystem-v2-0203.html
The reason I am asking is I reviewed few FHIR examples where PI wasn't used for patient internal identifer, for example:
https://www.hl7.org/fhir/patient-example-ihe-pcd.json.html
Lloyd McKenzie (May 25 2020 at 22:02):
You can't count on Identifier.type being present at all. However, if present - even with text, it is expected to be coded with an appropriate code if one applies. The example you reference is technically non-conformant. Can you submit a change request for it to be fixed?
Ramandeep Dhanoa (May 25 2020 at 22:04):
sure, how do I submit change request?
Lloyd McKenzie (May 25 2020 at 22:16):
At the bottom of every page in the spec is a "propose a change" link. The first time, you'll have to register (free) and convince us you're a real human. After that, you're free to propose changes any time you see something that isn't right/could be better. :)
Ramandeep Dhanoa (May 25 2020 at 22:17):
Gotcha, Thanks :)
Last updated: Apr 12 2022 at 19:14 UTC
