Details
-
Change Request
-
Resolution: Persuasive with Modification
-
Medium
-
FHIR Core (FHIR)
-
R5
-
Patient Administration
-
Practitioner
-
-
Brian Postlethwaite/Daniel Rutz: 17-0-1
-
Clarification
-
Non-substantive
Description
One aspect of interoprability of FHIR is not only the more common use-case of FHIR for training AI/ML models but the use of AI/ML agents in an interoperable way. Babylon would like to model their AI/ML agents as "machine practitioners", and perhaps model the different AI/ML models/versions as "practitionerRole"s.
Babylon as a provider would like to be able to attribute FHIR resources created to these "machine practitioners" and be able to use them for interoperability in cases providers exchange their electronic medical records or if providers want to offer their AI/ML agents to partners.
It would reflect correctly in an EMR what a human and what a machine practitioner performed and is essential for evaluating clinical safety and so on.
Supporting "machine practitioner" in HL7/FHIR wouldn't need any additional work by HL7/FHIR implementors but would need an update on the documentation of the FHIR Core Spec (v5) on http://www.hl7.org/fhir/practitioner.html and https://www.hl7.org/fhir/practitionerrole.html which are both managed by the patient working group.
During integration work I noticed that EMRs (like Athena) already model "machine practitioner" in the wild which Cooper confirmed for Epic, too, and such a documentation step in the standard would only reflect what is already used by providers.