By Rien Wertheim – This week, at the HL7 Working Group Meeting in Paris, I witnessed an interesting discussion on the positioning of the “FHIR registry” eithin the standard. The registry is a website where communities manage their FHIR profiles, value sets, extensions etc. They are a central concept in the FHIR standard, because without a registry FHIR will be much less a standard than it claims to be.
How is that?
One of the starting points of FHIR is the 80/20 rule, which in practice means that you cannot standardize the whole universe and that any standard should leave room for customization. There is no perfect world. There might a better world.
That is why FHIR leaves much open for use cases to mold the standard to their own specific requirements. They do this by making agreements about “conformance”. For example, FHIR has a very generic resource called Patient. In the Netherlands there will be a Patient_NL, or even a Patient_NL_Amsterdam, which derive elements from each other.
Someone in the working group meetings asked Grahame Grieve (the project lead of FHIR): “Suppose everyone starts making his or her own profiles and extensions, what the hell would that do to interoperability? How do I know if a profile doesn’t exist already?” Grahame’s answer: “The FHIR registry is on top of my most-important, damn-if-I-don’t task list.”
Everyone agrees that you need a registry. But this is where it gets interesting: will there be one registry, e.g. at fhir.org, managed by HL7 International? Will there be national registries? Or even local? Or vendor registries? These are scenarios of increasing horror in the perspective of (some) standardization people.
Someone of influence told me at the conference: we will never host our registry at HL7. They are a standardization organization, not a services organization. Besides, we want full control.
Let us presume that indeed registries are no different from anything else in standardization, or in fact in anything human, and that global unity is an illusion. What than can we do to maximize interoperability? The answer is that these registries should facilitate users in sharing profiles. Not because they feel they have to, but because it’s easy. These registries should therefore be able to facilitate searching for profiles in other related registries in a federated model.
And even then, it won’t be perfect. As Ewout put it in his policy making address at the Paris Working Group Meeting: we have to open FHIR registries, let everyone publish their specialized dialect of FHIR and reintegrate when the need arises.