I question whether any critical app or solution should use a managed FHIR server.
The big cloud providers all have 1,000s of customers all over the world using their servers. Surely there’s safety in numbers?
My feeling is that the opposite is the case here. With a large provider you’re only one client of many.
Most FHIR servers are incomplete. Work is ongoing and new releases happen frequently.
Any one of these releases could break a key piece of functionality for you.
- An $operation that changes its expected behavior
- A SearchParameter that doesn’t work any more
- Request headers that behave differently to yesterday
- Performance issues that deteriorate overnight
For your app or solution any one of these could mean pages not loading for doctors, background processes failing, clinical data getting lost or never being processed.
Bugs happen. That’s the nature of software.
But when you’re using a managed FHIR server, you don’t control the release cycle.
You can’t “roll back” when a new release breaks something.
FHIR is complicated. There are resource types, elements and search parameters that are so obscure you could be the only one using them with a particular server provider.
A new release could break for you and work for everyone else.
Ask yourself these questions:
- Am I notified in advance of a new server release?
- Can I test my solution end-to-end against the new release before it goes live?
- If a new release breaks something, will the server provider “roll back”?
- Do I have 24/7 access to critical technical support — by phone?
Managed FHIR servers are easy. They often fit neatly into existing company infrastructure.
Don’t let “ease of use” cloud your judgement when making those early decisions around which FHIR server to use.
“Managed” is not the only option.
---