As Eliot points out:
CDA/CCD documents amalgamate lots of data together. There are too many data points. That makes these documents extremely impractical for solving real world integration problems. They tend to be very brittle and difficult to accomodate changes. It’s one of many reasons why the cost associated with using CDA/CCDs for integration are so high with a low return on investment. I have seen the blood in the field.
FIHR, in contrast, has a much better design and is much more modular (A nice overview is here.) "The FHIR train has left the station. It’s picking up speed and no one, not I, not the HL7 organization, nor even Grahame Grieve who invented the thing can stop it. The concepts behind FHIR will have a life of their own. This train is going to be disruptive. If you run an organization in healthcare you need to know about it. Your only choice at this time is to get on the train, or step in front of it."
We remain sceptical. First, FIHR is owned by HL7. Second, FIHR is still based on the RIM -- which is what caused all the problems in the first place.