The practitioner resource has a simple list of communication codes (CodeableConcept), unlike the Patient and RelatedPerson, which have a list of possible languages, along with a boolean for preferred. While the use case for practitioner languages is different (in a different direction) the underlying notion of 'can speak if really necessary' vs 'a language I am comfortable in' still seems applicable. Therefore I propose that for consistency's sake, the structure should be the same.

Note: this task is created as part of defining a common pattern "HumanActor" which applies to all 3 resources. In the absence of a 'preferred' property, the pattern will have not much option but to mark all the languages as preferred for a practitioner