Currently a PractitionerRole only needs a Practitioner and active flag, and a Practitioner only needs a name and active flag. This means clients might have to chase down multiple references just to get information (i.e., a name) that could be present in a Reference.display directly. 
 
It would be good to understand the use cases for PractitionerRole over Practitioner, and to create profiling constraints on PractitionerRole based on this understanding. For example, we might start by following some of the practices from US Core and: mark the following as Must-Support:

• code
• location
• organization