We have a alignment challenge between CDA and FHIR wrt the use of nullFlavor to express certain concepts semantically.  I'm thinking we should adopt the principle that when ever there exists a valueSet that could be used across FHIR and CDA, our choice should be to "give up" the use of nullflavor in the corresponding CDA template in favor of using a single "valueSet only approach" to represent the semantics. 

This principle would start weening us off of the use of nullflavor whenever possible and as appropriate paced by the creation of valueSets to be used by FHIR for USCDI data elements.

For example, following this principle would cause us to the Gender Identity valueSet (2.16.840.1.113762.1.4.1021.32) in the Gender Identity Observation template rather than using the Gender Identity USCDI core valueSet (2.16.840.1.113762.1.4.1021.101) and the separate CDANullFlavor valueSet.  

This would be a simplification principle that would help us accelerate making choices in how we use CDA which make it easier to align with FHIR WHENEVER POSSIBLE.