USCDI definition of "sex" does not stipulate "not characterized as sex assigned at birth or birth sex" (as per US Core); it only states "Documentation of a specific instance of sex and/or gender information." This seems to be very broad both semantically (sex and/or gender) and procedurally (a specific instance of . . . information). It seems, in fact, to align quite nicely with what the Gender Harmony project called "recorded sex or gender" - a copy of a record of from some source that may not provide unambiguous semantics. I.e., this is a place to record a vaguely defined sex or gender concept, which may (or may not) be unambiguously specifiable in any terminology system.

The problem is that the USCDI specification also asserts that SNOMED is the applicable vocabulary standard. The element is defined as constitutionally vague, which suggests that it would be a very unusual code system that could support it, certainly not one as semantically precise as SNOMED CT.

Ideally, USCDI would withdraw this terminology binding - because it seem unlikely to actually specify values. 

As an interim fix, the binding in the specification should be relaxed (best: may, but should would be tolerable) to allow the transfer of poorly defined but possibly useful information. "Extensible" implies that the sender has some way to assess synonymy: since the whole point of RSOG is vagueness, this seems unrealistic, so "example" seems appropriate.