Stream: committers/announce
Topic: Expectations around including target types for elements
Lloyd McKenzie (Oct 06 2020 at 22:09):
MnM had some discussions about what the rules should be for including resources in choices and as targets for references and canonicals. We've had some challenges with different work groups following different rules for deciding what should be included vs. not. Some felt that the 80% rule applied (so only list a type if there's a belief that 'most' existing systems support the type), while others followed a looser set of expectations.
The outcome of the methodology discussion is as follows:
The bar for including a type in a choice is lower than the "80% rule" because the cost of a type being omitted is higher. Types should be included if the WG can foresee a reasonable likelihood of a type being used in a production system for the specific data element.
To give a concrete example, even if very few systems allow patients to author MedicationRequest instances, 'Patient' should still be a target of the 'subject' element because there's a reasonable likelihood of that being allowed by at least some systems. The term "reasonable likelihood" means that you don't have to allow totally ludicrous use-cases, merely those where there's a reasonable chance of them showing up in real systems.
If there are questions/concerns, please discuss on #committers.
Last updated: Apr 12 2022 at 19:14 UTC
