There has been considerable feedback about the requirement to use ValueSet as the means of conveying the permitted answers for questionnaire questions that are expressed as a choice. The gist of the feedback is that the use of value set is: complex, confusing, awkward and dangerous. The latter is asserted because it raises the potential for valule sets being changed outside the context of the question they're tied to and because questions might share value sets inappropriately or be impacted by changes to value sets that don't take multiple uses into account.

This last issue can be partially mitigated through the use of "contained" value sets, but only partially because not all questionnaire authors will make use of "contained" appropriately. And the "contained" approach does nothing to address the complex/confusing/awkward complaints.

ValueSet is used in Questionnaire because in some questionnaires, there is a need to share permitted answers across multiple questions within a questionnaire or, more rarely, across questionnaires or with DataElements referenced in questionnaires. Some questionnaires make use of externally defined code systems when defining their answer lists. As well, features of value sets such as support for coded ordinals, multiple display names and the ability to map to other code systems are all useful.

This afternoon, the Structured Data Capture work group discussed this issue along with a number of HL7 vocabulary experts. The consenus was as follows:

• Questionnaire should only use ValueSet in circumstances where there is a requirement to share codes across questions or Questionnaires or where the concepts represented are drawn from an externally drawn code system

• Questionnaire.group.question should permit a choice of either ValueSet or a list of Codings (but not both) to define the allowed list of possible answers. Coding was selected because of the need for both a unique "code" as well as a display name and the need to take advantage of code-like machinery such as ordinals and display name translations

• While Coding is used, the "system" property would be omitted.

• Guidance should be provided on when to use ValueSet and when to use the internal coding list as well as how to make use of ConceptMaps to map the codes defined in the context of a particular question (generating the "system" based on the url of the questionnaire together with the linkId of the question)

• That questionnaire should have a url that would function in a similar manner to the url on Conformance resources