Dwaraka Srinivasan asks:

Why is consent not considered as a type of Questionnaire? It looks like Consent and Questionnaire could be combined especially if certain questionnaires are valid only for a certain period of time and require validation / verification

I asked Lloyd McKenzie who leads the Questionnaire work:

All resources can be represented using questionnaire, given that questionnaires allow you to capture arbitrary data elements nested in arbitrary ways. In theory, we could drop every resource and just do everything with Questionnaire. However, if we did that, we’d get minimal interoperability – because everyone defines their own custom questions and answer sets (and frequently revises them fairly frequently too).

Questionnaires make great tools for standardized data capture and give you really robust control over user interface. You control the specific wording of the questions, the order of them, the allowed answer choices, any validation rules, etc. However, they’re not good for searching , decision support, or anything else that requires standardized representation of knowledge.

A common architectural pattern is to use Questionnaires to perform initial data capture (possibly with some of the answers pre-populated from other FHIR resources available in the server) and then, once complete, extract data from the QuestionnaireResponse into the appropriate resources (be they Observations, AllergyIntolerances, MedicationStatements, or Consents). You can then link the resulting resources back to the original QuestionnaireResponses using elements like derivedFrom or Provenance.

The Consent resource provides a standard computable representation of the information about a consent action in much the same way as an Observation can provide standardized representation of a blood pressure. Yes, you can gather both using a Questionnaire, but for data sharing, you’re still going to want the purpose-specific standardized data structure.