There are multiple opportunities and challenges with proceeding with short messaging flows to FHIR workflows outside of a singular language. There are different approaches depending on whether the implementer will use one dominant language, or is in context in which there must be multiple languages simultaneously supported. For the most comprehensive overview, see this talk by Rene Spronk.
Note: FHIR IDs (resource.id) are limited to ASCII. But, identifiers, code systems, and other key values are usable in Unicode.
Per the specification, there are several approaches that may be used, specifically:
- Language tag on Resource and Narrative
- Translation Extension
- Translations on CodeSystem and ValueSet
- Translatable extension on element definition
- Language Tags on HTTP requests
- Languages in Names and Addresses
What this means is that implementers must approach their multiple language needs differently.
For starting with FHIR Questionnaires as the primary source of the survey structure and text, there are experiments ongoing to use .PO or XLIFF files to hold translations and render Questionnaire resources with the translations. See this discussion on the FHIR chat for more about implementation methods.