Fix Version/s: None
The Summer 2020 updates to the MedMij FHIR Implementation guide clarifies/requires that senders SHALL provide a "clinically safe" narrative, except in the case of a few exceptions.
We request that this be optional (a SHOULD) except for those resources where there is a clear use-case in mind (and we'd be happy to discuss those examples).
For our system, including a narrative with each resource is non-trivial effort, while providing limited value based on the examples we've so far considered.
For additional background to our thinking and questions:
- Unlike other exchange formats, FHIR narratives are defined at an instance level rather than a datatype level (e.g., we'd have two separate narratives for a patient's latex allergy and their peanut allergy, rather than a single consolidated narrative for all of that patient's allergies). Narrative content (the Narrative.div field) is HTML content, so it wouldn't be straightforward for FHIR consumers to piece together narratives from multiple resources. How should app developers use narratives, and are they expected to handle concatenation / reformatting the full set of narratives?
- Discrete elements seem tricky to translate into narrative forms, and we foresee a few roadblocks. Are there recommendations for how these should be handled for those groups that do not currently rely on a Narrative Generator?
- Locale specific formatting and language selection - should Names/Dates/Etc. be formatted according to the locale of the FHIR server, or the locale of the client/receiver? In our system, we also track a patient's language preference directly.
- CodeableConcepts with multiple codings - we typically try to send all possible mappings for coded concepts (e.g. condition.code, where we might return an ICD-10 code and a SNOMED code). How should we convey these coded concepts in narratives? Showing only one code/display might result in unrecognized content, while including all names/codes would make a narrative cumbersome and repetitive. This is mentioned in the last bullet of 5.6.1 of the IG, but seems to require judgement per resource.
- More generally - using textual narratives over discrete data seems like a step backwards in terms of interoperability. Discrete data can be parsed and consumed by application developers with relative ease, whereas narratives / HTML content would be harder to consume in a meaningful manner. For example - calculating medication reminders based on discrete MedicationRequest elements could be implemented in the same way regardless of the FHIR server, whereas attempting to parse medication frequencies from HTML content is challenging. The FHIR specification and the existing profiles are already standardizing terminologies. Are app developers expected to move away from narratives in the long term, and are there specific use cases where narratives are preferred?