FHIR Chat · narrative constraints

Has anyone considered, either in the base resource or in a profile, constraining the format of the narrative? In particular, I'm thinking about whether ImagingStudy should recommend/require something like "[modality] [procedure type] [accession]". This way, a list of imaging studies from multiple sources would have a similar format, and would list nicely. Details of the actual string aside, is this an appropriate thing to do?

Do we have any expectations for what narratives should looke like? I recognize that Observation and others have such wide use that it is hard to come up with a standard narrative format, but what about for narrower-use resources?

Brian Postlethwaite (Jun 20 2016 at 21:20):

For generated narratives, it really doesn't matter as could be re-generated to conform, however for content that is not, maybe not practical.

Brian Postlethwaite (Jun 20 2016 at 21:20):

Would like to see it at least considered though.

Erich Schulz (Jun 20 2016 at 21:52):

wouldn't middleware that took a structured resource -> text be a better (more flexible/sustainable) solution?

Grahame Grieve (Jun 20 2016 at 22:19):

curerntly we have code that generates narratives in several languages, all open source and with permissive licenses. If someone else wanted to package that code into a middleware, or write their own, we think that's a good idea

Grahame Grieve (Jun 20 2016 at 22:21):

with regard to the original question: yes, it's been considered, and I've done it for CDA IGs before. HTML is kind of hard to constrain, so people tend to write narrative, but I have done, for instance, there SHALL be a table, there SHALL Be the same number of rows in the table as their are entries in the act as an XPath, But it's really hard to get a useful XPath that's not too restrictive

Elliot Silver (Jun 20 2016 at 22:24):

OK. As I was writing my question, the thought came to me that we could easily generate that text from the resource elements and ignore the narrative totally, but I wanted to see what the general approach was.

Grahame Grieve (Jun 20 2016 at 22:27):

nearly every resource example in the spec has a generated narrative using that code

Grahame Grieve (Jun 20 2016 at 22:28):

but in production, you might not be able to do that - that's why there's a status code with the narrative. If it contains additional information not in the data (sometimes) then you can't ignore the narrative