The approach we’ve been taking in cohort builder is that all of the underlying queries should actually be generic and therefore we are defining them in the reporting module, and we’re defining them in these (new) “definition libraries” that I haven’t had a chance to really write much about yet. See example.
So, we might be able to define the underlying queries for these reports in a module like emrapi. For example if we were going to have a “top 10 diagnoses during time period” report, this definition could go in emrapi. If it does require specific reference application metadata it could go in the referencemetadata module (but since the refapp is really a base, rather than an actual implementation, I hope we can avoid this).
The other thing is that you don’t need to create a “report definition” in the database, but instead we can put dataset definitions in a definition library (which is defined in code).
And +1 to Mike’s suggestion that instead of a “report” per se you focus on building individual UI components that can be reused across different distributions.