Open ArdonToonstra opened 3 years ago
Discussed with Arianne that it would be a better idea to stick to our current approach. Although a complete mapping to the MP dataset would be preferred and helpful, it does not seem to outweigh the negatives. It will be a lot of manual work that is prone to errors and will require work in the governance of it, while the mapping is already present in the zib mapping.
Perhaps we could include in the Profiling Guidelines that a general rule or recommendation is that nl-core- and use-case mappings are added next to the zibs mappings. So concepts that are defined similarly in the use case as by the zib would only get a zib mapping. Profile consumers should therefore be aware that they need to look at the zib mappings + the use case mappings.
@pieter-edelman-nictiz and @jduwel do you agree and do you think this should be included in the profiling guidelines?
There's a bit of a fundamental discussion that underlies this issue. It's not really clear what we want to use our profiles for, and if there maybe should be multiple profiles for different goals -- with mappings on different places.
In any case, if we decide to add use case mappings next to the zib mappings, this should be done in automated fashion.
The 'nl-core' profiles may add concepts that are used in one or more use cases. These concepts get a mapping to the functional model where these are defined (e.g. a mapping to a dataset/transaction in ART-DECOR).
For the medication-related zibs, concepts of the MP dataset are profiled in the nl-core profiles. These get a mapping to the MP9 2.0.0 dataset. This dataset inherits the zib concepts. Do we need to provide a mapping in the nl-core profile to the zib concepts too?
This makes it more explicit and gives a nice mapping overview if you select the MP mappings. It comes however at the costs of more work and more governance.