Dear All
As an involved clinician I have been thinking about advantages of using inheritance to relate DATA_STRUCTURES and CLUSTER. This arises from two things we have learned in clinical modelling:
-
At almost every point where a DATA_STRUCTURE is currently specified in the RM it would probably be inappropriate to model any structure other than a ITEM_TREE. This is because any change to this will invalidate existing data and we have learned that information models get more complex with time. So the construct of DATA_STRUCTURE at the moment offers no real value in return for the complexity it introduces as it stands at the moment.
-
A ITEM_TABLE or ITEM_LIST ‘constraint’ can more usefully appear at any point in the data – even forcing a ITEM_TREE is useful if you can see that specialisation should always allow hierarchical structures.
I am suggesting that we change the model to have ITEM_STRUCTURE inherit from CLUSTER. This would mean that ITEM_STRUCTUREs always behaved as clusters (paths etc) but had added features. A ITEM_LIST then becomes a constraint on a CLUSTER to only allow ELEMENTS. And ITEM_TABLE could now appear anywhere in the data.
For backward compatibility it would mean that ITEM_SINGLE and its ‘item’ attribute would have to be made obsolete. In fact an ITEM_LIST with cardinality of 1 offers the same constraint capability. At the moment ITEM_STRUCTURE and HISTORY inherit from DATA_STRUCTURE but I am not sure if that confers any benefit as I do not think that DATA_STRUCTURE is present anywhere in the model. Thomas may want to elucidate this aspect. It may mean that we need a new interface with multiple inheritance in Eiffel if that is important.
This does mean that openEHR is more closely aligned with CEN as well – that is to say all data would be clusters with more attributes where ITEM_TABLE was used and more features in software.
I am interested in the Techo’s response to this idea.
Cheers, Sam