Hi, I think ADL has proved to be a pretty good knowledge modelling language and that the growing number of archetypes represents a good deal of clinical knowledge. What is lacking is the ability to define relationships and their types between nodes or even nodes in other archetypes. I don't really know if this is possible via invariants - but I think this should be pretty easy with the introduction of a few keywords and definitely would not imply any changes in the RM (I hope).
There are a number of situations where I think this might be useful: 1) openEHR is criticised in 'other' rounds for the lack of computable semantics inside the Archetypes and Templates and that they say delegating all the semantics to terminology via bindings is not a safe and sound approach. Although I am fully aware of the power of openEHR with regard to semantic coherence, I think there is nothing wrong to define some basic semantics inside the archetype. That is actually part of the domain knowledge. 2) It is possible to build a 'real' domain ontology - say mini-ontology without depending on other formalisms. Some of you might say what is the point of building a domain ontology with ADL...Well I can think of better validation during data entry, processing, GUI design, querying etc.-nearly all aspects. 3) It is one of our strongest arguments that Archetype do not need an external terminology - can rely on its own local codes. External terminology just adds more semantics into. So far so good - But what happens to a mission critical DSS application when terminology server goes down? Or, in most cases, a terminology for that concept does not exist at all? Or exists but in another language? The formalism is already there - 95% of work has already been done to make ADL a great knowledge representation language. So why not do the 5%? BTW I am by no means a knowledge management expert, but had some experience with other formalisms (especially Protege) for clinical modelling. Cheers, -koray