Hi, Sam
During our ACPP meeting, you mentioned Alan Rector's remark on "modelling of belief" and "modelling of use". I checked Alan's homepage, but didn't find a paper (at least from the title) on this topic. If you have a pointer from Alan or someone else, can you please share with us?
During my searching, I came across this article " Clinical Guidelines as Plans: An Ontological Theory" [1] by Kumar et al. The ideas presented in this article echo a part of what we are trying to achieve here. I have the following points to add:
1. Need more relationship than classification
"Ontology of plans" containing classification of tasks, roles and parameters of a plan is an important part of explicit medical knowledge. However, not all medical knowledge and plan execution constraints can be expressed via classification. Mereology relationship is also an important part of medical knowledge. We will also need to identify some unique properties in modelling Adaptable clinical protocols pathways, for example, hasExpectedOutcome, hasMedicalGoal (intentions). RDF and OWL have equipped us to express those relationships.
2. Need rules and policies
We will need to use rules, and even proof in this area. The rules can be grouped according to their purposes and domain. For example, in the stroke management protocol test case, we will need rules to handle:
- How to selected tasks for a patient with one or more clinical problems. It could be via matching the patient state or a process state with the initialCondition described for a task
3. Different layers of knowledge and rules
We could further add different types or layers of rules to accomplish tasks such as
- How to validate the prescription to avoid adverse efforts when different tasks/drugs are recommended for the patient
- How to optimize resource allocation, maybe to turn on resource constraint rules when querying for tasks of a plan
4. Need a HL7 RIM Ontology?
As illustrated in the stroke management test case [2], we will need to describe tasks (act) and their attribute, participating entities and their roles. HL7 RIM, vocabulary and various schemas provided a significant knowledge base for describing concepts and relationships in Healthcare domain. We really should try to use HL7 terms as much as possible in our ACPP model. As you are also activity participating in HL7, maybe we can discuss if the HCLS group is a good place to have a HL7 RIM Owl ontology - we might need to take a close look of HL7 concepts and their semantics. I doubt an owl file that is directly translated from HL:7 UML diagrams and xsd files will be suitable for semantic web
I hope others on the HCLS list can shed some light or share their experiences on this subject as we move forward.
Helen
http://www.agfa.com/w3c/hchen
[1] http://ontology.buffalo.edu/medo/Clinical_Guidelines_as_Plans.pdf
[2] http://esw.w3.org/topic/HclsigDscussionTopics/HclsSubGroupACPP/Stroke
