Le 05/04/2018 à 15:24, GF a écrit : > Philippe, > > Can I understand that you file-guide are patterns that fit archetypes > so Healthcare Providers can compose whatever they want. > The file-guides insertions are context driven. > The system of file-guides acts like an Ontology for > clinical/administrative content. > Archetypes define how things are presented in a system-interface.
I use a level-zero ontology (say a set of 55000 terms as implementation of atomic concepts linked inside a semantic network à la "colitis" -is a-> "inflammatory disease" and "colitis" -located at-> "colon"). All information stored about a patient are expressed as trees. As a dependency grammar, each tree "tells a story" (for example a colonoscopy report, an encounter note, the list of health problems, etc). Once that said, there is no need for anything else when it comes to process this information. Now, we could want to guide/ease the way it is created by practitioners. Mimicking openEHR, I built my own flavor of Archetypes as flexible information schemas (but much simpler since they just have to host a "model tree", a set of constraints (mandatory or mutually exclusive elements, etc), and the corresponding user interface). Fils guides are made for situations when there is no reference model and this is just a very simple (yet powerful) "trick". The basic statement was that when an expert builds a reference tree, you usually have to "cross her habits" (the first branches) before "reaching her knowledge" (what surrounds the leaves). The other problem being that, for example, when writing an encounter note where you deal with neurology and proctology issues among other topics, you can hardly navigate a neurology reference tree, then a proctology reference tree. Now, imagine that you ask the gastroenterologist to build this reference tree, then you break it apart as a big set of branches, and finally ask your expert to define, for each branch, the most general path that makes it a valid proposal. For example, the branch with the 3 choices "benign|suspect|malignant" would be attached to the path "*/mass/aspect" (* being a joker for "any sub-path"). As you can guess, it is now possible to put all reference trees into the same bag and, each time a user validates an element, to get her current path (say "encounter/objective/findings/polyp/aspect") and to go find in this bag the closest fil guide path that fits (if any). In order to do that, the system will first try to find if there is a fil guide which path is exactly "encounter/objective/findings/polyp/aspect", If not, it will try to find a fil guide which path is "encounter/objective/findings/polyp/X" where "X -is a-> aspect" in the ontology, Then look for "encounter/objective/findings/Y/X" where "X -is a-> aspect" and "Y -is a-> mass" in the ontology, etc. It may finally elect "*/mass/aspect" as the closest valid path and put "benign|suspect|malignant" as proposals in user interface. If the user selects "suspect", then the path she is standing is now "encounter/objective/findings/polyp/aspect/suspect" and the system goes back looking in the fils guides bag... etc As you can understand, the system of fils guides is neither an ontology nor an information reference schema. It is fit for systems that can "tell structured stories in the wild" because they are based on an ontology and a grammar. Finally, the nice thing with fils guides is that they are really complementary with Archetypes. No need to say that, instead of "benign|suspect|malignant" it would have been possible to attach an archetype to the path "*/mass/aspect". Instead of simply being proposed the three possibilities to move one step forward in the tree (the 4th possibility being, of course, to go get something else in the ontology or to write a free text), the user would see a form popup, click whatever needed and see the whole sub-tree being added to her "encounter note" tree. In my system, the native "free text control" is a tree filling interface. So the aforementioned mechanism can also operate inside the free text area of an archetype. The other possibility is to use fils guides when reaching the leaves of an archetype. Some leaves may be "hard-connected" to open another archetype, but leaves that are not may have the chance of being flexibly connected thanks to the fils guides. To sum it up: If you have a vocabulary and a grammar (say an ontology and trees as a dependency grammar), you don't need anything else to tell and process discourses. In the deterministic area, you can use archetypes to standardize the discourse (it will eases user's task, eases processing by smart agents, etc) In the non-deterministic area, users can express themselves by building trees by hand and you can guide them in an opportunistic way thanks to fils guides. IMHO, that's pretty straightforward. Philippe _______________________________________________ openEHR-technical mailing list [email protected] http://lists.openehr.org/mailman/listinfo/openehr-technical_lists.openehr.org
