Continue working on ANF modeling guidelines for CDS KNART terminology artifacts.
Discussion items
Time
Item
Who
Notes
10min
Open Discussion
Keith and Stephanie
Keith: Is there anything that needs to be discussed before we resume work on the ANF modeling approach? Stephanie: Yes, based on his comments on her proposed revised method where had 3 clinical statement types vs. the 4 that were previously identified and Keith's apparent desire to keep the 4 as is, she recommends we establish a business rule that the 4 types (i.e., Action Request, Action Performed, Phenomenon Measurement, and Phenomenon Measurement Goal) must be the ones that will be used and no recommendations for changing them should be submitted. Keith: That wasn't is intent. The idea is that we have a set of top-level, general models and that we don't sub-class them. For the moment, we have a set of working components which are the Action Request, Action Performance, Phenomenon Measurement, and Phenomenon Measurement Goal. Stephanie: She had proposed a different set of top-level models, which was outlined in her email to Keith yesterday. Keith: The struggle is with Action Performed and Phenomenon Measurement. Using BP measurement as an example, there is the measurement itself and the performance of the measurement.
30min
Review of Stephanie's Proposal
Stephanie
Stephanie: She tried to identify the basic things that clinicians do when they see patients, which she believes can be represented by the following 3 types of clinical statements:
Phenomena: Combines Action Performed with Phenomenon Measurement
We had not previously modeled Phenomenon Measurement statements and when she attempted to do it, she became confused as to whether the clinical statement should be an Action Performed or Phenomenon Measurement.
The intent is to eliminate the confusion and difficulty in knowing how to split the two or eliminate the need to repeat the same information in both types of statements.
Orders: Same as Action Request, just renamed.
Goals: Same as Phenomenon Measurement Goals, just renamed.
NOTE: After she sent the proposal to Keith, she thought about which clinical statement type would be used for something like documenting that a patient was provided patient education, such as a patient diagnosed today as having pre-diabetes and education materials were provided to the patient. To accommodate this, she suggested revising it to Goals and Patient Education.
Keith: He sees it as an evolution of the previously identified 4 classes. Stephanie: Agreed.
Joey: Referenced (did not capture the rest of what he said but it might have been about CEM).
Keith: Maybe we have topic, circumstance, and value as a separate object. We can think of better ways to do them but we are trying to do a standard process.
Stephanie: But don't the components have to apply to all 3 of the clinical statement types? She can see where value fits for Phenomena but how does it apply to Orders and Goals?
Keith: The goal of the classes is to define them into a set of disjoint classes and we can't subclass them. We're trying to eliminate the proliferation of 1000s and 1000s of classes,which severely impacts data analysis and data retrieval. What would those disjoint things be? It's not the case if you create an element called Value that it has to be in all 4 (or 3) of the classes. But, for those that include it, the idea is you would come up with an action object that has sufficient generality (missed the rest of what he said).
Joey: He understands not having proliferation of classes but does that mean that we wouldn't have a proliferation of constraint models underneath them? Keith: Actually, you would have proliferation under them - we've pushed out the proliferation to a lower level.
Joey: He has come to terms with the observable entity and finding issue both being in phenomena measurement. In what most people think of observable entities with a value and quantitative answer (basically measurements), every one of them have a data type, such as quantitative and maybe with different units of measure. A finding is basically something with a different data value that indicates presence or absence. You have coded labs, such as red blood cell morphology, which is different from a quantitative one. The one thing that would be useful would be to define a data type where all these meanings of 0 to 1, 1 to 0, etc., and that data type was used in these types of things. Keith: You have demonstrated an open-mindedness to get your head around this idea, and that is exactly what he is looking for. As we work on this, there may be fuzzy areas that we need to evaluate and work through and determine if we need to revise our evaluation criteria. Again, it's an iterative process we hold ourselves accountable to. It's not necessarily natural as to how a clinician may think about it in clinic but it's a regular and reproducible way. For ANF today, we don't have the rules and principles to say where that is. Example: Is route of administration in the same category for administering a medicine as for a psychiatric visit for providing patient education? This is an example of the type of issue we need to think through. If we can't say something is reproducibly disjoint, we don't want to torture those using our model and make them have to figure it out.
Keith: He perceives this as a real challenge and would like for us to think through it for the Friday call.
Keith: There are certain things about a measurement that are necessary to interpret it and that don't necessarily go into unit of measure. For example, a BP measurement. You have mmHg but you also have other information such as right brachial artery. We have some choices. Things like patient was seated when BP was measured and patient urinated within 30 minutes of it being taken could be recorded as part of the action performed statement (i.e., those actions were performed in order to take the measurement), as part of the measurement, or both. He can see merits for all of them. How do we choose which is right?
Joey: He sees them as always the same thing and any action performed is always a phenomenon measurement. Think of a finding as absence or a surgery with the value of it taking place (occurring).
Keith: This is exactly the kind of thinking he was looking for. He didn't go to the point of surgery having a value result but maybe phenomenon measurement never lives independent of the action. There might be problems with the goals. We have to figure it out but we solve the problem of not requiring people to make a choice as to whether it's Action Performed or Phenomenon Measurement. We could end up with 3 classes vs. 4 and the measurement is part of the action performance vs. a separate thing.
Stephanie: Agrees with Joey and this is exactly what she proposed except that Keith seems to want to use Action Performed vs. Phenomena as she proposed.
Keith: This is progress already.
Patrick: He doesn't know the domain very well but sees where we're wanting to go and pushing the complexity to a lower level vs. at the top level. This is the same approach he takes in software development.
Keith: He would like to set up a longer meeting for us on Friday. For that meeting, let's discuss:
Independent of whether we call it Action Performed or Phenomena, where we do combine it, how do we make it work?
Stephanie: She can work through the depression use case she started and send it to Joey.
Joey: He can work what we have discussed today using some of the TSRs. Keith: Agreed that this is a good approach.
Action items
Stephanie Klepacki: Work on applying the depression use case she started last week and apply it using the revised 3 clinical statement types.
Joey Coyle: Same as above for Stephanie except use the TSR use cases.
No labels
0 Comments
You are not logged in. Any changes you make will be marked as anonymous. You may want to Log In if you already have an account.
0 Comments