Review and discuss the draft of the ANF modeling guidelines that Kirsten has created.
Discussion items
Time
Item
Who
Notes
60min
Review of Kirsten's Draft of ANF Modeling Guidelines
Group
Keith: What Kirsten is showing is exactly what he wanted.
Keith: We need to have a definition for a principle, such as for "procedure" and we may need a glossary. In HL7, timing includes repetition and frequency is often thought of as a repetition. Need to know if we should accept that definition.
Keith: We are not creating an exhaustive list but a list of categories. We need to number the tables so we can refer to them.
Keith: For ventilation order, there is an inconsistency. Device of nasal cannula is under instructions vs. including it as part of the topic. The thing being administered for medications is the tablet. (Unable to capture Kirsten's response.) Need to ensure it is reproducible and has the outcomes we want.
Keith: Is a topic a single thing or a set of fields? Talked about usage of constructors. Displayed an example of constructing a high-level object we would push in. For ANF, as long as it fits into a logical expression, we can pass to other objects. An instruction may be a logical expression or timing and priority are independent expressions. We pass in the known data, such as is it an internal or external data source. It would be in an array. It's designed to handle change at the level. Want to break it down into as small of pieces as possible but have an integrated layer. You have a clinical statement constructor and on one side you pass in a (missed what he said) and a circumstance.
Keith: Back to Kirsten's example, for instructions we want to know the top-level things that can be cleanly separated and that are general to every request. As an example, access route and access device are not general to every request and shouldn't be included as instructions but as part of the technique. Priority, timing, frequency, and indication are general to every request. For pharmacy, that technique might include route of administration. For technique, there is a technique that can be described for any request, and this is a hypothesis to be proven or disproven and if the latter, we do something different. We test it by using use cases. For medications, we have the nitroglycerin sublingual example and we need to put it somewhere. Given a generic request and we can specialize a field that is generic for a particular purpose, how should we handle the "PO" part? The topic would include "nitroglycerin 0.5 mg tablet," so for the part for sublingual, where would that go? Kirsten: It could be under technique. Keith: You could have a list of techniques to be applied or it might be a single expression.
Kirsten: Her concern is at some point, we will have to translate all of that into the use of certain terminology. For procedures where we are most likely looking at SNOMED, if we have to capture which kinds of classes and sub-classes, we can't do that in SNOMED. Keith: He would like for her to suspend disbelief somewhat but know that he has heard her concern. One of the challenges we have repeatedly is the rush to the solution before we define the problem. He wants a clean definition of the problem that is reasonably devoid of specific biases, such as SNOMED did it this way or we want to do it with the least amount of effort. Kirsten: The building of the TSR will occur next week. Keith: He accepts the heat, and you are doing as he requested and you have raised appropriate flags for delivery.
Kirsten: After our last meeting, she asked Claude about his thoughts on instructions and he wasn't sure. When they discussed it, he said they are going to expect one UUID for a procedure and one UUID for instructions. Keith: Yes to part A. For part B, the instructions side, if that's the best we can do, he'll be disappointed but that is what they need to meet their contract deliverables. Instructions are part of circumstance. We would have one UUID for each of those fields that encompass the terminology, such as one UUID for the technique expression, one for the device expression, etc. Preconditions would probably go into the technique. You probably wouldn't need a UUID for timing since it probably wouldn't be a terminology expression. Priority is probably a terminology expression. It comes down to what are the fields associated with a request for an action. For areas where you're not quite sure how to model an expression, you would substitute a UUID for it until it's better modeled. An example is route of administration for technique and you use the UUID for purposes of submitting a deliverable so they can be paid. Again, he'll take the heat for this and if it's one UUID for the instruction, he'll be disappointed.
Kirsten: Since Sarita will return on Monday, should Kirsten stop attending the calls? Keith: His preference is that Kirsten at least continue to attend a few of the calls for a transition. He welcomes her continued participation. Mike Lincoln has indicated that he will also attend the calls. At our last call, he had asked Sarita to provide a list of her issues/concerns.
Stephanie: Asked for a copy of the document that Kirsten presented today and for it to be emailed to the group. (NOTE: Kirsten did email it, and it is attached below).
Keith: Asked Stephanie if she was able to follow the discussion. Stephanie: Yes, except when Keith said timing for instructions would not have a UUID. Keith: Timing would not require a UUID because it is more of a field that is a specific field structure that is optimized for time. The DL is a powerful technique but isn't necessarily appropriate everywhere. Example: In SNOMED, they are going to create a concept for every strength of every medication that is manufactured and they want to include it in every medication concept. A better way to do that is to have a representation that supports numbers and algebra and use an integer that represents 100. We can do things like add them together, such as patient is taking 100 + 275 mg. Timing can have UOM, such as minutes, hours, seconds, or a specific calendar date and the type of calendar, etc. It may not be completely clean.
Keith: This is not ready to go to the "guinea pigs" on the Monday call. His hope is this time we can get it right. Let's get the circumstance and figure out what the fields are and instructions was one of those fields.
Add Comment