| 60min | ANF Modeling | Group | - Keith: He needs to be ensured that we will make progress over the next two weeks. He will be out of the office all next week. Joey: He isn't sure if Keith is aware but he wasn't working on this project during September but is back now.
- Joey: Going through the drug information, the drug name may or may not include a strength, may imply a route, etc. If they are not included, they have to end up in the circumstances. Keith: This is the part about being productive. You had an assignment last week to bring to today's call some examples of how to divide topic from circumstances and technique. If you have specific examples about your struggle with this, let's discuss.
- Joey: He is struggling to understand even what technique is. Keith: We have to define it for our process. We have clinical statements and need to decide where to put them. CIMI has a clinical statement model that has topic and everything else. We had an exercise to understand what should be part of the topic. If we can't decide that, the model is worthless. We must have supporting documentation with a clear definition of topic. Can we come up with reproducible rules? That is part of what he has been charging. He is a little frustrated because the point is you have come up with things that need to be broken out but you need to come up with the editorial rules and principles for how to do that. Does that make sense to you? Maybe he hasn't clearly communicated the rules. Does everyone understand the rules? Patrick: Yes, he believes so. The idea is for anything we need going forward, we need to have rules and principles for doing it. Keith: John, this is somewhat consistent with your concern two weeks ago, which was we don't have these rules yet. Do you agree that we need specific guidance for the distinction between topic and circumstances? John: Yes. Keith: You will represent someone on his side that represents a skeptic. There is a proposal that the CIMI model might and aspects of it are a starting point. John: Understands this as the starting point.
- Keith: The second step is to say we want to constrain the CIMI model as an ANF, and we recognize that a lot of the CIMI model might be what you recognize as CIF. They are based on what a clinician would naturally see, grouped together as part of the process of entering data into a form. For ANF, we might want to group measurements, such as BPs, as a consistent way of representation. We want to come up with the top-level categories of things, and the strawman is that the 4 categories are: action request, action performance, phenomenon measurement, and phenomenon measurement goal. Thus far, they have held out for every scenario presented to us except for one posed by Bob Greenes, where we might need a justification of other than 1 of the 4. But, we don't need to do that for the use cases we have before us. The action performance is less critical right now than the other 3. Do we have consensus that those 4 top-level categories make some kind of sense? Joey: Yes. He has things he would like to discuss that kind of covers what you are talking about now. Keith: He wants to finish so circumstances may differ, such as for phenomenon measurement and a goal for that phenomenon. You can identify a set of circumstances for each, but you wouldn't have to change the topic statement for that phenomenon; rather, it's just that the circumstances are different. An example is for systolic BP:
- In the simple case, it's pretty easy. Example: Topic=systolic BP; Circumstance=measured by X; Value=150. Other things are right brachial artery, for example. Are they aspects of the circumstance or the thing being measured?
- We might create a goal for this patient, such as within 30 days, have systolic BP from right brachial artery to be <130. Which part of this goes into the topic and which into circumstances?
- For this and other things, such as for medications, we need to say what is it, such as an action request, and which part represents the topic and which the circumstances?
- John: For each clinical statement, it's a brand new world to him. Keith: You have to iterate through a set of principles that ideally will be general and simple. That is the goal. For example, we could say that anatomy is always part of the topic or always part of the circumstance. Kirsten has worked with Claude, Stephanie has worked on it to some degree, but we can't keep starting over each week. For example, "Head CT with Contrast," is "head" part of the topic or the circumstances? Previously, Stephanie showed him a questionnaire on providing different use cases and can they be properly parsed into topic and circumstances.
- Keith: Does the goal make sense that we have simplified aspects of the CIMI model that has 4 general areas and we are required to define what goes into the topic, circumstances, and for circumstances, how to break them out?
- Joey: Yes, he understands the problem.
- Patrick: He isn't sure he will be on the call next week since Claude will be back but it makes sense to him. Having use cases and principles makes sense. Keith: Part of this is we must have the ability to iterate. Patrick: We need to have lots of use cases. Keith: Agreed. That is why he has been pushing to have the KNART white papers put into use cases. Stephanie has taken the cardiac one and started work on it. Patrick: He was thinking that they would definitely apply immediately to this.
- John: It sort of makes sense to him. Two challenges he see are: 1) For any clinical area, such as a CT scan of the head, we can come up with topic and circumstances but he doesn't know that we can all agree them. 2) The work may not transfer well to other clinical areas. Keith: You're going to need to suspend disbelief. We need to go through the process to see if we can attain it.
- Sarita: When this discussion first started between Kirsten and Claude, she was on vacation for a couple of weeks. For the time time today on this call, she is finally having clarity. She understands what he is presenting. We do need to work with the use cases to define the model. They have already been working with some of this for the cardiology use case. Keith: Yes, the process should be use case-driven, architecture-centric, and iterative and incremental.
- Kirsten: Yes, it makes sense to her. Going forward though, we'll need more clarity. We need to make some decisions on the fly, such as we think this is part of the circumstance. Keith: When you make a decision like that, there needs to be documentation on the principles as to why you made that decision. His job is to run interference so you can focus on the end game. Kirsten: Understood. Mentioned the documents we have been working on (missed part of what she said).
- Stephanie: She understands but will also need to suspend disbelief. She thought we were close with the last iteration of the document but John raised additional concerns/comments that need to be addressed. We need to come to an agreement on how to do it and it may not be perfect but we need to move on. Keith: You can pursue correctness forever. We need to be consistent, even if we are not correct. The whole premise of informatics is if we can't get reproducible data, there is no point spending money on KNARTs or CDS.
- Joey: Basically, you want to create an algorithm for what Kirsten and Sarita are doing. Keith: Yes, but it's based on reproducibility. Joey: We need to review the data they have produced, know what the specified rules are, and whether they have gone through the data and broken it out. He can review what they have done to understand their thinking. Keith: The nitroglycerin use case is to be used to propose what the things are. Is there an independence of techniques, such as when route of administration is a standalone technique or a sub-part of technique, and come up with a proposal that builds on the work that has been done? But the work that has been done doesn't cover aspects of that use because it has a lot intricacy. Joey: He talked to Claude first and reviewed the information with him. This has now crystallized in his mind.
- Joey: Displayed and discussed the work he has done on this.
- Keith: For documents like this, we need to ensure we are working in a common environment that is not Google docs.
- Keith: Doesn't think the state of the subject should ever be part of the topic. Joey: We have been putting the range for the topic, such as for a BP, as part of the topic.
- Keith: The topic needs to be representable within a clinical terminology using EL+ semantics using (?) domain. So a value is excluded from that because you get into things like absence, which cannot be represented in OWL EL. We need to separate the thing being measured from the measurement. Patrick, do you agree with what Joey said goes into the topic? Joey: All of our TSRs up to this point have included the value in the topic. Keith: No, the TSR has a place for value, and maybe you can display one, but his understanding is they are distinct and separate. Joey: It's not the topic code. The topic code is basically the conceptual stuff you are talking about and then there is the value that is part of the topic but it's not the topic code. Keith: The topic expression is the topic and everything else goes into circumstances. That is pretty fundamental from his perspective. Sarita: Agrees with Keith. Keith: That needs to be fixed. Patrick: If we are going to make these changes, they will drastically change the TSR and he definitely needs to attend these calls. Keith: He disagrees that it drastically changes the TSRs but yes, you can attend these calls.
- Keith: There is the topic and everything else that goes into the circumstances. For topics for phenomenon and phenomenon goal, they should contain the same kinds of things and the same for action request and action performance. Those need to be principles that are clearly understood. Topic is a class, but it's not an instance. It becomes an instance in a particular circumstance. Joey: Then there really isn't a need for topic. Keith: That is a point that John was making but we need a compelling reason as to why it won't work.
- (Missed part of the conversation.) Keith: Asked if there is something in the prior ANF document that states what is shown in bold above. (See what is shown below, which was included in the September 29, 2017 version of the document. It seems to be implied that the topic for both types of action statements and both types of phenomenon statements would be the same.)
- The Topics and Circumstances applicable to each type of clinical statement are shown below.
Clinical Statement Type Topic Circumstances Action Request Procedure Request Circumstances Action Performance Procedure Performance Circumstances Phenomenon Measurement Phenomenon Assessment Circumstances Phenomenon Measurement Goal Phenomenon Goal Circumstances - Joey: In the topic, he thought the technique was part of it vs. the circumstances. Keith: That is an open question - is route of administration part of the technique or not? He can create an expression that for acetaminophen 500 mg by mouth and put it in the topic or it could be part of the circumstances. This is not defined well. Again, we need to define that, see if it holds up for the various use cases, and adjust as needed. For the nitroglycerin example, there are lots of pieces that need to be determined if they are part of the topic or circumstances and if circumstances, is it for technique? Patrick: For technique, it would make sense to make it part of the topic if it wasn't dependent on any circumstances. Keith: For systolic BP that needs to be taken 30 minutes after patient has urinated, is it part of a precondition or is it part of a technique? Patrick: A precondition is something that everyone agrees to the default vs. a technique that is arbitrary and could be done differently by different people. Keith: But if the outcome is the same regardless of technique, it doesn't make sense to record it. Patrick: but, we could be measuring to determine if the outcomes are in fact different. Keith: This is something we're seeing with hypertension being over-diagnosed because people are uncomfortable due to needing to use the bathroom when their BP is being taken. The challenge is driving down the middle of the road vs. down the shoulder and making sure you don't drive off the road. How do we create a potentially arbitrary rule that says this is when a precondition is part of the technique, for example. More work is definitely needed. He is not confident that we'll have everything that is necessary finished in the next 2 weeks. We need 1 architecture, not 10; it needs to be use case-driven, architecture-centric, etc. We need to work through the nitroglycerin example fully. Reproducibility is our measure of success.
- Joey: He just watched SNOMED learning. They were saying the observable entity is the question and the finding is the answer. Keith: Disagreed. There is no reproducibility between the observable entity and the finding, and he demonstrated this at the last meeting. He went through this with Kent Spackman between finding and disease, and eventually he gave it up but thought we hadn't tried hared enough. SNOMED may continue to say the distinction does exist.
- Keith: Asked Joey to let him and Stephanie know if he runs into any roadblocks and to please make progress on this for Friday's call.
|