Discuss Claude's assignment and Keith's discussion with Stan Huff.
Discussion items
Time
Item
Who
Notes
5min
Web Meeting Information
Keith
Keith: Stephanie, we can setup recurring meetings in Ring Central and there won't be a code required for a leader.
30min
Claude's Reproducibility Assignment for Visible/ Indivisible
Keith
Keith: Claude, you had an assignment regarding the reproducibility of whether something assigned is visible or indivisible. Have you made progress on it or conclude that it is futile? Claude: Joey and I have discussed this. Joey: He doesn't have the panel vs. component assignment. He did the TSR assignment. The main difference from what Keith discussed last week is that the heart example is a normal physiological thing, whereas a wound is an abnormal finding. We do by location for millions of models. Keith: Is it normal or abnormal? Are women pregnant more than 50% of the time? Joey: Not for a normal woman. Keith: So that's not a normal state for women. The point is we're getting into the dangerous value judgment territory and it's not necessary. Is this really what we want to spend our time on? He has been arguing this for 15 years. Claude: We should time-box this. We are about to submit the model to HL7 and some of these discussions could be helpful. Keith: He had an hour-long conversation yesterday with Stan Huff, and Keith's perspective is not to contradict what Stan is saying. There is some nuance here. He is happy to launch into this and he can start with his explanation of where he went with Stan. Joey: He would like to hear what you have learned from Stan. Claude: Proposes we discuss this later in the call. It will be interesting to see how this reconciles with some of the things.
Keith: For a heart murmur, it is typically abnormal. Claude: He will venture a reproducible (missed part of the conversation as had to step away for a minute). The question is: why does it matter? Why do we need to say that this lacerate or macerated edge is part of a wound? Claude: Because it's a piece of information that further qualifies a finding, it's not a body structure but a finding. The laceration property of wound edge, etc., really has no meaning outside of what they are qualifying. They are basically providing some distinguishing characteristics of another part of the body. Keith: You're saying we would never want to search for a standalone wound edge. Claude: You can, but the querying is a little different. For example, I can search for all lacerations reported for this patient and it would return the value of laceration but each laceration would need to be tied to a wound. You could search for something such as foul-smelling discharge. Keith: Is the discharge always associated with a wound? Maybe someone is getting ready to go into isolation. This could be from an abscess, from the eye, from the nose, etc. Thus, discharge is the ting being evaluated and only being able to query that in association with a wound makes it harder. It could be the sinuses but not the nose, etc. Is starts to get complicated very fast. Claude: (It could be ) a qualifier of foul-smelling discharge and it could return or you could say me all findings of foul-smelling discharge. In this case, you would need two queries to find it: 1) give me anything that is a finding not tied to another finding or 2) give me everything tied to the component of foul-smelling discharge. If it is a qualifier on another finding, you would get everything back you could further refine as needed.
Keith: You are making the case that the distinguishing between an indivisible and visible finding is a real one, that wound edge is indivisible from the wound itself. Joey: You are doing it as a matter of convenience. you could basically say they are all independent findings but this patient could have 20 wounds. You want to be able to label them and describe them. Getting that information together makes it easier in the EHR. Keith: So for convenience, we can make a data structure that a user fills out and it becomes very onerous. Stephanie, have they made a case for this? Stephanie: It depends on how much information you want to capture without the clinician becoming frustrated with too much detail and not documenting anything. Sarita: You wouldn't have a murmur outside of a heart. Keith: (Stated something but I wasn't able to capture it.) Sarita: Again, she is concerned about making it too difficult for data captured but she understands the convenience intent. John: A wound edge from a circular saw (i.e., ragged) vs. from a different type of wound, such as from a razor blade (i.e., smooth) would be very different. Keith: The representation of a wound edge as being separate from the wound itself as being indivisible but he doesn't think that is what we are saying. Do we have to build something statically into the model? Again, he and Stan had conversations about this. Catherine: Maybe murmur isn't a good comparison but something where you could have more than one source of it in your body, such as retinal detachment when you would need to know which eye. We need to be able to link things together. Should we think of wound having a qualifier? Is it hierarchical? Rather than having wound with a qualifier, you could have a finding of discharge and it's associated with another finding. Keith: He thinks she is hinting at what he is alluding to. Claude: From Stephanie, we need to be careful not to get too complex. There is the notion of inherence of location. Here is my take: there is a difference from specifying where something happens vs. describing an aspect of some finding. Body site provides information as to where something occurred, such as a rash. For wound edge, we are describing the wound vs. the location. If we are to represent wound edge and keep it simple, we have tried to model it but there are consequences. Either I need to know where the wound edge points to or I need a pointer (not sure this was captured correctly). We need to define the relationship that qualifies this. It raises a question: we now have to tie these qualities back to the findings - is there a steadfast rule that can be applied? Keith: Wound edge is just a proxy for the bigger argument is trying to get to. Joey: If I did a BP on a patient while standing, I can do a patient standing finding, include the time it occurred, and associate the patient was standing when the BP was done. But, that is inconvenient. Keith: Right, but patient standing stands alone. We are closer to and better at describing the issue at hand.
Keith's Discussion with Stan Huff
Keith
Keith: Had a discussion with Stan Huff and other participants were also on the call, including Susan.
Keith: He is fine with having indivisible and divisible findings on the CIF but not on the ANF.
Keith: Basically, Stan said that modeling from first principles does not work. You cook up the theory in your mind of what the world should look like and you force everyone to implement it. Keith doesn't look at it that way, rather than through observations, process, and examples and we arrive at what is true. Stan said there are no principles we can point to up front. Where Joey said it's a matter of convenience is basically what Stan was saying. The examples he gives are always easier, such as searching for "50" makes no sense. Searching for dispensing of opiate medications with count >=50 makes sense. How do you draw the line of something that makes no sense outside of a finding with something that does make sense, such as for wound edge? Another example is associating retinopathy with diabetes. For the indivisible issue, if in fact it's an observation that stands alone, it has the potential to be divided. Whereas if you have a qualifier, it's more clear-cut that the anatomy cannot stand alone as a statement; it's like a sentence with no subject. His belief is that at the moment, it's not indivisible. Question to this group: Can you understand the concerns of the argument?
Catherine: She understands the problem and is interested to hear what Claude says. (I missed the rest of what she said.)
Stephanie: Agrees with Catherine. She is still getting up to speed on this.
John: He likes what Keith is saying, and it would be good advice to make this distinction in practice.
Sarita: Understands the problem and a similar issue came up with LEGOs regarding pressure ulcer. She anticipates we'll have the same type of issue for the KNARTs. Keith: Yes. (Missed the rest of what he said but he mentioned observable entity.)
Joey: A panel can contain individual information that is standalone, such as a BP panel that contains two independent observations (i.e., systolic and diastolic BP) that are collected together but put into a panel. Another example is the CHEM 20 that are all different lab tests but stored up at the panel level, like specimen. In the wound example, if this something that could be stated, it could stand alone and we could model it and have a wound panel and everything you could state about the wound would go into it. For example, the location of the wound would be at the panel level. Keith: If you do that, you don't have to make the distinction of what is indivisible and what is divisible. Gave an example of a fistula and tunneling that is the same phenomenon but is it really associated with a wound?
Claude: Two things: 1) Depending on the approach we take for the ANF, we have to make sure we can go back to the CIF and vice versa. 2) Of course we could do the panel approach that Joey described. You would include attributes and presence/absence but it has some problems. The panel is a convenience for grouping purposes but when you do it, the idea is that each component of the panel but, in the case of wound, the problem is now you can no longer interpret the wound edge independent of the wound and that is an important distinction. In the ANF, we need to somehow say it's related to this finding and you need to be able to go back, especially in the case of the CIF. This is especially critical for the component. Instead of the independent object, you could have a table where each row has a primary key. (Missed the rest of what he said.)
Keith: He wants to circle back to an assignment and if we go with the independent/dependent and we pass the sniff test for the panel idea, he wants to create a pros and cons sheets. A start is shown below. He would like for us to complete this. This could feed into reproducibility criteria.
Panel
Ability to independently represent present, absent, and unknown? Yes
Dependent
Ability to independently represent present, absent, and unknown? ?
Claude: He has a request for Sarita. We could take one of these paneling examples and see how we could think about in the ANF. In the CIF, we have a strawman. We can specify something is in the context of something else. For example, a wound is present but tunneling is absent. Keith: Asked Claude to take a stab at the pros/cons matrix. He may need assistance from the terminologists. Sarita: Said she can assist with this effort. Keith: It could be that you can't lump everything you want to say about the ANF. With the ANF, there is only one value statement, such as 1, present, absent, etc. If you want to say "I don't know if wound tunneling was present for whatever reason," that is un-representable to have one single finding, where part of it is known and the other part is unknown. To be able to associate findings, such as I know the wound's height but not its width, it becomes more like a panel. He is fairly confident we can represent this in a panel form. What do uber observations give us, what is the value? If CIMI wants to have an impact on FHIR and FHIR is skeptical with CIMI, we have to make a strong case. He knows there is complexity here that we have to be cognizant of and we have to do analysis so we know what we're getting into. Every piece of structure we want to add costs money and complexity.
Action items
Stephanie Klepacki: Change the separate recurring Monday calls from Skype to Ring Central.
Claude Nanjo and Sarita Keni: Take a stab at the pros/cons matrix.
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