Date

03 Nov 2017

Attendees

Scribe

Stephanie Klepacki

Goals

Review the results of the sample clinical statements that Joey provided and the results of how he and Stephanie proposed modeling them using revised ANF modeling guidelines.

Discussion items

Time	Item	Who	Notes
5min	Overview of Exercise	Stephanie	NOTE: Reference the attached email that contains the sample clinical statement that Joey provided and which he and Stephanie modeled. Joey's descriptions of how to model them are shown in black text, with noted differences between how he and Stephanie modeled them in red text. Stephanie: Keith, there are some differences for which your input is needed.
55min	Discussion of the Modeling Exercise	Group	Joey: Looked at the things that were always present in order to decide what should be in the topic. For the 3 BP examples, that was always "Systolic BP." Stephanie wasn't breaking out as a third component of Result (i.e., on same level as Topic and Details) to put things like the result/value and unit of measure but instead she included them as details. Keith: You both understand it and are doing it exactly right. There are some minor differences. He agrees with the grouping of the result and unit of measure. But, when he talks about unit of measure, he would still have a unit for something like a ratio. Joey: If you are using data types, it gives you context as to what the meaning of the value is. If the value is a data type of ratio, then (did not capture the rest of what he said). Keith: (Described a scenario with Joey that was not captured in the notes.) Keith: The intent is to make it that there is one form that the person that runs the query on a database is more likely to get accurate results. We might make a rule that you can't have concepts as part of the result. It is kind of like Jeopardy where you state the answer in the form of a question. Joey: What are the units, because you said there are always units, even for presence/absence? Would it be a count unit? Keith: Yes, or something, but for now, he's just treating it as a count. An example is you have at least one observation of diabetes for a patient. We need to discuss for something like Diabetes Mellitus, where it can be either present/absent/or I don't know yet. Is this making sense as to what he is thinking for standardizing the result? Joey: For coded ordinal results, such as for deep tendon where it's 1+, 2+, etc.? Keith: That is one we'll need to work through. You could say that the criteria meeting scale 4 is present. We might be able to do better than that and say the deep tendon scale reflect of 1-5 is a tendon unit. A lot of these ordinal scales are in fact quantitative but with a different resolution. In other cases, the distance between one and another isn't reproducibly quantifiable, such as is the difference between a value of 1-2 the same as a difference between 2-3? Joey: Both could work in that case, but he thinks one is easier because there are numbers. He would need to look at more ordinal scales but after a quick glance, he thinks it could still work. Stephanie: Asked for clarification of result - is it to be a component that is on the same level as topic and details? Keith: Yes. Keith: Wants Joey to: Work on an example clinical statement that has: upper bound, lower bound, precision, or concept that represents proxy for units but includes odd things like ratio. Keith wants to include precision in cases where there is a reason to say the precision so the upper/.lower bounds can be truncated. It also makes it consistent with HL7's interval data type. Explain and document the difference between technique and precondition. Find a variety of use cases to use for the next steps and include those in the appendix use case document that has multiple uses (see 2nd attachment). If there is something that needs to be added to the use case, see if there is a plausible way to do it. The nitroglycerin example is a good use case, particularly he would like for Joey to show how to model it when a patient takes two nitroglycerin pills but it has no effect. Need to consider if data needs to be represented in some way (is it useful)? If not, and we can say why, then sticking with one side or the other on that boundary is appropriate. Stephanie: There are a few issues on the email that she sent Keith (see 1st attachment) that need Keith's input, where she and Joey had differences in how they would model a clinical statement. Keith: Let's determine a time to meet on Tuesday. Copy Susan so she can help find a mutual time.

Action items

Joey Coyle: Work on an example clinical statement that has: upper bound, lower bound, precision, or concept that represents proxy for units but includes odd things like ratio.
Joey Coyle: Explain and document the difference between technique and precondition.
Joey Coyle: Find a variety of use cases to use for the next steps and include those in the appendix use case document that has multiple uses (see 2nd attachment). If there is something that needs to be added to the use case, see if there is a plausible way to do it. UPDATE 11/30/17 FROM JOEY: The use case document seems similar to the KNART work where you need to extract out the clinical statements from a scenario to then model in the TSR. Seeing that we already have a team doing that for us, and they have produced many un-modeled examples, I think we would move quicker simply trying to model what we have. The other advantage is that I can use these completed TSRs to begin to generate the L3 KNARTs.

Browser not supported