2017-10-19 IA Small Group Call

19 Oct 2017

Attendees

Not Present

Keith Campbell (on travel)

Goals

Review and discuss Claude's recommended revised approach to ANF modeling of CDS KNART clinical statements.

Discussion items

Time	Item	Who	Notes
30min	Discussion	Group	Claude and Sarita: We are not clear on what ANF is and the differences between it and CIF. Stephanie: Displayed an explanation from a CIMI newsletter that quoted Keith on the differences between the two (see below). Source: HL7 CIMI-IIM&T Lessons Learned document, 8/28/17 draft version (sent by email from Stephen Hufnagel on 8/28/17). In recognition of this durable reality, CIMI has adopted a strategy that at first glimpse might appear to resemble a "have our cake and eat it too" type of approach. CIMI's architectural framework provides the capability to define 1:1 round trip transformation between pre- and post-coordinated representations of a given medical record instance. Pre-coordinated instances could be transformed to its post-coordinated form, and vis-a-versa, without loss of information or detail. Keith Campbell refers to these two forms of equivalent information as Clinical Input Form (CIF) vs. Analysis Normal Form (ANF). *Clinical Input Form* (CIF): is really nothing more than a synonym for the representation of clinical information which is highly pre-coordinated to optimize and limit the number of clicks, tables, columns, indexes and joins. Another example is the CIMI "Assertion" modeling style. *Analysis Normal Form (ANF):* is really nothing more than a synonym for the representation of clinical information which is highly post-coordinated to optimize for sophisticated querying, decision-support logic and quality metrics implementations. Patrick: The definitions are incorrect because where it states "pre-coordinated," that is really post-coordinated and vice versa. Claude: We should include the 8 questions that Keith has mentioned previously, such as who, what, where, etc., within the ANF modeling document. Patrick: Agreed, and we should include the principles that Keith has discussed previously. Claude: (Did not capture what he said.) Sarita: We have transitioned from the ANF capturing everything to separate, multiple fields. Claude: We should say what we think ANF is, which puts the burden on Keith to determine if it's wrong and why. By not writing down the requirements, we cannot determine if there is a conflict with them. Sarita: Yes. If you have a medication order such as "Administration of 250 mg of Acetaminophen 1 tablet oral daily," and I give you the post-coordinated expression in SNOMED, which is the ANF, and you create the CIMI CIF of that, then we determine if they are isosemantic and whether all of the information is captured in both formats. To me, that is something that would make sense to do. Claude: There is ANF in theory and ANF in practice. If I administer a medication twice daily, I could theoretically do it with ANF entirely. He proposes we do blue sky. John: He is leery about doing that. Sarita: We have been given an assignment and we have to complete it. How do we move forward with this work? Her suggestion is since Kirsten and Claude have spent a lot of time on refining differences between the topic and circumstances, for them to continue that work. Claude: Let's not limit to topic and circumstances as that may not be the right approach. Let's determine what we think is the right approach. We can keep a doubtful eye on the separation, and there may be a different way to do it.
85min	Review of Draft of Claude's Approach	Claude	NOTE: See Claude's email dated 10/20/17 that summarizes his revised approach that he discussed during this call (subj: Re: 10/19/17 IA Small Group). Claude: He would like us to walk through an exercise, as shown below. Clinical Statement: Administer 2 tablets of Acetaminophen 100 mg TAB PO TID This is what it could look like in ANF where the clinical statement can be represented by 1 single code by the terminology: CS.code = Definition status = Sufficiently defined IS-A = 18629005 \|Administration of drug or medicament (procedure)\| ( METHOD = 129445006 \|Administration - action (qualifier value)\| DIRECT SUBSTANCE = XXXXX \|Product containing acetaminophen 100 mg each oral tablet (clinical drug)\| FREQUENCY = XXXXX \|Three times daily (qualifier value)\| ROUTE OF ADMINISTRATION = XXXXX \|Oral (qualifier value)\| DOSE QUANTITY = 2 tablet ) CS.code = some expression for a respiratory procedure, lab procedure, etc. Pros/Cons Support for concrete domains Different terminologies for different things - ONC recommendations, not in scope of SNOMED No tooling support for SOLOR expressions Solution Pull out the object of the action into a separate field CS.code = some expression CS.other fields This is what it could look like for CIF where the CS can be represented by 1 or more codes: (NOTE: It is believe Sarita said this is how it would be represented in the information model, and one model is needed for each type of procedure vs. in the ANF, it would have only the 4-5 clinical statement types (e.g., Action Request) we have defined thus far, where those X types of procedures fit into 1 of the 4-5 clinical statement types.) CIF1: CS.code = 'Administration of substance' CS.medication = 'Acetaminophen 100 mg TAB' CS.doseQuantity = 2 tablets CS.routeofAdministration = PO CS.frequency = TID (3 x in 24 hours) CIF2: Some model for respiratory procedure CIF3: Some model for lab procedure Sarita: Agrees with Claude's approach and that is how she thought the ANF was supposed to work, which is a different approach that the current ANF modeling document is taking.
5min	Next Step	Group	Claude will continue working on the draft approach he discussed during this call. We'll regroup tomorrow at 11:00 ET and review it.

Action items

Claude Nanjo: Finish draft approach for ANF modeling of KNART clinical statements.