Representing knowledge in Archetypes/Templates or external rules

Technical (archive)
1

Dear All,
I have encountered a question regarding knowledge representation during archetype designing.
Let’s say a list of some special drugs that patient has used is important for Evaluation. I have two options here: First, to apply one external rule during the data entry, and mark the patient as e.g X-affected. Secondly, to represent this knowledge in Templates/Archetypes I am designing. It can be in the form of an Item-Tree-Medication-List using internal codes ( to specify medications that may be selected in the drug list). And to check this node at runtime or during evaluation of EHR for storage.
It may even be a third option that I have not been considers so far i.d using internal rules (First Order Logic) within archetypes.
The first option, sounds more general to me, and the benefit is that whenever I want to update rules, I don’t need to change the arcehtypes/templates. And the second option sounds more natural to me in that I express the knowledge in archetypes directly, makes it more understandable for clinicians and can be modeled by themselves.
Which one is more acceptable based on openEHR concepts?Do you see any flaws in these solutions?Is there any better idea?

Regards
Pariya

MSc; PhD Candidate
Department of Computing Science and Engineering
Chalmers University of Technology
http://www.chalmers.se/cse/EN/people/kashfi-hajar

2

Dear Pariya,

I would vote for the first alternative. If we start to mix up the information model and the decision support rules too much we will end up with a quite chaotic system.

Regards,

Mikael

3

Hi Pariya,

If I have understood correctly, you are suggesting something like this…

Relevant medication

Atenolol Y/N
Furosemide Y/N
Aspirin Y/N

This is a typical instance of a ‘Questionnaire-type’ question. In contrast, all of the current major health information representations use a ‘Clinical Statement’ pattern. e.g openEHR, HL7v3, Snomed-CT

Relevant medication

Medication = Atenolol
Medication = Furosemide
Medication = Aspirin

i.e the list is created as a series of assertions, rather than being a set of Y/N answers to a fixed list.

For interoperable queryability, it is important that the clinical statement pattern is maintained when the data is persisted e.g Asprin= Y must be mapped to Mediation=Aspirin.

Handling the negative e.g. Aspirin=N, is more problematic. In general, clinicians do not need to record negatives for long-term queryability but some negatives need to be recorded for medico-legal purposes and are more common in research questionnaires.

Part of the difficulty here is that the use and purpose of questionnaires is often highly localised and the mapping requirements for each Question to an equivalent ‘clinical statement’ have to be analysed on a case-by-case basis. Just be clear that this is a feature of Questionnaires and not of openEHR - a recent HL7-CDA paper describes an identical problem.

Our current working practice is to dual-model a Questionairre within a template, using a single archetype to represent all of the question-type nodes (often within section for clarity) and to model any equivalent clinical statement archetypes in a separate section.

e.g.

SECTION.Questionnaire
OBSERVATION.PreopQuestionnaire.v1
Relevant medication (DV_CODEDTEXT)
Atenolol Y/N
Furosemide Y/N
Aspirin Y/N

Relevant diagnoses (DV_CODEDTEXT)
Heart disease Y/N
Diabetes Y/N

SECTION Clinical Statements

ACTION.Medication.v1
EVALUATION.problem-diagnosis.v1

Unfortunately, this means that we have to specify how and when each Y/N answer resolves (or not) to a clinical statement.

In general questionnaires are horrible but they are a reality in modern clinical recording practice which we have to deal with. I think in future years we will find ways to automate the process a little better.

Ian

Dr Ian McNicoll
office / fax +44(0)141 560 4657
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian@mcmi.co.uk

Clinical Analyst Ocean Informatics ian.mcnicoll@oceaninformatics.com
BCS Primary Health Care Specialist Group www.phcsg.org

2009/6/1 Pariya Kashfi <hajar.kashfi@chalmers.se>

4

Hi Ian,

SECTION.Questionnaire
OBSERVATION.PreopQuestionnaire.v1
Relevant medication (DV_CODEDTEXT)
Atenolol Y/N
Furosemide Y/N
Aspirin Y/N

How this would be implemented?
Just by setting Occurrences to Repeating-Limited 3-3 for Relevant medication (DV_CODEDTEXT) can I have those three Y/N options? How would that be reflected in the GUI? what I can see now is just one time of this drop down list in the GUI represented in the Archetype Editor

-Regards
Pariya