To me a “questionnaire” is a vague notion. There can be a lot of different “questionnaires” in health. From the GP’s in Thomas’s example to a Apgar score, to a clinical guideline and even a checklist. Those are all a set of “questions and answers”, but the scope and use is totally different. In paper questionnaires we will find a mix of many, maybe all, of those, crammed into what the local practice have found to be useful (= “Frankenforms”). To try to put all of them into a generic questionnaire-archetype is of no use.
Examples:
The GP questionnaire referred to by Thomas is in the quoted question about “ever had heart trouble” merely a help for the GP, and of little use for computation. But if it is supplemented by more specific questions, based on answers by the individual, then the final result can be “occasional arrhythmia with ventricular ectopics”, which is a relevant information for later use and should be put into a relevant archetype. So is it a “questionnaire” or a guideline for the consultation? Not relevant IMO, it’s the content, that’s relevant.
Patients with haemophilia in Oslo university hospital are offered a questionnaire online to register whether they’ve had incidents of bleeding, what caused it, if they needed medications and if so, the batchnumber of the medication. This is followed up by the staff both for reporting of used medication, and for the patients next follow-up out-patient control or admission. Questionnaire or not? Not relevant – it’s what the information is and what it is for, that is important. Find relevant archetypes for them, OBSERVATIONS or ADMIN-ENTRY for this, I guess.
Even checklists are a set of questions and answers. “Have you remembered to fill out the diagnosis?”. “Is there a need to offer the patient help to deal with the cancer diagnosis?”. Main thing is to analyze what the resulting answer is representing, and the use of it. Decision support? Clinically relevant? Merely a reminder? Put them into a template, using appropriate archetypes.
I agree.
‘questionnaire’ is many things, but not at the same time.
In any case any EHR needs to be able to cope with all kinds.
From ones with one or more qualitative results: such as the checklist
To the validated Score where individual results are aggregated in one total score.
It must be possible to create one pattern that can deal with all kinds.
An EHR needs to be able to cope with all kinds of data, “questionnaire” or not. However I’m not so sure a modelling pattern that works for everything that could be labelled a “questionnaire” is achievable, or even useful.
Modelling patterns are sometimes extremely useful, for instance for facilitating modelling by non-clinicians or newbies, but sometimes they aren’t very practical. One of the problems is that clinical information in itself is messy, because healthcare information doesn’t follow nice semantic rules. Clinical modelling must above all be faithful to the way clinicians need to record and use data, not to a notion of semantically “pure” models.
Finding “sweet spots” by identifying patterns that are sensible, logical, and above else work for recording actual clinical information is often an excruciatingly slow process of trial and error, exemplified by the substance use summary EVALUATION and the physical examination CLUSTER patterns of modelling, which both had taken years of trial and error long before I got involved in them.
If we can find patterns across some kinds of “questionnaires”, like clinical scores, great! However, since there isn’t a standardised pattern for paper questionnaires, it’s not likely that it’s possible to make one for electronic questionnaires. Outside the RM/AOM, a generic pattern archetype for every questionnaire with variable levels of nesting, variable data points, etc isn’t possible, nor would it in my opinion be useful. It would put all the modelling load on the template modellers, which arguably would be more work than modelling the same structures as made-for-purpose archetypes.
Some rules of thumb have developed over time though:
Model the score/assessment/questionnaire in the way that best represents the data
Use the most commonly used name for identifying it
Model them as OBSERVATION archetypes, unless they’re clearly attributes of f.ex. diagnoses, in which case they should be CLUSTERs (example: AO classification of fractures)
Make sure to get references that support the chosen structure and wording into the archetypes
In my opinion this pragmatic approach is likely to capture the data correctly, while at the same time minimising overall modelling workload.
I understand that at the clinical (health care provider) level each wants/needs to do its own thing in each unique way.
But:
Modelling Templates using recurring patterns (Archetypes) has a (a lot of) value.
Doing the same things the same has value.
In this way we can create an ordered way to query the data safely.
What means ‘the way that best represents the data’?
Is it the way clinicians see it how it is presented?
Or is it the way we can query the best, most safe way?
Requirements for both are not the same.
Questions in questionnaire are observations.
But what is it, when a Scale makes use of existing data in the database and calculates an aggregate result?
(E.g. BMI)
Isn’t the latter an evaluation of existing observations by means of a rule?
In the case of BMI the weight and length are real observable properties of a (human) body.
Question: Is the BMI an observable property? I think not. It is an aggregate, an evaluation.
Questions in questionnaire are observations.
But what is it, when a Scale makes use of existing data in the database and calculates an aggregate result?
(E.g. BMI)
Isn’t the latter an evaluation of existing observations by means of a rule?
Just on this question - a BMI might be computed, but it's still just a datum relating to an 'individual', as ontologists would say, so it's an OBSERVATION in openEHR terms. An EVALUATION is a opinion generated by comparing fact(s) about an individual to a knowledge based in order to classify the individual in some way, e.g. 'overweight'.
In the case of BMI the weight and length are real observable properties of a (human) body.
Question: Is the BMI an observable property? I think not. It is an aggregate, an evaluation.
not observable in the literal sense, but the view we take in openEHR is that an Observation is the apprehension of data relating to the individual, by means of examination and / or instruments. A BMI clearly falls under this category of information.
It can be the basis for an Evaluation if compared to some BMI normal ranges, to generate an Evaluation such as 'overweight', as above - this is an inference. A machine might do this.
Aside:
Someone will probably bring up scores like Apgar, and say they are some sort of inference, since the constituent scores are based on statistical/clinical pictures of what is healthy or not (e.g. HR >= 100 bpm etc). Philosophically speaking, this is true, and in theory they should be an EVALUATION. For practical reasons they are generally modelled as OBSERVATIONs, since they tend to act as a means of reporting physical examinations (in a quantitative way), and they get used as triage variables for determining which treatment path to follow.
In a more perfect world, scores might have their own ENTRY type, but I don't think the lack of it has done any harm to date.