thomas.beale at openehr.org
Tue Apr 11 10:30:57 EDT 2017
For those who are interested in the background....
the underlying design of the Entry types
were the result of an epistemological analysis of kinds of information
that could be generated in clinical work. That required an analysis of
the 'clinical cognitive loop' as I now think of it, which is similar to
a scientific investigation (theory building), but goes further and then
intervenes in reality, based on the theory of what is going on.
So briefly, the types are:
* Observation - data gathered, including what is called 'subjective'
data, e.g. patient-reported pain. It is assumed that both manual and
machine means are used to gather data, and that the data may be of
unlimited complexity. However, the data are 'about' the one
individual, i.e. the patient (patient's kidney, skin, CV system, etc)
* Evaluation - inferences made by human mind or machine, by comparing
the individual data to the current knowledge base - i.e. standard
medical knowledge, diagnostic guidelines, etc - any knowledge that
is 'about' categories, e.g. 'patient with high BP', 'patient with
raised blood glucose 2h after challenge', etc. These can be
understood as some kind of 'opinion', since there can always be
errors in matching the observation to the knowledge, or even knowing
which observations matter and so on (think: episodes of House).
Typical clinical words for Evaluation are 'assessment', 'diagnosis',
but even just identifying a 'problem' is a kind of evaluation. An
evaluation that has been made can be considered some kind of
clinical decision on which actions can be based.
* Instruction - request to various actors to perform specific actions,
based on the Evaluation(s), typically medications, other therapies,
education, further observation.
* Action - record of actions actually performed.
* Admin_entry - record of administrative statements recording passage
of patient around the care system.
The actual data structures we defined for these types are in some cases
quite structured, e.g. Observation, Instruction, Action
This was based on the idea that these kinds of Entry have data of a
certain shape. For example, Observations are by definition time-linked
samples in past time, so the data structure is a time-series.
Additionally, the data/state/protocol triple format mimics the reality
As we gained experience with these Entry types over the years, it became
apparent that there are some grey areas, e.g. path labs and radiology
can send back results containing 'interpretations' and so on. If we were
to redo the analysis today, we might potentially have somewhat more
flexible structures, and rely a bit more on archetyping to specifiy the
epistemic category of each Entry. However, I would say that the current
system has held up pretty well, and the grey zone is pretty small.
It should be additionally understood that concretely defined data
structures are what make it possible for software engineers to build
clinical software that can work with the data.
Compare this with software based on most of the extant message formats
or other more freeform models - there is no support for even the most
basic things like time-series of data, or state machine transitions of
orders. While there are sometimes debates about whether some data item
should be an Observation or Evaluation, once the choice has been made
(at least for international or national archetypes), everyone has the
same representation of that data, and software is guaranteed to be able
to process its basic structures (time points, data/state/protocol etc).
So, of course what is there is not perfect, but it seems to have held up
reasonably well, and new additions like Task Planning
<http://www.openehr.org/releases/RM/latest/task_planning.html> will make
Here are some FAQs about the Entry types
that talk about the grey zone etc.
On 11/04/2017 01:54, Bert Verhees wrote:
> It is clear to me.
> Not only our senses observe, mostly we use devices to observe.
> Sometimes very complex devices, like MRI, their are a lot of
> calculations in those devices, billions of calculations.
> BMI is something that often takes two devices to measure, but it could
> be possible to create a single device which does this. So in that case
> it would be an Observation? Than it will also be if not a computer
> does the math, but a human being.
> To judge however if a BMI is a sign of obesity, that is subjective.
> For a normal person, a BMI of 28 could be obese, but for a
> weightlifter, because of all those heavy muscles, a BMI of 30 still is
> not obese. So that would be evaluations.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the openEHR-clinical