I just can't resist.

Rather than moving in a philosophical direction, my humble suggestion is
to build and use model(s) that serve the needs of effective/adequate
communication. There is a trade-off between descriptive power and number
of constraints placed on the model. Required/standard attributes act as
constraints on models and systems that implement these models.

In asking the question in the subject line (during a psychiatric mental
status exam), the intention is not to elicit an exhaustive list of
universally agreed upon attributes - but rather to examine the first few
similarities that come to mind.

Similarly, why do we care about modeling 2/3 tons of Apple vs. 3
cart-loads of Oranges (continuous/discrete and across variables) via a
set of common attributes (e.g. Data_value, Quantity)? Why should these
particular attributes be modeled differently from other attributes such as
Color, Size, pH, Taste, Shape, Price, etc?

More below.

On Mon, 24 Sep 2001, Bart Koppers wrote:
...
>  for example: a code (could be an integer, but a string would do fine too)
> or a link to a certain dosage of a drug, in which that dosage contains
> information like "twice a day", might not be a numeric value or whatsoever.
> It is however, quantifiable.
...

Bart makes a very important point. Once you introduce a
"subtype=Quantity", you are burdened with drawing numberous lines between
what are considered "Quantity" and those that are not-Quantity. This
necessary extra-work must be justifiable in terms of more
effective/efficient communication (for example, certain
operations/calculations are meaningful if the attribute is a subtype of
"Quantity".)

Back to the original question:

> > Another modelling question, as part of our development of the openEHR
> > "convergence model" for EHRs:
> >
> > In GEHR there is a DATA_VALUE subtype called QUANTITY, which models
> > physical quantities. See bottom for abstract semantics.
> >
> > HL7 has a type called QTY also, with similar semantics.
> >
> > The question is: do we really need two types, to model discrete and
> > continuous quantities? For example, DISCRETE_QUANTITY and
> > CONTINUOUS_QUANTITY.

If you followed my reasoning thus far, you will not be surprised by my
interest in asking - So What? What can I do with discrete quantity that I
cannot do with continuous quantity? Why should I even bother knowing the
difference between the two? Where is that list of terrible things that
will happen if I use "Continuous" for everything?

Even more importantly, what does Quantity mean (in the functional sense)
within the GEHR model? What can one do with attributes of the Quantity
type that cannot be done with Color attributes, for example?

> > Currently, the type of value in QUANTITY is REAL,
> > which theoretically accommodates INTEGER, i.e.
> > the possible values of discrete measurements, but it hides the true
> > nature of discrete v continuous thing being measured; in particular, we
> > have to add semantics to the class to allow it to be specified as
> > discrete or continuous.

I think I understand what you are saying. But, some examples will help.

Thanks,

Andrew
---
Andrew P. Ho, M.D.
OIO: Open Infrastructure for Outcomes
www.TxOutcome.Org (Hosting OIO Library #1)
Department of Psychiatry, Harbor-UCLA Medical Center
University of California, Los Angeles

Reply via email to