I tried to post the following about a week ago, but it was rejected. It
may work now...
I have just joined this list, and it seems that a few pieces of
information might be
of use, going by the discussion I see. Apologies for repeating something
someone else
has said, but I haven't the time to read the whole archives..
Alvin Marcelo wrote:
> Again, dear researchers, you have to reach out to the practitioners in
the
> list because they are the ones who will give life to your models.
For the record, the current GEHR efforts underway (in Australia - see
gehr.org; in the
UK, see the Synex project, a CEN/GEHRish system on chime.ucl.ac.uk) both
have a number
of clinicians. Our Ocean kernel work in Australia involves 3 GPs
currently, all of
whom are practising and all of whom are very IT-literate as well.
> The list is looking for a model for an OpenEMR kernel.
>
> There are now five models on the block: GEHR, HL7 3, VistA, CEN, and
> CORBAmed.
Modern GEHR and CEN are not so far apart. We believe the latest GEHR
model deals with
a few things that CEN does not yet. I am going to develop a rough
equivalence table at
some stage. The GEHR model aims to be implementable, whereas, although
the CEN one
could be, it does not really help implementors too much. It also has a
couple of
things which we do not, which we may want to put in at some later stage,
but we are
going to conduct clinical software trials first, and test the waters.
> We won't go into which is better or which is worse (and why). If there
was
> a definite answer to this, then there would be only one model (the
best).
> But now there are five which means there is no consensus as of yet.
Yes, and it means that there have not been (nearly) enough prototypes
and clinical
usage trials.
> Knowing this, I ask again:
>
> Will the proponents of each of the models please...
>
> show us the bridge that links their models to an actual EMR
> implementation, ie, how do we use your model to build a simple primary
> care record system...
The GEHR model, as published on www.gehr.org (architecture doc) is shown
in BON
(Business Object Notation; we will convert it to UML when we get a spare
5 mins!);
this is a direct graphical representation of Eiffel. In other words,
each class,
relationship is found in an Eiffel class somewhere. If you look at the
classes in the
textual Eiffel IDE (we use the ISE tools) then you will see more
details. If you were
to look at the Ocean kernel implementation classes, which inherit
directly from the
GEHR classes (names are prefixed with "G1_"), they finish the job by
adding code. This
is normal seamless object-oriented development. It is nicest with
Eiffel, but is
certainly possible with C++, Java and so on.
> ....in SQL speak, how do we build standard SQL tables from your model?
Are you trying to implement an in-house solution? A better way to go,
based on the
fact that it is a comprehensive OO model is to use an object database,
which is
responsive to OQL (object SQL). Many clinicians have come to this
conclusion, due to
the difficulties they have had trying to squash the relatively complex
object
structures into a relational DB.
Another way to do it is to wait for the Ocean GEHR kernel, which will be
available as
open source, and will move to an open source mode of development
soon(ish). Around the
middle of this year. It will provide a full kernel component, with COM
and Corba
interfaces (including to accept SQL like queries).
Actually, you can implement a decent OO model over an RDBMS, but you
can't use 3rd
normal form tables, you hhave to use ann object/relational schema.
Fujitsu/CA's ODB-II
does this.
> Note researchers: the question is not "why your model is better
> than another". It's how to migrate from your model to the
implementation.
Just to re-iterate: this is just standard software engineering. It gets
done all the
time.
regards,
- thomas beale
