John,


| Thank you for taking the time to answer me individually.  An honor I do
| not deserve.  A couple of questions.  I don't believe that the "actual
| record" is in the DTD.  I'm not quite sure what you mean.

No, but you have to have a structure in order that you interpret something
for what it is. XML health records will be great to read and possible to
process in so far as the structure is agreed - this is the role of the DTD
(XML Schema is more likely to be used as it is much richer for data).

|  Second, my
| impression is that it is a triviality to translate XML into SQL or an
| object database, so that XML and a "database" are mutually very
| compatible.

Two issues - it is possible to translate the XML Schema into a database and
store the document - but this will potentially require a new database (or
set of tables) for every document.

Second - you can agree the schema. This is what we have done in GEHR - but
it is entirely generic and is meant to be stable. We don't want to change
the GEHR architecture often as this will require major rewrites for everyone
using it (the problem with HL7 messages at present).

So we have a third level - the archetype - which sets out what can and
cannot be in a Blood pressure - for example.

| That is, XML doesn't rule out a "database approach" does
| it?

No - but can you see the difficulty. I think the PRA mob set out to develop
a message in markup and that is fine - it is a good approach. Now they see
the EHR being a collection of these documents. This will need more thinking
if they are to get the functionality out of the system that we expect from
an EHR. It will also require lots of functionality of the XML database -
which may be on the way.


| Where am I coming from?  I have come to believe that open source in
| medicine, in order to succeed, must offer clinicians a "lingua franca"
| similar to what the C language offered the Linux and the Apache
| project.  That will not be easy, at all.  I think that something very
| like XML will be that "lingua franca".

XML is the data - not the "linga franca". The linga franca will be XML
Schema language - when people start to push that - then we will know if we
have a solution entirely in XML.


| Specifically, I don't think that
| any of the object languages, such as Java, Eiffel, etc., and the common
| object facilities such as CORBA will attract clinicians to this
| process.

I certainly do not think that XML Scema language will do that either. In
GEHR we see the clinicians working with a tool to extend and aggregate
archetypes to meet their needs. This is the clinical interface to GEHR - no
language at all. This work is being done by DSTC in Brisbane and is very
exciting.

Sam

| >
| > John,
| >
| > I agree, get on with implementation trials but not theoretical
standards -
| > and lets do it in an open way. We believe in GEHR that we will be able
to
| > read and write PRA documents but we have a database philosophy with a
fixed
| > schema and archetypes. XML will require that the schema and the actual
| > record are expressed in the DTD - this will be a challenge. Finally, the
| > security of a text based approach which is check by passing will always
be
| > more difficult to implement than a database approach if the aim is to
| > process the information.
| >
| > Sam
| >
| > ----- Original Message -----
| > From: "John S. Gage" <[EMAIL PROTECTED]>
| > To: <[EMAIL PROTECTED]>
| > Sent: Friday, 14 April 2000 6:19
| > Subject: Re: Proprietary Snares
| >
| > > The thing is that the version of the HL7 Patient Record Architecture
| > > that is currently being balloted is XML (non-proprietary) and because
it
| > > is document oriented and very close to the domain may be the best way
of
| > > representing clinical data.  Each of us has to get on a particular
horse
| > > and ride it.  Clinicians may gravitate toward PRA.  I await the storm
of
| > > controversy.  (UMLS and PRA and mutually compatible).
| > > John
| > >
| > > "Alvin B. Marcelo" wrote:
| > > >
| > > > John!
| > > >
| > > > You have just opened an interesting thread.
| > > >
| > > > Everyone knows about the UMLS. It is public domain and it was
intended
| > to
| > > > map terms lying under the same concept into one heading (a concept
| >    -- UI).
| > > >
| > > > Example:
| > > >
| > > > Query Term:      colon
| > > > Concept Name:    Colon
| > > > UI:              C0009368
| > > > Source:          CST95
| > > >                   MSH2000
| > > >                   MTH
| > > >                   RCD99
| > > >                   SNM2
| > > >                   UWDA99
| > > >                   AOD95
| > > >                   CSP98
| > > >                   LNC10M
| > > >                   SNMI98
| > > >
| > > > The UI itself is free (and so is MSH2000 which is MeSH). But if you
use
| > SNM
| > > > (SNOMED), then you will have to abide by copyright rules.
| > > >
| > > > The main criticism against UMLS that it is not as thorough as its
| > "sources"
| > > > (eg, SNOMED). But who cares? For our purposes, it may be sufficient.
And
| > > > the nice thing about it is that it is continually evolving.
| > > >
| > > > We, as a community, can decide to adopt the UMLS for what it already
| > > > contains...and as a community, add on to the UMLS metathesaurus for
what
| > it
| > > > lacks. A case in point would be the porting of the current SQL
tables of
| > > > the open source projects into UMLS-compliant terms...
| > > >
| > > > The best way to understand UMLS is to use it directly. You may apply
for
| > a
| > > > free account at umlsks.nlm.nih.gov
| > > >
| > > > Alvin
| > > >
| > > > At 04:38 13/04/00 -0400, you wrote:
| > > > >On the one hand, things like this in PRA are very satisfying to me
and
| > > > >to everyone else with a moderate to severe OCD, but they have a
flaw:
| > > > >the code system is proprietary.  So, I like the document model of
PRA,
| > > > >and I think it will prevail.  And I like the "attention to all
possible
| > > > >eventualities" spirit of PRA.  But I am dashed in the face by a
bucket
| > > > >of ice water (so to speak) by the discovery that I am being drawn
into
| > > > >proprietary-land against my will.  I guess that I could "register"
my
| > > > >own coding system with HL7, but I doubt that that would fly.
| > > > >Does GEHR operate at this level of detail yet (a level of detail
that
| > is
| > > > >very satisfying as I have pointed out) or does it exist at a level
just
| > > > >above this one and does it ensnare one with anything proprietary
(aside
| > > > >from Eiffel)?
| > > > >John Gage
| > >
| > >
|

Reply via email to