Calle Hedberg wrote:
> One of my problems with Thomas' Archetype strategy is the perceived
> difference between "domain people" and "IT people" (I'm not sure where
> "medical informaticists" and other hybrids belong here).
Simple enough: most people have their main professional interest in one area or
the other, so I would call them either clinicians or IT people. Lots of
clinicians know something, or are even proficient in programming; some very few
actually have degrees and professional experience in both medicine and
computing; for these, both names apply. I.e., I am not trying to exclude any
person from being able to wear two hats. When I say "IT people", I mean trained
and experienced software engineers, the kind who write the "hard stuff". So for
example, people here are interested in using things like Zope and Apache. It
takes a software engineer of some kind to write these tools, just as it requires
a "real" physician to actually treat another human being.
> My main problem with this is the perception that "the domain" somehow is
> capable of efficiently authoring and processing concept models.
Good observation.
> My experience is that the fracture lines usually cuts straight through both
> the
> "medical domain" and the "IT domain". To put it bluntly: Doctors and nurses
> are perfectly capable of arguing any concept or concept model to death -
> often because all parties are pursuing different interests. Similarly, "IT
> people" are equally diverse
Exactly. I never said the clinical domain would be efficient, or even clear in
their work. All I said is that if you at least give the domain a chance to
configure its own systems via formal concept definitions, then a) they have
control over their work, b) they can't complain that the IT people got it all
wrong, and c) software developers can get on and build software, without
worrying about having 100s of 1000s of clinical concepts in their software
models.
We're not about trying to improve human nature. (That's the follow-up project to
GEHR. I predict we get massive funding and have a train-crash catastrophe...)
> In a few cases, "IT people" might be capable of driving the development and
> processing of concept models. In a few other cases, "domain people" might be
> able to do the same. In the majority of cases, though, I think you need
> alliances of both. These alliances will shift over time - just like they do
> in e.g. mainstream politics - and the key factor should always be what
> drives the development and learning processes best.
Of course you need alliances of both, you are right, but in the methdology I
propose, how these people work together is much better defined, since for most
domain concerns, domain people do not care about the system at all - they just
know that it processes archetypes, and go on their merry way.
Of course, when specific systems are being engineered, then requirements are
written, analysis & design are done etc, and all the usual clinician/developer
involvement occurs. In some cases (e.g. open source) these are often
fortuitously the same people.
> To take a practical example: The medical establishment ("domain people") in
> SA has during the last 5-6 years made very little progress in developing a
> Data Dictionary for medical informatics. Six months ago this process was
> given a major boost by "IT people" cutting through all the committee
> hair-splitting and simply PRODUCING a Data Dictionary through
> merging/streamlining nearly 1,500 data elements from a number of national
> and provincial Minimum Data Sets into a set of around 700 data elements. It
> had significant gaps and errors, but for the first time users throughout SA
> had something they could use. The responsibility from following up from
> there was handed back to the "domain people".
I would suggest that the clinicians were not given the appropriate tools by the
IT people. Let's face it, IT people cannot build a workable clinical data set
from scratch - somewhere they have to get domain input, as it appears they did
in this case, by using existing data sets.
> For me, this had two lessons: Firstly, that both "domain people" and "IT
> people" will have different chances of making significant progress,
> depending on the concrete circumstances. Secondly, that 700 data elements
> defined and in use are better than 350,000 elements somewhere in the
> official pipeline.
Well, the figure of 350,000 is from SNOMED, and UMLS would have similar numbers,
as would other termsets. All of these exist today.
- thomas beale
