On Tue, 28 Nov 2000 13:55:09 Horst Herb wrote:
...
>> This means that with the OIO, there is democracy at the level of
>"information model".
>
>Sorry, this might be a dumb question, but how would you actually data mine
>such a chaotically growing structure salad at a public health level?
Hi Horst,
Great question! This has been the fear that has driven much of the top-down,
centralized efforts to control medical vocabulary. One of the main hypothesis of the
OIO project is that the "natural" course of language evolution will sort out the
sementic chaos in electronic health information as it has in all other human
languages. In the psychiatric research arena, for example, diagnostic categories,
criteria, and assessment methods are changing so rapidly that any information system
must be able to keep up with these modifications on a month-to-monthn or day-to-day
basis.
The idea is to build tools to facilitate the extraction of "meaning" across
divergent terms rather than imposing a fixed set of terms on a pre-specified
"knowledge domain". First of all the medical knowledge domain cannot be
"pre-specified", as you know, that change is the rule rather than the exception. So,
rather than assuming that any group of experts can be so smart that they can define
all the "terms" that are needed for the next 6 months - and build a system that
*requires* the existence of this fixed set of terms, the OIO is built to operate on
the "chaotic" reality of the the (medical) knowledge domain. Another problem with an
imposed set of terms=closed vocabulary is that the definition will inevitably _drift_
over time. This is necessary to allow the system to describe something that was not
in the vocabulary when the vocabulary was created. I am giving light treatment to the
problems facing controlled vocabularies - I am sure others can point to many others.
Unfortunately, because of limitations in previous information technology (hardware
and software), the controlled vocabulary approach was the *only* type of system that
can be implemented. (I am not going to go into that here but if anyone is interested,
I will also be happy to share my views on that.) Consequently, most of the field
assumes that controlled vocabulary is the only way to prevent chaos and
non-interpretable data.
So, how does OIO derive meaning from "democracy"=potential chaos? First of all,
although it does not *impose* structure, it encourages it. This is done in 2 ways: 1)
metadata reuse - this means rather than creating a new structure de novo, any user
that chooses to use an existing structure (=OIO form) rather than creating something
from scratch will save time and be able to use existing "tools"(=OIO Reports) that
have already been created to work on that structure. The reuse can either be complete
(i.e. using the entire form) or partial (i.e. using a modified form). Also, there can
be a mix of existing and new forms used in a particular study or patient. 2)
"Linkers"-metadata that connects between items across forms (and across standard
vocabularies). These "Linkers" are reusable and will be subject to archival and
retrieval through online libraries to facilitate reuse. They are produced as needed
to analyze data and merge data sets. Through incremental accumulation and extensive
reuse, they form the "secondary structure" that holds together the dynamic and
extensible domain space within OIO.
>This is
>ultimately what we want to achieve, isn't it?
Yes! That is why we need a strong mechanism in place to "link" together synonyms and
define transfer functions (interpreters) between terms within and across vocabularies.
This is irregardless of whether we have a closed vocabulary or not. The fact is that
there are and will always be multiple vocabularies even if they are all "standard"
vocabularies. The OIO aims to be a "vocabulary" neutral infrastructure that offers
this linking mechanism.
>Gather information on a broad
>base to facilitate evidence based medicine, isn't it?
Right again. Unfortunately, there are many many different outcome indicators now and
in the future. How is it possible that any vocabulary - however learned and
insightful the authors are - can anticipate all the possible measures and outcomes
that must serve as the "evidence" for evidnece-based medicine? Perhaps such an
approach may work for surgery or medicine to some extent (if all we ever care about is
mortality) - but in psychiatry - where we must measure depression and anxiety severity
in many different ways - it is clear that the supporting information tool must support
the "evidence" - not constrain it.
>I fail to see how
>that can be achieved without centrally defined data structures and term
>glossaries.
I point you to the vast majority of studies that have been published without using a
"centrally defined data structure". I raise this merely to point out how successful
the establishment has been able to equate "useful information system" with "centrally
defined data structures". Just because the information system does not constrain the
information content with a "centrally defined data structure" does not mean that the
investigators and clinicians will not use terms and instruments that are well-defined,
reliable, and valid for their data collection needs. This is reminiscent of
"centrally planned economies" where it is claimed that individuals acting autonomously
will not make the best use of available resources. This cannot be further from the
truth.
However, what is needed is a way to minimize the "information/transaction costs" - so
to speak - of re-using existing structures. This means information on available "data
structures"=OIO forms must be readily available and easily re-usable - so that re-use
and consistency is a viable option to all users of the system. That is exactly what
the OIO Library and OIO's plug-and-play metadata are all about.
Andrew
---
Andrew P. Ho, M.D.
OIO: Open Infrastructure for Outcomes
www.TxOutcome.Org
Assistant Clinical Professor
Department of Psychiatry, Harbor-UCLA Medical Center
University of California, Los Angeles
Join 18 million Eudora users by signing up for a free Eudora Web-Mail account at
http://www.eudoramail.com
