Some time ago, on behalf of the PHCSG Clinical Interests Group, I posted
ideas about archiving data from GP records to GP-UK and asked for
comments. The response was fantastic and from this collaboration a paper
was produced on the subject which was presented at the PHCSG Annual
Conference at Cambridge. The role of the GP_UK contributors was
acknowledged.

I feel that this e-mail discussion process is one of the best ways to
develop our ideas about what we want from computer systems.

Once again, as part of the PHCSG activities, I am presenting a few
thoughts on the subject of what sort of decision support we want. Any
comments, suggestions, references worth including etc. would be very
gratefully received.


Paper follows as plain text..
DECISION SUPPORT - DIFFERENT ROUTINES FOR DIFFERENT DISCIPLINES?

Introduction
Decision support systems have been under development for some time, the
increasing sophistication of the electronic medical records probably
means that they will be implemented in a more widespread fashion over
the next few years.

There are two main mechanisms in use in decision support systems.  Many
of the original systems were based on statistical probability, in
particular using Bayes theorem.  In such systems the statistical
probability of various conditions or diseases is held within the
computer and the decision support system utilise information given by
the user to work out the particular probabilities of any individual
circumstance.

The other method is to use rule based systems working with a knowledge
base containing large amounts of medical data in the form of statements.
The rules, gathered from experts in the field, are used by a mechanism
called an inference engine to explore the knowledge base to come up with
possible solutions to the questions being posed by the user.  Later
developments of decision support systems have tended to be a mixture of
rule based systems with some statistical probability incorporated within
the rules. Both mechanisms require that medical information is contained
within the system.  The purpose of this paper is to explore whether or
not medical information, contained within these systems, needs to be
different for different disciplines.  It will be argued that three
factors vary from discipline to discipline, particularly between
hospital medicine and medicine in primary care, and these differences
mean that a system suitable for one discipline is not suitable for
another.  The three areas are:

*       The differing prior probabilities of the problems of a condition
according to the discipline

*       The differing diagnostic processes undertaken by different
medical disciplines

*       The differing emphasis on disease management and Health
Promotion.

Difference In Prior Probabilities
Hospital doctors do not experience the same probability of the problems
of the disease as general practitioners.  In the hospital setting the
patients are already a selected group and therefore the incidence of any
particular condition is at variance with the incidence experienced by
general practitioners who do not deal with the same selected group.  Any
system which relies on the problems of disease presenting to hospital
clinicians will produce answers which are appropriate to the selected
population in the hospital.  This will be different to that experienced
by the general practitioners.

The Differing Diagnostic Approaches
General Practitioners deal with known patients, often with trivial
complaints, on a frequent and recurring basis.  Hospital doctors deal
with selected patients, more likely to have significant disease, on a
one off basis.  General Practitioners also have to cover the whole range
of physical and psycho-social disease.  Hospital doctors specialise in a
particular discipline.  As a result, GP's are trained to work in a
'hypo-thetico-deductive' manner.  This  allows for quick decisions based
on a number of clues.  These decisions, or hypotheses, are then tested
to prove or disprove their validity.  This proofing process can take
place over repeated visits over a longer period of time.

Hospital doctors work according to pre-defined algorithms - exploring
the problem in a logical and structured manner by eliminating
possibilities until a preferred differential diagnosis is obtained.

Another difference in the diagnostic process is the determination of the
end point.  Hospital doctors will tend towards reaching a definitive
working diagnosis.  GP's are trained not to strive to reach a definite
diagnosis, but merely a management plan which may not include an actual
diagnosis at all.

Thus the structured approach of the hospital doctors diagnostic process
lends itself to the structured lends itself to structured algorithmic
mechanisms of classical decision support.  GP's may benefit more from
the less structured 'watchdog' approach.

Decision support systems for GP's will have to cope with the whole range
of physical and psycho-social conditions.  The emphasis will need to be
on highlighting possibilities rather than confirming diagnosis.

Hospital doctors will require more in depth, but speciality limited
systems which lend towards arriving at a differential diagnosis.

The differing emphasis on Chronic Disease Management and Health
Promotion:
Hospital doctors are only concerned with the speciality in which they
practice.   Thus the neurologist will be interested in chronic disease
management of epilepsy, but would not be concerned about that same
patient's problem of asthma.

GP's have the need to simultaneously monitor all the diseases from which
the patient may suffer.

Any Health Promotion offered by hospital doctors tends to be
unstructured and patchy.  GP's have specific responsibility for taking
an overview of any patient, and providing Health Promotion concerning
all aspects of the patient's lifestyle.



--
Dr. Glyn Hayes


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%