Bruce: No doubt, the question of how we determine what constitutes optimal
decision making is central not only to clinical medicine but to human
activity in general. Certainly it is of enormous importance to EBM movement,
which is often promoted as a "tool" for effective decision-making. As we
argued elsewhere (and on this discussion group as well) , EBM therapeutic
(and diagnostic) summary measures cannot alone be used in medical decision
making (MDM), but require integration within a normative framework of
decision analysis (DA). EBM summary measures are necessary but not
sufficient to make optimal decision. One can assume that EBM summary
measures affect either probabilities of a given clinical event or outcome
associated with it. Incidentally, when a decision tree is defined in a such
way, interesting (new) relationships between measures of benefits and harms
with respect to a given decision arise(such as, never administer daignostic
test if the treatment harm is greater than benefit, which in turn is
expressed as relative risk reduction, etc). In this sense, EBM summary
measures should be looked at normatively. Extending this argument, one may
also note that our interpretation of the validity of evidence is also based
on axiomatic postulates about what constitutes sound design of the study
(and not on empirical data). Thus, (in my opinion, at least), EBM should be
understood normatively (in a broader sense than implied in my initial note).

However, if you are trying to improve decision making regarding a specific
clinical intervention, and your model ends up providing specific
recommendations (with respect and over your "usual" intervention), then you
may be using EBM in a prescriptive fashion (just as decision aids are
commonly developed to summarize evidence on benefits and harms regarding,
say, hormone replacement therapy, but they cannot tell us should we
recommend it or not).
Again, Bruce started extremely important discussion in which the views of
the other members of the group would be quite welcome.
 
ben
--------------
-----Original Message-----
From: Guthrie, Dr Bruce
To: [log in to unmask]
Cc: [log in to unmask]
Sent: 9/23/99 5:54 AM
Subject: RE: deciding if decision analysis is good

Dear Ben,

Thank you.  I like the distinctions you make, but I'm not sure where 
decision analysis as described in EBM fits in (= "clinical decision 
analysis").  It strikes me that "clinical decision analysis" is 
explicitly based on a normative model (rational choice theory - 
although I dislike the use of the word rational since it immediately 
implies that any other theory or method is irrational).  

However, as described in the EBM textbooks it is prescriptive - it is 
what we should do to make our decisions better, or at the very least 
what we should aspire to or should use when decisions get difficult.  
In that sense it is a decision aid within your definition.  If you 
disagree with this, then you appear to be saying that we should 
judge that decision analysis is best because it's a neat theory.  
It's not amenable to test or evidence.  This doesn't feel right to 
me.

Deciding if it is any good could take two strands then.  Firstly, do 
I agree with the axioms of the normative theory on which it is 
based?  Secondly, even if I do then how would I assess that clinical 
decision analysis (CDA) helps people to make better choices?  Before 
I could assess evidence for this, I think I need to be able to define 
what a "better choice" is and I'm not sure that I know how to do 
that.

Any ideas?

Bruce

> Please don't confuse decision aids with decision analysis. The former
are
> adjuncts, counseling aids "design to help people make specific and
> deliberative choices" among different treatment options. The latter is
"is
> an explicit, quantitative method of clinical decision making that
involves
> SEPARATION of the probabilities of events from their relative values,
often
> called utilities". 
> 	It is customary to think about three theories of
decision-making: 1)
> normative theory describes how (rational) people SHOULD (or ought to)
make
> decisions, and is based on axiomatic mathematical or statistical
concepts
> (usually, the best course of action is the one that maximizes expected
> utility), 2) descriptive theory recognizes that people often violate
> normative principles of decision making and is concerned with
understanding
> HOW ("is vs. ought to") people actually make decisions, 3)
prescriptive
> theory deals with the question what should we do to IMPROVE our
decision
> making (such as developing decision aids).
> 
> 	Normative models are evaluated by their theoretical adequacy,
that
> is, the degree to which they provide acceptable idealizations or
rational
> choice.
> 	Descriptive models are evaluated by their empirical validity,
that
> is, the extent to which they correspond to observed choices.
> 	Prescriptive models are evaluated by their pragmatic values,
that
> is, by their ability to help people make better choices.
> 
> 	Now, it is not clear at all how to define "gold standard" of
> decision-making. People have argued about that for decades, and I am
not
> sure that there is any good consensus about it.
> 
> 	Would be interesting to hear views of the members of this
discussion
> group.
> 
> 	ben

Bruce Guthrie,
MRC Training Fellow in Health Services Research,
Department of General Practice,
University of Edinburgh,
20 West Richmond Street,
Edinburgh EH8 9DX
Tel 0131 650 9237
e-mail [log in to unmask]


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