Reporter Volume 25, No.27 May 5, 1994

By MARK WALLACE
Reporter Staff

"Despite or perhaps because of all sorts of technological
advancements, our cognitive skills remain the preeminent part of our
work as physicians," Jerome P. Kassirer, editor-in-chief of the New
England Journal of Medicine, told the audience gathered to hear him
give the annual Stockton Kimball Memorial Lecture at Spring Clinical
Day, held April 30 at the Buffalo Marriot. "But despite the importance
of the diagnostic process, we spend precious little time trying to
understand it," Kassirer said.
	Kassirer's talk, "Clinical Reasoning: Teaching It and Learning
It" focused on what he called "the mental process that physicians use
when they perform the task of diagnosis."
	Robert Reisman, president of the UB Medical Alumni Association
and a longtime friend and schoolmate of Kassirer, introduced the
speaker and praised him as "everything a physician should beQa
clinician, teacher, and researcher who is dedicated to medicine. He's
an excellent physician, often referred to in those most admirable of
terms as a 'physician's physician,'" Reisman said.
	Kassirer, a 1953 UB alumnus who earned his medical degree from
the UB School of Medicine and Biomedical Sciences in 1957, began by
pointing out that he felt it was a great privilege to be giving the
Stockton Kimball Memorial Lecture partly because Kimball was dean of
UB's medical school when Kassirer was a student there. Kassirer
praised UB and the Buffalo area, saying "I treasured my experience in
medical school and the two years I spent at Buffalo General Hospital."
	Regarding the diagnostic process, Kassirer said, "The more we
study it, the less likely we are to make errors. I want to explain how
the process goes, as well as the faults that affect even the best of
us."
	Kassirer pointed out that while physicians often think of
themselves as they only people who use diagnosis, nothing could be
further from the truth. Diagnosis is a process used by all sorts of
people, from politicians to mechanics, Kassirer said.
	The diagnostic process involves, first of all, inferences that
are made from observable facts about the nature of a system
malfunction, Kassirer said. Such inferences yield "a series of
provisional approximations that are revised continually," he said. But
he added that it is important to pay attention to the distinction
between a perception and an inference, and to note how much inference
goes on in any act of perception. "As an example, if you see a ball
bounce into the street, that perception may cause you to immediately
slow down your car because of your inference that the ball could be
followed by a child," Kassirer said.
	He pointed out that there are two basic types of
problem-solving methods, which he termed "weak" methods and "strong"
methods. Weak methods are general problem-solving methods that do not
depend on a particular domain of expertise, are widely applicable
across different problem areas, and are characteristic of novice
rather than expert behavior, Kassirer said. Strong methods, he said,
are those used by domain experts and involve rapid recognition of
patterns, and efficient formulation of problems in "semantically
meaningful chunks." Such methods go directly towards the solution of
particular problems and are domain dependent and narrowly applicable,
he said.
	Kassirer said there are four basic components of the
diagnostic process: hypothesis generation, context formulation,
hypothesis refinement, and hypothesis verification.
	Generating the initial hypotheses and formulating a context
for those hypotheses provides a model for gathering more information
about the problem, Kassirer said. Often based on minimal cues that
come in clusters, these initial hypotheses "can range from the general
to the specific," Kassirer said. "At this stage, hypotheses are
intermingled in a peculiar sort of way."
	After developing a context for understanding why one has come
to these initial hypotheses, the problem -solver enters the stage of
hypothesis refinement, Kassirer said. At this stage, various
hypotheses are deleted, added, or made more specific. This phase is
dominated by probabilistic reasoning as the problem-solver tries to
determine which of the generating hypotheses are most likely correct.
Kassirer said. By using precompiled rules and information about the
source of various problems, the problem solver tries to achieve a
hypothesis that has near certainty, he said.
	When one hypothesis clearly begins to dominate, the
problem-solver has reached the stage of hypothesis verification,
Kassirer said. The hypothesis has to be tested for coherence and
accuracy, and to see whether it explains both normal and abnormal
findings. The competing hypotheses also have to be proved false at
this stage, Kassirer said.
	Research on the diagnostic process has a number of important
uses in education, Kassirer said. It enables better teaching of that
process, and helps students learn how to recognize clusters of cues at
the beginning of the process. Research on diagnostics also helps
explain the rationale behind the traditional routine "workup" with
which physicians prepare for every diagnosis, Kassirer said.
	But research on diagnostics also helps teach students that in
certain cases it may be appropriate to skip around, to start from
laboratory data, or, in the case of urgent problems, to interrupt
normal processes of information, Kassirer said. Further, students can
be taught to avoid certain heuristics that can lead to error. Teaching
the quantitative approach to combining clinical data, as well as the
use of precompiled condition-action pairs, can also be improved by
research on diagnostics, Kassirer said.
	Kassirer listed a number of basic principles for teaching
clinical problem-solving, which included using case presentations and
giving students data in "chunks" that allows them to participate in
the diagnostic process. Kassirer stressed the importance of using real
cases in order to reflect what he called "the messiness of clinical
medicine." Other basic principles include stressing causal relations
when possible, and identifying for students the reasoning strategies
employed, as well as potential errors in reasoning. It is essential,
Kassirer said, to teach by example. Students learn reasoning by seeing
others reason, he said.
	The clinical reasoning process can help minimize errors at all
points in diagnosis, and can help avoid such consequences of error as
delays in treatment, unnecessary exposure of patients to various
tests, inappropriate treatment, and the anxiety that results from each
of these consequences, Kassirer said.
	"All this will hopefully lead us to a general model of problem
solving that can help us identify good or bad reasoning," Kassirer
said.