aam@pucc-h (Dwight McKay) (10/11/84)
(* bug-buster *)
Does any one have a reference to a study (or studies) outlining how
medical professionals go about diagnosing an illness in terms of strategies
they apply to the problem? I'm looking for something along the lines of the
book "How to Solve Problems" by Winograd (?), where different methods of
problem solving were presented as way of teaching the reader to develop a
greater skill in solving problems.
My wife who is attending Veterinary School here at Purdue is
interested in such a reference to assist her in developing a strong
diagnostic skill. Any of you expert system designers know of such a study?
--
-- Dwight McKay, PUCC user services
-- {decvax|harpo|ihnp4|inuxc|seismo|ucbvax}!pur-ee!Pucc-H:aam
-- "I want my UTS!" UTS - UNIX for people who hate to wait...robison@eosp1.UUCP (Tobias D. Robison) (10/17/84)
I cannot suggest references, as requested, but would like to
comment on the interesting fact that most doctors use strategies
of deduction that are not followed in most sciences, businesses,
or professions. [I'm not a doctor, just a consumer of medical
services. If anyone feels I have misrepresneted an honorable
profession, please feel free to comment...]
Most doctors tend to diagnose as follows:
They will consider some, but usually not all of the symptoms and
other diagnostic data that are available to them. They will then
deduce the most common diagnosis that fits most of the information
considered. ("Common" is used in the sense of "occurring most often".)
If this diagnosis fails as a working hypothesis, the process will be
repeated. Usually at this point more diagnostic data is available,
or is collected. Once again, the most common diagnosis to fit the
data considered is obtained (usually excluding the diagnosis
that was rejected).
There is an important exception to this procedure, in that priority
will be given to testing for some well-known possibilities that
require speedy treatment.
Interesting points about this procedure:
- trying to decide what is relevant diagnostic data is itself an
iterative procedure. I think there is no recognized minimum set of
indications that EVERY doctor will use; I've seen diagnoses made on
the basis of hardly any data at all. A practical consideration here
is cost. Any reasponable comprehenive set of data would be too
expensive to diagnose a common cold.
- diagnoses need not fit all of the available data. Many symptoms
reported by people are subjective, and many medical conditions can
be present with a great variety of symptoms. Most people who do
professional deducing would be more likely than doctors to rule a
possibility out or in on the basis of a clear, incompatible symptom.
- This reasoning procedure has a high success rate because:
+ it detects the most common disorders routinely.
+ most medical conditions can still be treated even
if they are only detected after several mis-diagnoses.
This reasoning process is quite unsettling to anyone who is used to
doing decuctive reasoning in his or her work, and who also has an
unusual medical problem. Such people will wait impatiently for their
doctors to collect a comprehensive set of data and then select the
matching condition, while the doctors are actually testing a
sequence of progressively less likely hypotheses.
- Toby Robison (not Robinson!)
allegra!eosp1!robison
or: decvax!ittvax!eosp1!robison
or (emergency): princeton!eosp1!robisonherbie@watdcsu.UUCP (Herb Chong, Computing Services) (10/17/84)
I may be remembering incorrectly, but there was a film called
"The Chips are Down" made about 10 years ago (maybe less) that
spotlighted a package then under development to do medical diagnosis
using a primitive knowledge-based system and a whole lot of rules
of thumb heuristics. It has been long enough ago since I have seen it
that I do not remember who makes it, where the work was being done,
or who was doing the work. The film was a documentary on the micro-
electronic revolution and some of the AI work being done at the time.
Things shown were (I believe) the CMU robot, SHRDLU(?) and this medical
diagnosis system. It may not be what you're looking for, but it may give
a direction to proceed and/or meet your needs.
Herb Chong, BASc
Computer Consultant
Department of Computing Services
University of Waterloo
I'm user-friendly -- I don't byte, I nibble....
UUCP: {decvax|utzoo|ihnp4|allegra|clyde}!watmath!watdcsu!herbie
CSNET: herbie%watdcsu@waterloo.csnet
ARPA: herbie%watdcsu%waterloo.csnet@csnet-relay.arpa
BITNET: herbie at watdcs,herbie at watdcsusmith@umn-cs.UUCP (Richard Smith) (10/20/84)
[]
In response to aam and his wife in vet school:
Dr. Lawrence Weed wrote a book called something like "Problem Oriented
Medical Diagnosis" that my wife vaguely remembers from medical school.
She seems to remember positive things about the book, but we don't seem
to have a copy of it around the house anywhere.
Last summer we encountered a fellow in West Virginia who is using a
software package on an IBM-PC called the "Problem Knowledge Coupler"
(TM of 'PKC Corp') which claims to be an embodiment of Weed's work.
This system claims to be somewhat related to the PROMIS system developed
at some rural New England state university (U Vermont? UNH?).
QUESTION TO THE REST OF US: Is anyone familiar with this
work? My vague impression is that the 'automated' part
may be little more than inspired use of a database of clinical
articles.
If anyone knows, I'd love to find out.
Rick.smith@umn-cs.UUCP (Richard Smith) (10/20/84)
In response to robison and the fine art of diagnosis: My secondhand impression of medical diagnosis matches yours. My wife is a family physician and from what I've seen it's obvious that physical examination is an arcane art. My wife seems to derive lots of data simply from poking, prodding, and listening with a stethoscope. The larger part of one's medical education is in a sort of hospital 'apprenticeship' learning these things firsthand (i.e. med school clinical rotations, then internship and residency). Naturally you're most familiar with things that your teachers know and with the typical ailments at the hospitals you work at. Rick.
jwb@ecsvax.UUCP (10/21/84)
Dr. Weed was Chairman of the Dept of Medicine and the University of Vermont
for a while. He and Dr. J Willis Hurst of Emory University Medical School
were two of the major proponents of so called "problem oriented medical
records" probably better called "problem structured medical records".
These days, most medical records contain elements of this. Some feel that
Dr. Weed and others were somewhat overzealous, concentrating too much on
the form of the record and not enough on the content. I once went to a
Medical School Grand Rounds (case presentation) where Dr. Hurst was the
invited guest and discussant. He spent a lot of time on the relative size
of the writing of the various problems and sub-problems (I am talking about
the size of the letters used to write them) and not much time about what was
wrong with the patient. I am surprised that someone from CMU has not
commented on their project to study the thinking processes of Dr. Jack
Meyers of the U of Pittsburg with an eye toward including these processes
in an expert system for diagnosis. Is this project still ongoing?
Jack Buchanan (MD)
Medicine and Biomedical Engineering
Univ of North Carolina at Chapel Hill
decvax!mcnc!ecsvax!jwb {usenet}
jwb.mcnc@CSNET-RELAY {or whatever the correct ARPA or
CSNET syntax is}lepreau@utah-cs.UUCP (Jay Lepreau) (11/05/84)
Toby Robison writes:
...while the doctors are actually testing a
sequence of progressively less likely hypotheses.
I'd claim that's exactly the process that we use when debugging.
Only when initial hypotheses fail do we bring in the successively
heavier data-gathering artillery such as tracing.
This reasoning process is quite unsettling to anyone who
is used to doing decuctive reasoning in his or her work...
Although scientists might find this so, it shouldn't bother computer
"scientists" at all.jlg@lanl.ARPA (11/07/84)
> Toby Robison writes: > ...while the doctors are actually testing a > sequence of progressively less likely hypotheses. > > I'd claim that's exactly the process that we use when debugging. > Only when initial hypotheses fail do we bring in the successively > heavier data-gathering artillery such as tracing. > > This reasoning process is quite unsettling to anyone who > is used to doing decuctive reasoning in his or her work... > > Although scientists might find this so, it shouldn't bother computer > "scientists" at all. I don't think it would alarm anyone who does deductive reasoning a lot. The method described IS deductive reasoning. As Sherlock Holmes once observed: 'when all that is impossible has been removed, whatever remains, no matter how improbable, must be the truth.' This doesn't prevent checking out the most probable (or the most easily tested) first. It is, in fact, the most efficient way to approach the problem. What might be unsettling to some would be the failure of most doctors to explicitly define the possibilities they are testing. Most doctors follow a systematic proceedure for examination and evaluation of patients in which the various possibilities are examined 'automatically'. There is no need to explicitly elaborate a hypothesis until the examination proceedure nears its end, when only certain possibilities remain.
robison@eosp1.UUCP (Tobias D. Robison) (11/09/84)
This is a followup on the discussion of how doctors reason when doing diagnoses: >I don't think it would alarm anyone who does deductive reasoning a lot. >The method described IS deductive reasoning. As Sherlock Holmes once >observed: 'when all that is impossible has been removed, whatever remains, >no matter how improbable, must be the truth.' This doesn't prevent checking >out the most probable (or the most easily tested) first. Sherlock Homes did not, in my opinion, describe what doctors do. In the first place, many tests are available to doctors, some simple and inexpensive, to rule out the improbable. Usually these tests are not performed until the more likely cases are checked out. A good example is a diseased gall bladder. Its common symptoms are similar (depending upon how people report them) to lower backpain, ulcers, and other forms of gastric distress, including viruses. Doctors almost always will do the more painful, and more expensive ulcer test first (barium X-ray), before checking for gall bladder disease, which is less common. Sherlock Holmes always reasoned on the basis of very little information, but he was careful to collect all he could at a given moment, and then was ready to deduce from that the ONLY possibility, however improbable. Doctors will collect some of the information easily available to them, and then deduce the most probable cause, no matter how many possible causes are still not ruled out. Please recall that I'm not flaming about all this. Anyone who has suffered from one of the less likely possibilities will prefer that more deductive reasoning were used sooner; but I can appreciate that doctors have a system that works a high percentage of time, and also minimizes the number of tests required, at the cost of delaying correct treatment to arelatively few cases. I'm not sure that any alternative would be better. - Toby Robison (not Robinson!) allegra!eosp1!robison or: decvax!ittvax!eosp1!robison or (emergency): princeton!eosp1!robison
riks@athena.UUCP (Rik Smoody) (11/12/84)
Physicians order some tests quite liberally even when it will not help to refine the major hypothesis. These are (usually, sigh) relatively inexpensive tests which have a chance of discovering unsuspected diseases. The idea is that most people do not get regular checkups, so give the whole body the once over while you got it in the office. Rik Smoody