In article <000301bdae67$45ebfcc0$590d93c3@-jeff>, Jeff Green
<[log in to unmask]> writes
>
>Even in the 60's Tim de Dombal's program on a BBC was giving more accurate
>diagnoses of appendicitis than the doctors at the LGI.
>30 years on I have little doubt that, after triage, computers could be
>programmed to do a more accurate job of diagnosis and make better treatment
>decisions if that's what anyone actually wanted them to do.
>
A further paper was published in '82. This showed the computer as right
in 91% cases whilst the registrar and houseman were 79% and 71%.
However the only computer specific part of this was the Bayes (b. 1793 -
well before silicon computers) Theorem calculations (i.e 1 + 1 = 10 etc
but very fast). Otherwise it was a collection of human experience
(data), human pecision in defining terms and human vision in getting it
all together. Basically it could all be duplicated by a human given pen,
paper and enough time - the computer is just an electronic abacus and a
reminder of the consistently specific statistical input that was
designed by the human.
Dombal himself said that the computer prediction should only be used as
a 'special investigation' after a human diagnostic prediction.
The treatment decision bit didn't come into it.
I think in many cases the generic 'computer' is still a solution looking
for a problem (not counting Quake 11 that is).
Regards
George Myszka
Peterborough GP
-------------
Z J M
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|