Numerous decision-making tools exist to assist physicians in diagnosis
management. However, the accuracy of available clinical information i
s often ambiguous or unknown and current analytical models do not expl
icitly incorporate judgmentally defined information. A model encompass
ing both physician judgment and probability analysis was developed to
accommodate such data. A problem requiring sequential diagnostic testi
ng was structured utilizing the analytic hierarchy process (AHP). The
case presented involved a patient complaining of upper abdominal pain
who, after initial evaluation, did not need immediate surgery. Physici
ans were faced with identifying the optimal sequence of diagnostic tes
ting. The criteria used for test selection included minimizing risk, p
atient discomfort, and cost of testing and maximizing diagnostic capab
ility. Although at the onset the ''best'' test choice was unknown, the
clinical picture indicated four test alternatives: upper gastrointest
inal series (GI), abdominal ultrasonography (US), abdominal computed t
omography (CT), and upper gastrointestinal endoscopy (END). Based upon
the relative preferences of the criteria utilized, the AHP analysis i
ndicated that upper GI series was the optimal first test. Given a nega
tive test, posterior probabilities were calculated using Bayes' theore
m, resulting in a new estimate of diagnostic capability. The AHP analy
sis was reiterated, identifying abdominal ultrasonography as the optim
al second test. This analysis may be repeated as many times as necessa
ry. Sensitivity analysis demonstrated that changing criteria preferenc
es may alter the choice of tests and/or their sequence.