   Last update: 08-Jul-2004 Arch Hellen Med, 19(6), November-December 2002, 688-699 APPLIED MEDICAL RESEARCH Diagnostic methodology. 1: Clinical inference under uncertainty E. ANEVLAVIS “Agia Olga” Konstantopouleio General Hospital, N. Ionia, Athens, Greece

Information (evidence) gathering and evaluation, hypothesis formation (differential diagnosis and diagnosis) and therapeutic decision-making, based on the most probable diagnosis, constitute the process that clinicians follow in everyday practice, which is impregnated with uncertainty. Clinical diagnosis is the method by which the necessary conclusions are achieved, using the hypothetico-deductive syllogism IF A THEN B and the productive rules which, by expressing judgment, they are true or false with a certain degree of belief (DB), ranging from -5 (absolute certainty of falseness) to +5 (absolute certainty of truthfulness). When the DB is +5 the a posteriori probability equals the positive predictive value. When the DB is -5 the a posteriori probability equals the negative predictive error and when the DB is 0 there is no information from the evidence E and thus the a posteriori probability equals the a priori probability of hypothesis Y. The main target of clinicians is to arrive at the diagnosis using as few tests as possible and with minimal inconvenience for the patient. Therefore, they search first for the evidence with the maximum informative (diagnostic) content, which is expressed as the difference of the probability for the hypothesis Y to be true given the presence of evidence E and the probability for the hypothesis Y to be true given the absence of the evidence Ε. Based on that evidence they update the probabilities of the hypotheses which are, at this stage, considered as initial probabilities. Next, the maximum and minimum value that each of them can take is calculated, based on the remaining evidence in favor (DB=+5) or against (DB=-5) the hypothesis and for its two true values (YES, NO). Next, the clinicians search for the hypothesis that has the maximum value, among all the minimum values (maximin) and checks whether there is a hypothesis with a maximum value that exceeds the maximin. If there is not, then this hypothesis is the most probable (diagnosis), given the evidence that was used. If there is such a value, none of the hypotheses can be considered as probable (diagnosis) and the process is repeated, checking the diagnostic content of the remaining evidence, in the same way as described above.

Key words: Clinical inference, Degree of belief, Diagnostic content, Productive rules, Uncertainty.