Decision analysis: dealing with uncertainty in diagnostic testing.

Decision analysis is a process for systematically analyzing complex choices by considering all pertinent information. In this paper, we discuss how uncertainty associated with diagnostic testing can be included in a decision analysis using pay-off tables and decision trees (decision-flow diagrams). Variables associated with diagnostic test interpretation (such as pre-test and post-test probability of disease; test sensitivity, specificity and predictive values; fixed cut-offs versus continuous measurement scales; test dependence associated with the use of multiple tests) are considered. Several decision criteria and output measures are discussed (including MAXIMIN and MAXIMAX criteria, opportunity costs, expected monetary values, expected utility, sensitivity and risk-profile analysis, and threshold analysis). The application of decision analysis to diagnostic testing for Johne's disease and traumatic reticuloperitonitis of cattle, and for canine heartworm disease are used to illustrate both population- and patient-oriented applications and criteria for ranking the desirability of different outcomes.