The multicriterion nature of diagnosis

Developing diagnostic strategies is a difficult task, complicated by the multicriterion nature of diagnosis. Sequencing and evaluating tests, as well as performing subsequent diagnosis, are difficult in and of themselves, and achieving accurate diagnosis becomes an ambitious goal. Several new diagnostic tools can compute accurate decision trees using many types of assumptions, thus increasing analytical power. This increased power allows the consideration of the optimization of diagnostic strategies by following several different approaches. They also allow the comparison of the various tools available. This paper provides a framework for developing diagnostic strategies or decision trees according to multiple criteria. These criteria and decision trees fall into three broad categories - preprocessing criteria, optimization criteria, and postprocessing criteria. The authors also discuss guidelines for comparing diagnostic strategies for differing criteria. Interactions between many of these criteria make comparisons difficult or invalid; therefore, the authors provide rules for equivalence as prerequisites for making valid comparisons