Understanding the accuracy of tests with cutting scores: the sensitivity, specificity, and predictive value model

Abstract

While researchers usually are concerned about psychometric properties of psychological tests estimated using large samples, most clinical decision-makers must evaluate the accuracy of test results for individuals. This is particularly true as regards tests that have cutting scores to determine, for example, whether to assign a particular diagnosis or accept an applicant into a training program. This paper reviews a conceptual model that may foster improved understanding of test outcomes for individuals. The terms "sensitivity," "specificity," and "predictive value" are defined, and the relations of positive and negative predictive values to population base rates are emphasized. Examples from the psychological literature are presented to illustrate the utility of these concepts in clinical decision-making with psychological tests. Implications for test users, test developers, and instructors are discussed.