Assessment of Diagnostic Tests and Instruments

Abstract

Assessment of accuracy and reliability of screening tests such as detection of cancer and depression is important in clinical and research studies. Although a diagnostic test may not be accurate, the criteria for the disease of interest are well defined, such as cancer cells for detecting breast cancer using mammography. In many studies, we are also interested in less transparent, multidimensional constructs such as quality of life. The latent and multidimensional nature of the latter construct often requires that we probe into various attributes of the construct such as physical, psychological and social functioning of an individual in the case of quality of life. Not only are the criteria less well defined, indeed in many cases the gold standard simply does not exist, but the multitude of assessments also needs to be synthesized to create a few meaningful scales to facilitate assessment of such latent constructs for clinical and research use. In this chapter, we discuss methods for assessing the accuracy of diagnostic tests and validating measurement scales for latent, multidimensional constructs. In the latter case, we also provide a brief overview of the construction of an instrument for assessing such latent constructs to help readers gain familiarity with this increasingly important tool for modern clinical practice and research. We use real study data to illustrate the methods discussed, along with the software used to facilitate the analysis.