Screening, Sensitivity, Specificity, and So Forth: A Second, Somewhat Skeptical, Sequel

Abstract

This article is concerned with the sensitivity, specificity, predictive values, and other metrics associated with screening tests. It has direct origins in two previous articles. In this third article, the author of the first article writes about topics and issues that were addressed only minimally in his previous article and expands on topics raised by authors of the second article. In particular, attention is turned to wording and terminology that can be idiosyncratic and confusing with regard to screening versus diagnosis as well as to issues associated with reference (“gold”) standards and screening tests, and to the importance of cutpoints and prevalence in relation to metrics associated with screening tests. The primary aims are to help readers attain clarity about topics that they might have felt unsure about; gain reassurance about conceptual difficulties in the field that, once recognized for what they are, can become less problematic because it is possible to be confident about not being confident; and, where appropriate, adopt an appropriately skeptical attitude about screening tests and their associated metrics. Examples are drawn from the use of ankle–brachial and toe–brachial indices for identifying peripheral artery disease, although wider applicability is intended.