To the Editor.—
The article by Harris on "The Hazards of Bedside Bayes" (1981;236:2602) presents an important caution. However, it misrepresents the inherent variability in data on diagnostic tests that can be collected in well-designed studies, and may exaggerate the effective variability of data on tests now in the literature.The article fails to make clear that the sensitivity and specificity values associated with a diagnostic test are subject to variation on two independent dimensions: (1) the test's capacity to discriminate a given disease from nondisease, and (2) the decision criterion (confidence level or cutting score) that is adopted for declaring a test result to be positive. Thus, a test with constant discrimination capacity can have measured sensitivity values that vary from 0.0 to 1.0, and specificity values that vary from 1.0 to 0.0, as the decision criterion varies from conservative to liberal. This relationship is reflected in the so-called
Swets JA. Sensitivities and Specificities of Diagnostic Tests. JAMA. 1982;248(5):548–549. doi:10.1001/jama.1982.03330050032020
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: