To the Editor.—
The article concerning postmortem evaluation of clinical diagnostics during five decades1 is a good effort to analyze disparate data that can be resistant to marshaling by ordinary statistical methods. However, I think that the true-negative category, within the constraints of this study, is largely a passive statistic that sheds little light on the diagnostic process. This occurs because the act of making a clinical diagnosis leads to a mutually exclusive binomial distribution of that diagnosis vs a statistical universe of other diagnoses, many of which are trivial. On the other hand, "agreed" (true positives), "underdiagnosed" (false negatives), and "overdiagnosed" (false positives) are intuitively satisfying categories that present a more realistic picture of actual practice.Another statistical method avoids the true-negative dilemma by using only those cells of the fourfold (2 × 2) tables that can be occupied by at least one positive diagnosis (clinical or autopsy).
Bates HR. Sensitivity, Specificity, and Predictive Values in the 'Sensitivity and Specificity of Clinical Diagnostics'. JAMA. 1989;262(3):350–351. doi:10.1001/jama.1989.03430030038020
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: