A closer look at diagnosis in clinical dental practice: part 2. Using predictive values and receiver operating...

When a clinician is planning to use a diagnostic test or procedure, it is important to establish the likelihood that an individual patient is affected by the condition or disease; this determination depends on predictions that are affected by various features of the diagnostic procedure. In this regard, sensitivity and specificity are limited because they describe the results of a procedure in a dichotomous way: the result is either positive or negative. However, many clinical procedures are not dichotomous, such as probing of periodontal pockets or assessment of radiographs for caries, and in these situations, a range of features is examined to produce a degree of certainty regarding the presence or absence of disease. This article examines predictive values and receiver operating characteristic (ROC) analysis, an algorithm that combines various statistical features of diagnostic procedures to assess the effectiveness of nondichotomous procedures without imposing an arbitrary threshold.