Specificity

Calculating Specificity Scores

Specificity is calculated based on the relationship of the following two types of dichotomous outcomes: (1) the true state of affairs and (2) the outcome of the test or collection of tests. The true state of affairs is known either via experimental assignment of some condition [Page 1572]or through classification based on some gold standard test. The outcome of the test is typically referred to as being positive (indicating the condition is present) or negative (the condition is not present). Based on these two types of dichotomous outcomes, there are four possible outcome variables, which are defined as follows:

Specificity is computed from the pool of individuals who do not have a particular condition or disease. Specificity is the number of true negative cases divided by the number of true negatives plus the number of false positives. This is distinct from sensitivity, which is computed as the number of true positive cases divided by the number of true positives plus false negatives.

The specificity of a test is typically not fixed but might vary depending on the cutoff used to define a negative test outcome. To demonstrate this, consider an example in which cognitive performance was measured from 100 head injury cases, 50 of which are instructed to perform to the best of their true ability level on measures of cognition (symptom validity test) and the other 50 are instructed to perform poorly (malinger). Table 1 shows, from left to right, simulated data showing cognitive performance levels and the corresponding number of patients scoring at that performance level who were instructed to perform their best. If a negative test is defined as being any patient with a performance level of 40 or higher on the symptom validity test, then 35 of the 50 patients instructed to perform their best would have a negative test outcome (sample size from the groups with scores of 40 or 50). Given this cutoff, then, the specificity of the procedure would be calculated as: [35 (true negatives) ÷ [35 (true negatives) + 15 (false positives)]] = 70%. In contrast, if a negative test is defined as a performance level of 30 or higher, then 45 of the 50 patients instructed to perform their best would have a negative test outcome (resulting in a specificity of 90% [45 ÷ [45 + 5]]).

...