Jp. Braun et al., Overestimation of the predictive value of positives by the usual calculations of the specificity of diagnostic tests, VET RES COM, 24(1), 2000, pp. 17-24
The specificity of a diagnostic test for a given disease, i.e. the percenta
ge of true negatives, can be calculated from either the results from a grou
p of healthy animals or possibly from a group which also contains diseased
animals which are free of the particular disease for which the test has bee
n performed. The specificity may be much lower in the latter case and the p
redictive value of positives thus greatly reduced. In the example of creati
ne kinase being used for the diagnosis of muscle diseases in dogs, the spec
ificity at the thresholds of 105 and 150 U/L (upper limits of the 95% and 9
9.7% interquantiles) decreased from 0.98 and 1.0 to 0.66 and 0.78, respecti
vely.