These numbers, although derived by somewhat dry, boring mathematics, are quite useful to doctors, veterinarians, and personnel involved in outbreak detection and investigation. The quantity 1-1.053 x 10-6 (essentially 1.0) is the probability that the individual who classifies negative does not have SARS. That is reassuring for outbreak investigators and for the patient. On the other hand, the number 0.00018, the probability that the individual has SARS given that the algorithm classified him as positive, does not help. Although 0.00018 is 18 times greater than the prevalence of disease, meaning that the chance that this individual has SARS is higher than before the classification, it is not high enough to change the management of the patient. The evaluator would conclude that the performance of the case detection algorithm with 90% sensitivity and 95% specificity is not good enough for this purpose.
Was this article helpful?