Sensitivity and Specificity

Diagnostic tests are performed by veterinarians on a daily basis. These tests may range from physical examinations (i.e. listening to heart and lungs, measuring body temperature, etc.) to measuring antibody levels in blood by enzyme-linked immunosorbent assay (ELISA). Diagnostic tests are any process or procedure that informs the medical decision-making process. Although useful in many situations, every diagnostic test has limitations; understanding those limitations can assist the practitioner in using diagnostic testing protocols appropriately.

Diagnostic tests are evaluated by their ability to correctly classify individuals as having the disease or condition (i.e. sensitivity) or not having the disease or condition (i.e. specificity). Sensitivity is the ability of the test to produce a positive result when the individual being tested actually has the disease or condition of interest. Conversely, specificity is the ability of the test to produce a negative test result when the individual being tested actually does not have the disease or condition of interest. Situations determine whether sensitivity or specificity is more important to the practitioner. For example, it is important that the diagnostic test used in herd-level testing for bovine viral diarrhea virus (BVDV) be specific, since a false positive could lead to consequences such as unnecessary expenses related to testing and culling. However, when BVDV is correctly diagnosed in a herd, a highly sensitive test is desirable when testing for persistently infected (PI) calves, because missing one PI calf (i.e. false negative) allows the virus to persist in the herd.

Considering which is more detrimental, a false positive or a false negative, can help the practitioner determine a diagnostic testing procedure that maximizes and minimizes the risk of each error, respectively. For example, in cases of highly transmissible disease, or diseases that are treatable early in stages of development but become less treatable as the disease progresses, a false negative can be very detrimental. Thus, highly sensitive tests that reduce the risk of a false negative would be warranted. When a positive diagnosis leads to significant economic or regulatory consequences, or when treatment for the condition is not benign or without risk, a false positive diagnosis would need to be avoided. In these cases, a highly specific test would be advantageous.

Sensitivity and specificity are only characteristics of the diagnostic test, however. Neither describes the pre-test probability of disease, nor gives the practitioner any idea of the probability that an individual who tests positive is actually infected or diseased. These limitations point to the need to understand predictive values, and their utility in interpreting diagnostic test results.