Using evidence to determine diagnostic test efficacy

Various measures are used to summarize a diagnostic test's efficacy in deciding on the disease status of patients.

Sensitivity and specificity indicate the probability of a positive or negative test result, given that the patient either has or does not have, respectively, a specified disease, and are the most commonly reported measures of test efficacy. Predictive values indicate the probability of a disease, given the patient's diagnostic test result is positive or negative, and are more useful in clinical decision making. Predictive values are useful only if the prevalence of disease in the study sample is representative of the prevalence in the population of diagnostic interest. A likelihood ratio is the sensitivity of a test divided by one minus the specificity and is used as a multiplier of the pretest odds of a disease to obtain the posttest odds. The resulting posttest odds can be converted to a posttest probability of disease that is identical to the predictive value of the test.

An estimate of the posttest probability of disease that is more relevant for an individual patient can be obtained by adjusting the pretest estimate, taking into account patient characteristics and clinical experience. Using the Herren and colleagues' work, diagnostic test efficacy measures are presented to assist practitioners in determining which tests may be accurate in "ruling in" or "ruling out" disease.