Accuracy and predictive values in clinical decision-making
Mark J. Eisenberg, MD, MPHAddress reprint requests to M.J.E., Division of Clinical Epidemiology, Jewish General Hospital, 3755 Côte Ste-Catherine Rd., Montreal, Quebec, H3T 1E2, Canada.
In clinical practice, the accuracy and predictive values of a diagnostic test may differ substantially from values cited in published reports, owing to a lower prevalence of most diseases in clinical populations than in study populations. To correct this problem, published assessments of diagnostic tests should standardize accuracy and predictive values to account for disease prevalence.KEY POINTS
The accuracy of a test varies directly with the prevalence of the disease in question, and the upper and lower bounds of accuracy are determined by the test's sensitivity and specificity. When disease prevalence equals 50%, a test's accuracy is exactly midway between its sensitivity and specificity. If a test's sensitivity and specificity have the same value, its accuracy will also equal this value, regardless of disease prevalence. A test's positive and negative predictive values are also strongly affected by disease prevalence. Positive predictive values are high when disease prevalence is high, and they are low when disease prevalence is low. Negative predictive values have an inverse relationship.