TABLE 2-3. Positive Predictive Value of a Positive Tuberculin Skin Test Assuming 95 Percent Sensitivity

 

Predictive Value (%) at the Following Specificity

Prevalence of Infection (%)

95%

99%

99.5%

90.0

99.4

99.9

99.9

50.0

95.0

99.0

99.5

25.0

86.4

96.9

98.4

10.0

67.9

91.3

95.5

5.0

50.0

83.3

90.9

2.0

27.9

66.0

79.5

1.0

16.1

49.0

65.7

0.1

1.9

8.7

16.0

0.05

0.9

4.5

8.7

0.01

0.2

0.9

1.9

SOURCE: ATS/CDC (2000a), with recalculation to correct predictive values for 0.01 percent prevalence.

test has been estimated at 99 percent or better in areas where exposure to other mycobacteria is uncommon and at 95 percent in areas where such exposure is relatively common (ATS/CDC, 1999a, Appendix B).

A third measure helpful in assessing the usefulness of a screening test is its positive predictive value, which is defined as the probability of a disease or condition in a tested person given a positive test result. A test’s positive predictive value is affected by the prevalence of the disease or condition in the community of those being tested.10Table 2-3 illustrates how prevalence affects calculations of positive predictive value for the tuberculin skin test.

Another way of understanding the effect of disease or condition prevalence relies on Bayesian analysis, as shown in Table 2-4. Given the sensitivity and specificity levels assumed in the table, the positive predictive value—the probability of infection given a positive test result—drops from 49 to 8.7 percent when the prevalence of infection in the community drops from 1 to 0.1 percent.

Thus, even for a reasonably sensitive and specific test, the lower the prevalence of a disease or condition, the higher the proportion of false-positive results. In very low-prevalence areas, a majority of positive test results will be false positives.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement