lihood of true transmission of M. tuberculosis infection in the interval between skin tests. The influence of serial skin testing on the specificity of the test is unknown. To the extent that false-positive results are random biologic events, the false-positive rate will remain relatively constant regardless of the interval between tests, and the number of people in a given population with false-positive test results will gradually grow over time in inverse proportion to the testing interval. Thus, frequent testing will gradually result in the accumulation of a large proportion of the population with false-positive results regardless of the incidence of true transmission of infection.
If false-positive results are not true random biologic events but, rather, reflect individual immunologic responsiveness to tuberculin, persons predisposed to such false-positive results may be removed from a serially tested population and the specificity of the test may improve with time. Most authors have concluded that serial skin testing programs tend to overestimate the incidence of new tuberculosis infection in the populations being tested,(26,27) and it seems likely that false-positive results are random biologic events which will be magnified as the interval between serial tests is decreased.
In the absence of documented nosocomial epidemics of tuberculosis, almost all estimates of hospital transmission of tuberculosis infection are less than 1 percent yearly. If the previous estimates of specificity are accepted, the predictive value of positive tuberculin test done at yearly intervals in such populations will be less than 50 percent. If the assumption that the false-positive rate is a random biologic event is correct, this positive predictive value would be further reduced by testing at 6- or 3-month intervals.
The major value of a serial skin testing program is to alert the system that an abnormally high rate of tuberculosis transmission is occurring.(28,29,30,31,32)Table B-2 shows the range of actual transmission of infection based on the observed rate of skin test conversion assuming a 99 to 99.5 percent specificity. Skin test conversion rates below 1 percent probably reflect very low actual transmission and the majority of skin test conversions are probably false-positive reactions. Conversion rates above 2 percent probably represent actual transmission of infection.
Almost all information concerning interventions in tuberculosis infection without disease are based on studies in populations with a high prevalence of tuberculosis infection and thus a high positive predictive value of the tuberculin skin test. A good example is preventive therapy. Estimates of efficacy and efficiency of preventive therapy are based on