Estimation of disease prevalence from screening program

Abstract

The increased use of screening tests for drug use or antibodies to the AIDS virus has raised questions about the reliability of the results of these procedures. These concerns arise because the prevalence of persons with the disease in the general population is far less than that in a pre-screened group. For instance, if there are persons in a high risk category for the disease in question, there may be high fraction of false positive classifications among the test results. In this dissertation, we consider the two estimates from the screening test. First we consider the problem of estimating about the disease prevalence from screening test. We propose the score method based on likelihood theory for constructing confidence interval of disease prevalence. We also compare this one with other method. As a performance measure, we obtained the coverage probability and error tail probabilities. Second, we consider accuracy of the screening test. The sensitivity and specificity of the screening test has been widely used to evaluate the accuracy of the test. However, test results could be highly wrong even though these rates are quite high. The predictive value of a positive test result(PVP), C=Pr(disease condition positive|positive test), is currently regarded as the measure of accuracy of the screening test. We propose a score method for estimating of PVP, which is motivated by Gart and Nam(1988). We derive the score statistic and examine the empirical coverage probabilities of the proposed method under various situations through simulation studies.