The correction for bias in prevalence estimation with screening tests.

The concern that the South Oaks Gambling Screen (SOGS) and other screening tests have a relatively high rate of false positive errors which results in overestimation of the true prevalence in general population studies is shown to be unfounded. False positives are seen to be a necessary but not sufficient condition for overestimation. It is demonstrated that the proper research question is whether the sample prevalence estimator is biased, and, if so, in which direction. One solution to the problem of bias is shown to depend on the availability of estimates of the error rates of the test.