Screening tests are often used to estimate the prevalence of disease at a single point in time (point prevalence) in a herd or a given population of subjects. This can be easily done when the sensitivity and specificity of a diagnostic test are known and these two properties of a diagnostic test remain constant for every population of subjects being tested, irrespective of the prevalence of the disease being tested for. Based on study of nine infected dairy herds with paratuberculosis, where all adult cattle in the herds were tested by multiple diagnostic tests for paratuberculosis, we found that the sensitivity of several tests changed depending on the severity of the disease in the population being tested. As a result, the standard equation for estimation of disease prevalence, first described by Marchevsky and cited in most epidemiology textbooks, produces incorrect results. Specifically, use of the Marchevsky equation resulted in under estimation of paratuberculosis prevalence in herds with few test-positive animals, and over estimation of paratuberculosis prevalence in herds with a high number of test-positive animals. This observation has important implications for disease control and eradication programs when screening tests are used to estimate the prevalence of disease in different populations.