Consider the problem of estimating the confidence interval of defect rate of a specified population through a screening test. Relationship between defect rate of the population and sensitivity and specificity of the screening test is discussed. More general statistical problem is introduced with the assumption of unknown imperfect specificity. An improved version of Rogan & Gladden(1978)`s estimator is proposed and discussed. Simulation studies are performed to compare the efficiency of the proposed estimator with current estimators.