We discuss the problem of estimating the number of cases not identified in studies of two or more screening tests in which not all individuals are fully evaluated. In particular we consider studies in which only those individuals who are positive on a screening test have a complete diagnosis. We show how to adjust for individuals who are not fully evaluated despite recommendation for referral and for cases not identified among those negative on all screening tests. The assumptions under which these adjustments are valid are discussed. A proposal is made for analysis of studies using three screening tests. Estimation of and inference about the disease prevalence and the test sensitivity when disease status is ascertained on a random sample of those negative on both screening tests is discussed. Copyright © 2001 John Wiley & Sons, Ltd.