Knowing the efficacy of a diagnostic tool, whether it is a test used alone, a sequence of several tests or a group of clinical criteria, is essential to study and choose decision strategies. The validation of diagnostic and screening tests is thus necessary to conceive decision schemes. When a gold standard is available, the characteristics of a test can be estimated directly. However, the true individual disease status of the animals is often unknown, particularly in absence of a gold standard or when the gold standard cannot be used because of economical, practical or ethical constraints. In these cases, specific statistical methods like latent class models implemented through a Bayesian approach must be used. Our work aimed at estimating the uncertainty due to the use of diagnostic tests as decision tools. The first chapter presents the issues and practical details of the struggle against animal disease and the epidemiological tools available to estimate the characteristics of the tests and to compare them. In the three following chapters, these methods are applied to three different contexts in which the conception and the evaluation of decision tools are needed: the screening of porcine brucellosis in breeding hogs, the screening of Brucella ovis infection in exported rams and the screening of bovine tuberculosis in Côte d’Or, Dordogne and Camargue (France). The last chapter consists in a global discussion about how to choose a decision tool.
