A formal framework is proposed for defining data fusion processes and particularly a notion of qualified gain in a data fusion process is proposed: gain in representation, completeness, accuracy and certainty. These notions are applied to a medical monitoring and diagnosis problem where a dynamic Bayesian network (DBN) is used to modelize time series of observations and evolving states. The model aims at giving a daily diagnosis. Our experiments are under way by using data of an already existing system collected on kidney disease patients. Results will be characterized using our notion of qualified gains.