The frequent evaluation of the risk of fall for elderly people require reliable and abundant information about the person. Since it is not possible to gather regularly all the interesting pieces of information, observations are made over time and stored, which lead to a problem of information aging. This article deals with outdated information in a base of information about an elderly person. We propose a solution that includes a bayesian network-based model of knowledge and a reasoning module in charge of the detection and the management of contradictions and doubts about information.