An information base must be consistently revised and updated upon the acquisition of new observations that contradict the obsolete ones contained in the information base, given a representation model. In this paper, we present a novel approach for dealing with the information obsolescence problem when a Bayesian network is our representation model. We design a polynomial-time algorithm that detects in real-time, contradictions between observations and then identifies among the observations contained in the information base those that have become obsolete given the new observation. Since we work within an uncertain environment, our algorithm is based on a new approximate concept, \({(1-\epsilon )}\)-Contradiction, which describes the tolerable probability of having a contradiction between given observations. Finally, we demonstrate the effectiveness of the proposed approach in detecting obsolete information on an interesting case study: the Student’s Death Diagnosis. Our experiments show that our approach gives systematically and substantially good results.