The Impact Monitoring (IM) of Near Earth Objects (NEOs) is a fundamental part of the planetary defense strategy. Current NEO IM systems (Aegis, NEODyS and Sentry) scan the Confidence Region (CR) of each observed object looking for Virtual Impactors (VIs) with a time horizon of about 100 years. This procedure is performed regardless of the uncertainty with which the orbit of the object is known, and without considering whether a scattering encounter is present in the propagation time span. In view of the likely future increase of the IM workload due to higher future NEO discovery rates, it might be more reasonable to adapt the predictability horizon of the impacts to each object, taking into account the orbit uncertainty and the close encounters experienced. In this paper we discuss the problem of estimating a reasonable predictability horizon when multiple close encounters are present and start to address the problem proposing a formal mathematical definition of scattering encounter.
