In this paper, we describe the methodology we used to model temporal aspects of a reasoning for interpreting clinical data in real-time. We assume that 1) aggregation of similar observed situations and 2) forgetting non relevant information are two essential mechanisms involved in this class of reasoning processes. These two mechanisms require domain and context specific knowledge. There were gradually introduced by refinement of initial domain-independent temporal entities and inferences. We implemented our temporal model in order to integrate it into a working prototype, mainly reactive in its reasoning, for the management of ventilator therapy, and we are in the process of testing it in an Intensive Care Unit (ICU).