Since the heat wave of the summer of 2003 in France, regional and national commitments to face the struggle against the peaks of activity and the overcrowding in health care systems have multiplied in order to guarantee a better patient health care. Our objective in this paper is to identify the overcrowding situations through key indicators, model patient pathway using workflow approach and finally offer agent-based decision support solutions for managers to provide better control of activity. In case of uncertainties in hospital emergencies, a coalition of agents is formed to collaborate and negotiate to provide decisions concerning the dynamic orchestration of workflow and minimize the waiting time of patients during their care. This work is a part of the project HOST (Hôpital: Optimisation, Simulation et évitement des Tensions), funded by the French National Research Agency (ANR).