Evacuation drills are critical to evaluate emergency preparedness and infrastructure capacity. Before conducting drills, it is necessary to design the evacuation routes that people are likely to follow in a real evacuation. In this paper, we present a path-oriented optimization model for evacuation planning. To solve this model, we propose a column-oriented approach in which a master problem assigns people to evacuation paths, while an auxiliary problem generates feasible evacuation paths. To recreate possible evacuation drills, we embed the column generation method in an optimization-based simulation procedure that mimics the evacuation dynamics and unveils critical evacuation routes that are likely to become congested. To illustrate the applicability of our method, we recreated a real evacuation drill conducted in a university campus. Our results support evacuation planners that often make infrastructure decisions with the goal of enhancing evacuation dynamics with the commendable goal of saving lives.