In building emergency evacuation, the perception of hazard can stress the crowd, arouse their competitive behaviors, and trigger disorder and blocking as they pass through a narrow passage (e.g., narrow exit). This is a serious concern threatening evacuees’ survivability and egress efficiency. How to effectively manage risk of such undesired situations is a critical problem in evacuation planning. Based on advanced simulation, behavioral studies and psychological findings on crowd evacuation, this paper establishes a probabilistic graphical model for egress risk analysis, especially considering egress blocking effect on crowd movement. In this model, a hazard event (e.g., fires) is the cause of crowd escape. The undesired event of disorder and blocking is then characterized as an outcome of excessive stress on evacuees due to surrounding hazards, resulting in a drastic decrease of crowd movement in a probabilistic fashion. Several unknown parameters exist in this model, and they have clear psychological meanings, such as the social bond of evacuees, the stress level of crowd in a hazardous condition, etc. Based on this probabilistic graphical model, statistical inference can be implemented and the unknown parameters in the model can be estimated.