Abstract: The behavior of patients flow in a hospital where the patients sequentially pass through a set of healthcare units is analyzed. Focusing on hospital bed allocation， a nobuffer tandem queuing system with blocking is modeled， with the performance indicators such as patients’ waiting time and bed occupancy rate given. The blocking probability of the system is derived by an approximation method. A mathematical programming model with bed cost constraints is developed to determine how many beds are assigned to each stage/division to minimize the system blocking probability. And a neighborhood search combined with simulated annealing (NSSA) is developed to solve the model. The influences of the bed allocation and the weight of beds on blocking probability are investigated. Numerically experimental results are compared with the results in different literatures， showing the superiority of the proposed method.

Key words: bed allocation, blocking, nobuffer, tandem queuing, capacity allocation, simulated annealing