The patient admission scheduling (PAS) problem consists of assigning patients to beds over a given planning horizon to maximize treatment efficiency, patient satisfaction, and hospital utilization while meeting all necessary medical constraints and considering patient preferences as much as possible. Due to the curse of dimensionality, solving large-scale integer programming (IP) models of the PAS problem to optimal is challenging. In this paper, we focus on how to reduce the size of the IP formulation of the PAS problem to improve the solving efficiency. We employ two-stage optimization method where each stage builds a reduced model by constraints aggregation to improve the typical formulation of modeling the PAS problem. Experimental results on widely used benchmark instances indicate that our method can obtain newly improved solutions for 6 out of 13 instances. For the other 5 instances, we obtain matched optimal solutions while attaining this in shorter computational time.