Managing hospital inpatient beds under clustered overflow configuration

Abstract

Inpatient beds represent one of the most critical resources in hospitals, of which the availability has been under increasing pressure in China. Such beds are usually to be either used exclusively (“dedicated”) by each department or shared (“pooled”) among departments. This paper considers a generalized bed configuration known as the clustered overflow configuration. The departments are partitioned and managed as clusters. Each department has its dedicated beds and can also admit patients to overflowed beds shared among the departments in the same cluster. A mixed-integer stochastic programming model is developed to optimize the partition and bed-allocation decisions. The objective is to minimize the weighted total cost of rejecting patients, holding patients waiting, nursing cost, and partitioning cost. A simulation-based metaheuristic approach (SMA) is proposed to solve the problem. A niching genetic algorithm (NGA) framework is first proposed to optimize the partitions, and then each partition is evaluated by optimizing the underlying bed-allocation plan through adaptive hyperbox algorithm-based local search (AHA-LS). In AHA-LS, two speeding mechanisms are proposed to effectively identify promising partitions and discard the worst ones. One is an original sequential simulation budget allocation (SSBA) mechanism to sequentially allocate the simulation iteration to evaluate the bed allocation decisions for a given partition; the other is optimal computing budget allocation (OCBA) to allocate the simulation budget to evaluate the bed-allocation plans. The elite clusters form a cluster pool, based on which a set covering model is proposed and solved to obtain a better solution. Case studies are conducted based on real data collected from a public hospital in Shanghai, China. The numerical results collectively demonstrate the applicability and efficiency of the proposed method. Sensitivity analyses are also performed to gain managerial insights.