Abstract: Pregnant women need to go to one hospital and apply for admission afterthe 9th pregnant week. The scheduler should make the instant decision of admission or not. Once the women are accepted, the obstetrics unit will provide long-term consistent cares for women from the antenatal care to perinatal stage. It is a challeng for the scheduler to make such a decision due to the features of multiple resources requirement, multi-reentrance, and uncertainties in both demand arrival time and service time of the critical resources. Improper decision will lead to the unbalanced distribution of resource utilization rate of obstetrics and gynecology, and eventhe diversion of pregnant women to other hospitals. To solve this problem, this paper develops the dynamic programming (DP) model to control the resource utilization distribution. However, it is difficult to establish structural properties of the optimal control policy due to the complex structure of the model. Therefore, this paper proposes a discrete-event simulation model and multiple threshold-based admission control policies. The parameters of the policies are improved with the objective of minimizing the weighted cost of beds’ idle and overloaded and patients’ diversion penalty by using the extensive search method.Thenumerical experiments show that the optimal threshold performs much better than that from practices. A sensitivity ana lysis is performed and finds that the match between the capacity and the demand can further reduce the criteria value.

Key words: simulation, admission control, threshold-based rules, obstetrics units

摘要： 孕妇在怀孕九周后会选择一家医院申请建立孕妇档案卡(简称建大卡),管理者需决定是否接收。一旦该医院做出接收决策,孕妇将在该院接受妊娠期内的产检并在该医院分娩。孕妇在妊娠期内所需的医疗资源的多样化、尤其是对关键资源即病床需求时间及服务时间不确定性,增加了决策的难度,从而导致妇产科资源利用率分布不均衡,影响医疗服务质量。为了解决该问题,本文首先建立动态规划模型来最小化病床的空闲、加床和转院的惩罚成本。由于动态规划模型非常复杂,难以用解析方法来获取最优控制策略,因此本文建立离散事件仿真系统,以病床的空闲、加床和转院的惩罚成本最小化为目标,采用广域搜索对多种阈值控制策略进行参数优化。数值实验表明,本文提出的阈值策略比医院目前采用的阈值策略有显著改进,进一步进行了敏感度分析,发现随着到达率逐渐增大,期望成本首先逐渐降低然后在某一水平上波动,加床成本的上升对于期望成本的影响更大。

关键词: 离散事件仿真, 建卡优化, 阈值策略, 妇产科