Bed census prediction combining expert opinion and patient statistics
提出两种概率模型，将医生的预计出院日期与基于数据的住院时长分布结合，生成床位占用分布，帮助医院优化床位管理、人员配置和手术排程。
Predictions of bed census are crucial for hospital capacity management choices, encompassing ward sizing, staffing, patient bed assignments, and surgical scheduling. Presently, these predictions heavily rely on doctors’ estimated Expected Discharge Date (EDD). This paper introduces two probabilistic models that integrate EDD with Length of Stay (LoS) distributions derived from data. By employing the Poisson binomial distribution and probabilistic convolution, we generate full census distributions. Applying our approach to real hospital data demonstrates its ability to provide precise predictions, leading to valuable managerial insights. • Predictions of bed census are vital for hospital sizing, staffing, and scheduling. • Current predictions mainly depend on doctors’ estimated discharge date. • We present two models combining expert optinion and with ength of stay data. • We apply the Poisson Binomial distribution. • Our models offer valuable insights, proven with real hospital data.