High Fidelity Modeling and Two-Time-Scale Analysis for Hospital Inpatient Flow Management

Cornell University, Ithaca NY

Investigators

Abstract

The objective of this award is to develop an analytical framework, known as two-time-scale analysis, to predict time-dependent performance measures for a class of unconventional stochastic networks. These networks model the management of hospital inpatient flow with patients modeled as arrivals and inpatient beds as servers. This new analytical framework will overcome many challenges that cannot be solved by existing methods for large-scale queueing systems. They include (a) the arrival process has a periodic arrival rate, (b) the service times are endogenous, depending on length of stay (LOS), admission and discharge times, and (c) service times are extremely long compared with the time-variations of the arrival rate. Using the framework, the PI and his collaborators will develop methods to predict the time-dependent mean queue length and mean waiting time in steady state for both single- and multi-cluster models. The models under study handle several realistic and critical features including day-of-week phenomena, non-geometric LOS distributions, and allocation delays. The multi-cluster model has a novel partial sharing mechanism, where the sharing among different clusters is triggered by patients' long waiting times. If successful, this research will produce an analytical and numerical tool that can evaluate various operational and strategic policies including discharges, nurse staffing, and capacity planning. The tool can provide insights into (i) efficient inpatient flow management to alleviate emergency department overcrowding during certain hours of a day, (ii) resource allocation among different wards to optimize multiple patient-centric performance measures, and (iii) coordination within the entire hospital and with other related healthcare systems such as step-down care facilities.