Design, Modeling, Evaluation, and Optimization of Intensive Care Networks in Metropolitan Cities

Description

This project will provide solutions towards increasing the efficiency of health care networks, in particular networks for providing intensive care, as such cases are usually more urgent and more resource-constrained than other hospital units, with an intensive care bed requiring around six times the staffing cost of that of a regular hospital bed [1] and a high likelihood of fatality if prompt care is not received. We aim to minimize patient rejection within an intensive care network given a fixed number of available beds at each intensive care unit (ICU), while maintaining standards of fairness for each patient type. For the first time, a comprehensive study of design, modeling, evaluation, and optimization of multi-clustered ICU networks in metropolitan cities will be conducted. This will aid health care administrators with network planning, resource provisioning, and optimizing patient flow. To achieve our objectives, this project will use ideas from stochastic and queueing theory. To maximize ICU network efficiency, we will use the idea of resource pooling so that some patients may be referred to multiple ICUs in the network, and design new policies accordingly to determine the optimal order in which admission to these ICUs should be attempted. We note that the idea of hospital "clusters" already exists in Hong Kong and elsewhere [2,3] and we will base our network designs on these existing clusters while examining the possibility of multiple tiers of coordination, including within a cluster and between different clusters. This will extend the PI and Co-Is' existing work on single-cluster networks and provide a systematic approach to resource sharing in multi-clustered ICU networks, as opposed to ad-hoc approaches currently seen in many hospital regions. Our analytical approach will also allow administrators to examine "what-if" scenarios with increased ICU demand, for example an outbreak of acute infections, rather than relying on possibly limited or incomplete historical data. This work is motivated by recent research achievements of the PI and practical needs of the Co-Is' professional institutions and will take advantage of the PI and Co-Is' complimentary skills. The PI's role will be to oversee the project and provide the mathematical methods underlying the study, whereas the Co-Is' role will be to provide opinions on modeling assumptions and to review the potential applications of modeled outcomes. We will test our ICU network policies and optimization solutions on an example network using previously collected patient data from two Hong Kong ICU clusters.