| Hospitals are an integral part of a society's readiness to respond to man-made and natural disasters. Capacity planning greatly enhances the capability and effectiveness of treatment provided to the injured resulting from a disaster. The real-time hospital capacity estimates can be used for patient/ambulance routing, resource planning and emergency operations management. Clearly, case-specific models based on steady-state conditions are insufficient in such a dynamic environment. Hence, a methodology to handle the generic, real-time, and dynamic phenomena has been developed to provide accurate capacity estimations.; This research has addressed three major requirements that are not mentioned in previous research. First, the methodology needs to be generic so that it can represent a large range of hospitals with various sizes and capabilities. Second, in addition to long term performance, the dynamic nature of both hospital operations and patient arrivals in a disaster needs to be captured. Third, the capacity estimation has to be made in real time to ensure its usefulness for disaster relief efforts.; Several steps are taken to meet the above challenges. First, a generic simulation model is developed to take into account the hospital resources, operational efficiency, and types of injuries. Next, factorial simulation experiments are designed to cover a large range of hospitals. To ensure real-time applications, the simulations are executed off-line and the steady-state performances are regressed into a parametric response surface model by using both linear and non-linear regression.; Based on steady-state regression models, a double exponential parametric metamodel is developed to capture hospitals' dynamic performance during the transient period. This metamodel is further improved to be able to utilize the continuous patient arrival rate function. As a reinforcement of the metamodel, a sequential methodology is developed to estimate the dynamic patient arrival rate in real time.; Finally, the capacity estimation methodology is illustrated in an earthquake setting. Results show the viability of the approach and demonstrate promising potential for further analysis of hospitals' dynamic behavior under other emergency situations. More importantly, the developed methodology can be easily applied to other industries such as manufacturing and service. |
