Health care system is one of the most critical units in case of disasters. Floods cause an increase of emergency patient flow that may overwhelm hospital resources. In this paper, we present a simulation model that evaluates health care emergency plan and assesses the resilience of the Ile-de-France region in case of a major flood. We combined in the model the health care process with a Markov chain flood model. The results can be used to elaborate an optimized strategy for evacuation and transfer operations. We provide a case study on three specialties and quantify the impact of several flood scenarios on the health care system.
Disasters, natural or man-made, can result in significant economic loss and human casualties. Disaster management operations consist of the reallocation of resources (e.g. health facilities, transportation) to respond to the disaster’s emergency while covering the daily emergencies. One of the main difficulties in disaster management is the lack of resources (Hoard, Homer, Manley, Furbee, Haque, and Helmkamp 2005), and understanding the key resources and their management is primordial to adjust their utilization during a disaster.
In most of the emergency preparedness plans, health facilities represent a key resource and must accommodate the patients resulting from a disaster (Agca 2013). However, hospitals may be themselves at risk of damage from internal or external sources generated by a disaster. For example, in the Ile-de-France region, flood risk is relatively high due the geographical position. As any disaster, flood causes an increase of emergency patients flow that may overwhelm hospital resources (Takahashi, Ishii, Kawashima, and Nakao 2007). Moreover, some hospitals in the region are at risk of submersion and damages to electricity and water coverage. The patients treated in the impacted hospitals need to be transferred to other hospitals.
Hospital evacuation is more constrained than mass evacuation due to the patients’ health conditions and the necessity to relocate them in appropriate facilities. In the literature, hospital evacuation operations have been approached in different ways: project management, mathematical modeling, simulation models and hybrid models (Taaffe, Johnson, and Steinmann 2006). In disaster management, simulation models address a variety of problems (e.g. prevention, response, transportation) to evaluate several outcomes (e.g. costs, mortality) (Altay and Green 2006). Some simulation models focus directly on the building architecture (e.g. exits and staircases) that are used during evacuation (Hunt 2016).
In (Voyer, Dean, and Pickles 2016), a simulation model is developed to compare the impact of different resources on the evacuation operations. The results indicate that an increase to the transportation resources (number of ambulances or the transit rate) has a smaller benefit to evacuation than a change in the available capacity of the safe hospitals. The study in (Yi, George, Paul, and Lin 2010) focuses on the analysis of the available capacity in safe hospitals in Florida, ans estimates the absorption ability of the region in case of flood. In (Taaffe and Tayfur 2006), simulation is used to evaluate the effectiveness of an evacuation plan for one hospital under various scenarios and resources (e.g. patient types, nurses, number of ambulances). Moreover, emergency patients’ flow also varies during disasters such as flood events. However, only few studies integrate the uncertainty of disasters in the simulation models (Stilianakis, Consoli, et al. 2013). (Bankes 1993) suggests using interdisciplinary simulation models using for example meteorological or geological principles.
In France, the French White Plan (Plan Blanc) is an emergency management plan in case of a sudden increase of activity in a hospital (Chen, Guinet, and Ruiz 2015). If the increase of activity involves several hospitals, an Extended White Plan is triggered to coordinate both impacted hospitals and the hospitals receiving the evacuees. One of the main decision makers in the development and application of the Extended White Plan in a given region is the Regional Health Agency (Agence R´egionale de Sant´e ARS).
Our aim in the project with ARS Ile-de-France is to develop a simulation model to evaluate the performance of the regional hospitals in case of a flood event.
We present in this paper a discrete event simulation model that includes two major parts:
- A health care process on a regional (macroscopic) level.
- A flood model using Markov chain to represent the flood dynamic and thus capture the dynamic variation of patients’ flow.
The paper is organized as follows: first we describe the general approach and the various data used in our model. Then we detail the flood modeling and the health care process. Finally, based on a real data set, we present examples of results of the model quantified by key performance indicators.
General Approach and Input Data
The main objective of the DES model we present in this paper is to evaluate the emergency plan of the regional (Ile-de-France) health care system and assess the region’s resilience in case of flood.
We define the region’s resilience as the ability to treat all scheduled patients and emergency arrivals within the region (i.e. with no transfer to hospitals outside of the region). In other words, the resilience is achieved if the non-flooded hospitals can treat their patients as well as emergency patients and the patients coming from the flooded hospitals. In the context of very limited capacity, such solution may only be achieved by predefined management rules. For example:
- Discharging patients in order to free up as many resources as possible, before and during the flood.
- Preventive evacuation of high risk hospitals based on geographic location and electric fragility.
- Transfer of flooded hospital patients according to pre-established preferences.
Unlike most other disasters (natural and man-made), these rules are feasible in case of flood because of the alert period given by the weather forecast and water level measurements. However, the effectiveness of the flood management rules is highly variable depending on the flood dynamic (water level and speed) as well as the emergency patients flow.
Therefore, to evaluate accurately the preparedness plan and the decisions before and during flood, the proposed approach (presented in Figure 1) combines a patient flow model with a dynamic flood model using Markov chains.
In this approach, two main parts are distinguished: the patient flow model and the flood model. The patient flow model is implemented in order to simulate all care pathways in all hospitals at the macroscopic level over a long horizon (one or two years); the flood model is a dynamic short term event (few days to few weeks). When the flood alert starts, health care processes are adjusted by including the flood management rules until the resorption of the flood. Such processes are considered as degraded care pathway (evacuation of the hospital and/or patients transfers). Input data (green boxes) feed the aforementioned models: hospital data (capacity per specialty, flood risk evaluation...), patient data (referred hospital, length of stay, arrival time...) and geographical data (maps of departments within the region, flooded zones...). Finally, several Key Performance Indicators (KPI) are measured (regional resilience probability, number of transfers...) as an output of the model (orange box).