Conference PaperPDF Available

Improving hospital evacuation planning using simulation.

Authors:

Abstract

Hospital evacuation in the event of a hurricane is a complex and unpredictable process. Recent natural disasters have called attention to the importance of a timely evacuation plan. The success of an evacuation greatly depends on developing and evaluating alternative plans. However, there is no standard approach to address the issues of a hospital evacuation. This research describes the development of a simulation model and initial analysis to assess the effectiveness of an evacuation plan given different scenarios and resources.
Proceedings of the 2006 Winter Simulation Conference
L. F. Perrone, F. P. Wieland, J. Liu, B. G. Lawson, D. M. Nicol, and R. M. Fujimoto, eds.
ABSTRACT
Hospital evacuation in the event of a hurricane is a com-
plex and unpredictable process. Recent natural disasters
have called attention to the importance of a timely
evacuation plan. The success of an evacuation greatly
depends on developing and evaluating alternative plans.
However, there is no standard approach to address the is-
sues of a hospital evacuation. This research describes the
development of a simulation model and initial analysis to
assess the effectiveness of an evacuation plan given dif-
ferent scenarios and resources.
1 INTRODUCTION
Hospitals are usually considered a safe haven and support
system for the people involved in an emergency situation.
As the foundation for many emergency response plans,
hospitals are rarely considered subjects for evacuation.
This research focuses on the event of a hurricane in which
a hospital must decide to evacuate and relocate its patients
and staff to nearby shelters.
The objective of our research is to propose simulation
modeling as a tool to understand, analyze, and improve
hospital evacuation plans. This research evaluates the ef-
fects of varying transportation, sheltering, and staffing
plans for hospitals, while observing the effects on evacua-
tion time and number of patients evacuated.
In surveying South Carolina and Florida hospitals,
data were collected on the present state of evacuation
plans. As of 2004, the Department of Health and Envi-
ronmental Control in South Carolina requires all hospitals
to develop and/or update their evacuation plans (DHEC
2004). However, risk managers have limited actual
evacuations to aid them in refining and updating their
hospital’s plans. Some hospitals actually are forced to
perform costly mock evacuations to evaluate many of the
responses our research is modeling. We feel this research
will benefit hospitals because simulation provides a
method of testing many different scenarios over many
replications to observe the outcomes that are critical to a
successful evacuation.
Hospital evacuation planning involves several com-
plex, interrelated steps. The evacuation process can be
thought of as a set of activities, some of which are con-
strained by resources, but all of which must be completed
for success. Taaffe and Tayfur (2006) propose optimiza-
tion-based models that measure the ability to effectively
evacuate patients from a hospital based on evacuation
cost, clearance time and patient risk. Yet, with cost, time
and risk as competing objectives, there is not one clear
recommendation. Moreover, their work assumes all
events and tasks during the evacuation are deterministic in
nature. In addition to the uncertainties surrounding a hur-
ricane event, the available resources for accomplishing an
evacuation plan will be unique at any given time. While a
mathematical programming approach may be applicable
at a tactical level, there is a need for conducting simula-
tion analysis to understand the interdependencies of op-
erational-level decisions (Law and Kelton 2000). In this
paper, we provide insight into these interdependencies
through the use of simulation.
2 RELATED LITERATURE
Researchers have typically focused on general population
evacuations as they pertain to the use of roadway infra-
structure to move people away from a hazard (see, e.g.,
Sheffi et al. 1982, Pidd et al. 1996, Hobeika and Kim
1998, Franzese and Joshi 2002, Chang 2003, and Cova
and Johnson 2003). Some researchers have addressed as-
pects of the evacuation problem such as decision making
procedures (see, e.g., Tufekci 1995, Gladwin 2001, and
Sorensen et al. 2004) and emergency preparedness train-
ing (Pollak et al. 2004), and Frantzich (1997) considers
risk evaluation where hospitals serve as support for first
responders. Vogt (1991), McGlown (1999), and
McGlown (2001) even consider the decision-making
IMPROVING HOSPITAL EVACUATION PLANNING USING SIMULATION
Kevin Taaffe
Matt Johnson
Desiree Steinmann
Department of Industrial Engineering
Clemson University, 110 Freeman Hall
Clemson, SC 2934-0920, U.S.A.
5091-4244-0501-7/06/$20.00 ©2006 IEEE
Taaffe, Johnson, Steinmann
process regarding evacuation of health care facilities and
special needs populations. However, the problem of de-
veloping robust hospital evacuation plans using quantita-
tive techniques is still largely unresearched. Taaffe et al.
(2005) discuss the many issues and complexities inherent
in not only hospital evacuation planning but also plan
execution. Recently, the United States Government Ac-
countability Office released a report that summarized pre-
liminary observations of the issues surrounding health
care facility evacuation due to hurricanes (GAO 2006).
However, the report does not address how to improve,
suggest, or implement more robust evacuation plans.
Consideration of hospital evacuation occurs when a
threat to the population grows to include the hospital it-
self. However, the hospital under consideration is usually
an integral part of a broader emergency response plan to
deal with those injured or exposed to the threat, which
could result in a decision not to evacuate despite the threat
to the hospital. While hurricanes and floods were the
primary concern for mass evacuation planning for many
years, a more concentrated effort is now being given to
hazardous material spills and terrorist incidents (see, e.g.,
Rogers 1994 and Lindell 2004). Broadening the number
of possible threats expands both the scope of the problem
and the number of hospitals which may be at risk.
One of the more difficult problems with transporta-
tion and sheltering plans is when the threat grows to in-
clude the “safe” facilities or hospitals. The U.S. hurricane
season of 2004, for example, severely taxed resources
both by direct impact and by the extent and frequency of
the storms. It was not unusual for evacuees to find them-
selves the target of a storm after evacuation (Hoffman
2005). Given that hurricane-force winds can occur over a
wide swath, there is a strong likelihood that multiple hos-
pitals will undergo evacuation procedures (and each of
these hospitals will be subject to potential critical systems
failures). To date, there has been no documented research
to address the logistics of such an evacuation.
3 MODELING HOSPITAL EVACUATIONS
As previously stated, Taaffe and Tayfur (2006) compare
various resource and vehicle transport allocations for a
hospital evacuation based on a deterministic optimization
model. Due to the level of detail included in the model
representing the evacuation decision-making process, it
may become prohibitively difficult to formulate each de-
cision mathematically. Even assuming that we can ade-
quately represent each task or operation, we must also
consider the variability in the duration of each of these
sequential events that the hospital must complete during
an evacuation. As a next step, this analysis could be ex-
tended to account for stochastic elements directly into the
optimization model. However, this will not be a straight-
forward extension, and additional research will be re-
quired in this area. Instead, we incorporate the uncer-
tainty of task duration and timing through the
development of a simulation model. The next section
presents the assumptions that were used in the analysis.
Based on data inputs and evacuation guidelines from
several coastal hospitals in South Carolina and Florida, a
model has been developed that determines the average
evacuation time and the overall clearance (or evacuation
completion) time, based on the parameters of a proposed
evacuation plan. In this analysis, we only test a subset of
possible representations, and alternative evacuation plan-
ning models are currently being investigated with input
from the operators and risk managers at the hospitals. We
selected our scenarios based on the operating conditions
at Cape Canaveral Hospital (Florida) and Beaufort Me-
morial Hospital (South Carolina), both mid-sized facilities
with 150-200 beds. It is assumed that there is one hospital
for which we are evaluating its evacuation plan options.
This hospital can send its patients, medical staff, and ba-
sic supporting equipment (e.g., IV hookups) to various
sheltering facilities in the region. The logistics of trans-
porting advanced medical equipment and supplies is omit-
ted. Moreover, the sheltering facilities are assumed to be
hospitals, and they will typically have most of the pa-
tients’ medical equipment needs.
In this model, those patients who will be part of the
evacuation plan are addressed. Also, all tasks related to
preparing the patient for evacuation staging have been ag-
gregated into a single stochastic delay. This could include
preparing the patient for moving from the patient room,
processing any/all paperwork regarding the move to a
sheltering facility, moving to a first-floor staging area,
etc. The evacuating hospital faces limited resources in
terms of the number and size of transporting vehicles,
staging area for transport, support staff to accompany
transferred patients, and bed capacities at the sheltering
facilities.
3.1 Model Structure
Using discrete-event simulation modeling, we can meas-
ure the effectiveness of evacuation policies by modeling
human behavior and other stochastic decisions that may
not be handled adequately in the optimization model. The
goal is to design a set of experiments that systematically
test alternate flows, staging, and scheduling of events dur-
ing an evacuation. While there may be a universal set of
experiments of interest regardless of the specific hospital
being evacuated, there will still exist a need to study
unique hospital characteristics to be able to accurately as-
sess a plan‘s performance.
There are several approaches for model development,
and we describe one potential methodology for incorpo-
rating the administrator/staff/patient decisions into the
simulation. In Figure 1, we assume that there are three
510
Taaffe, Johnson, Steinmann
main areas of control within the evacuation process: storm
control, patient / medical staff control, and administrator /
risk manager control.
Figure 1: Simulation Model Structure
Using simulation, not only can we more accurately
account for the stochastic nature of human decisions, but
we can also characterize the uncertainty in the timing, se-
verity and duration of the hurricane event, and provide 3-
or 6-hour storm-track updates. It would likely have an
effect on the speed at which an evacuation can be imple-
mented, the time at which an evacuation is initiated, and
the number of patients that can be safely evacuated. This
functionality would be provided within the storm control
function area.
Within the patient / medical staff function area, we
will have the ability to monitor individual patient rooms
and obtain evacuation status on each room (e.g., room oc-
cupied or not, patient type, evacuation decision (release,
transfer, stay), and expected release or transfer time, if
applicable). If there is additional patient/medical staff in-
teraction that affects release or “ready for release” time,
this information can be included in the model.
The third function area is administrator / risk man-
ager control. This is a very important piece in the model-
ing process in that we can allow for policy changes in the
midst of a hurricane event. This will provide hospital
management with the ability to make decisions that
change the responsibilities of doctors, nurses, and staff in
how they are currently handling patients and the overall
evacuation.
Ultimately, we would like to include appropriate de-
tail in all of the control areas listed in Figure 1. In fact,
most of the body of code has been included in the base
simulation model. However, we have chosen to focus on
the patient / medical staff control function for this first re-
search paper on hospital evacuation.
We have developed the simulation model using the
commercially-available simulation software package
Arena, which is especially suited to representing process
flows. We recognize that hospital evacuations can be in-
fluenced by the interactions of different persons (patients,
nurses/staff, risk managers, and local emergency man-
agement personnel). As an area of future research, we
plan to test the use of agent-based modeling as a means of
more accurately representing these human behavioral de-
cisions (see, e.g., Deadman 1999, Bonabeau 2002a, and
Chen 2003). We may also be able to incorporate optimi-
zation within the simulation models that we develop.
Highly detailed, discrete-event simulation models can run
extremely slow, rendering near-term planning exercises
fruitless. Through our extensive testing, we will incorpo-
rate the results and learning experiences into a more ro-
bust evacuation planning procedure.
3.2 Model Assumptions
We assume that there are three acuity levels. An acuity
level 1 patient is denoted as any patient who is a candi-
date for early release, and these patients can expect to be
released 24 hours earlier than normal to reduce the num-
ber that need to be transported to another facility. How-
ever, not all acuity level 1 patients will be released, based
on any number of reasons where care cannot be provided
away from the hospital. An acuity level 3 patient repre-
sents a critical care patient, such as those either waiting
for or in recovery from a serious operation, or those that
have an extreme ailment. All other patients would fall
into the larger, middle group, which we denote as acuity
level 2. There are no priorities placed on the order in
which patients are evacuated (i.e., all three patient groups
will be evacuated simultaneously based on the availability
of the appropriate transport vehicles). In future research,
we will recognize that certain acuity level 3 patients may
require immediate evacuation to obtain the care that they
need, possibly preempting a planned transfer of acuity
level 1/2 patients.
Once a nurse is assigned to assist in transporting a
group of patients, he or she will remain with the patients
until the end of the evacuation to provide necessary care.
Based on feedback from hospitals participating in the data
collection effort, it is assumed that one nurse is required
for every 5/5/2 patients of acuity level 1/2/3 for transport-
ing patients.
In this model, there are three sheltering facilities,
each of which can accommodate any of the patient types.
We also include an overflow shelter to accommodate ad-
ditional evacuated patients, when originally-anticipated
shelter capacity is not provided for any number of rea-
sons. The evacuating hospital can have up to 50 patients
of each patient type, and the sheltering facilities are also
assumed to have up to 50 beds available. Vans and ambu-
lances are available for transport, and we will vary the
number available across different experiments. Each ve-
hicle can travel at speeds between 30 and 45 miles per
hour, and all facilities are assumed to be 100 miles away
from the evacuating hospital. We do not consider any
costs in this model.
Administrator / Risk
Manager Control
Patient / Medical
Staff Control
Storm Control
511
Taaffe, Johnson, Steinmann
3.3 Model Detail
In this research, we are proposing simulation modeling as
a tool to understand, analyze, and improve hospital
evacuation plans. The model allows for adjusting any of
the assumptions listed in Section 3.2 that provide input
data or define a particular test scenario, including initial
patient count prior to storm creation, nursing staff levels
and allocation, evacuation transportation capabilities, and
number of available shelters. The analysis provided in this
paper will concentrate on evacuation times, both for indi-
vidual patients and the overall evacuation completion
time.
In modeling the hospital evacuation, we created five
submodels that capture data input, patient control, evacua-
tion, storm updates, and risk manager control. Within the
data input submodel, the system creates patients and at-
taches each an array of assignments. Two important pa-
tient attributes are acuity level and expected release time.
Based on either of these attributes, a patient may leave the
hospital before evacuation begins. Also, according to
acuity level, patients require a certain amount of time and
staff members in order to evacuate the hospital.
Currently the number of staff members is pre-defined
within the resource configuration. Each group represents
a floor of the hospital and the corresponding patients by
acuity level. For example, nurses assigned to floor 1 are
responsible for acuity level 1 patients.
Patients continue to the patient control submodel
which monitors room occupancy and patient evacuation
status. After being assigned a bed, the patients wait for
storm updates. If the patient’s expected release time oc-
curs before evacuation begins or within six hours of
evacuation, then that patient prepares for release and exits
the hospital system. Otherwise, patients will take part in
the hospital evacuation. Additional patients can also ar-
rive to the hospital prior to the evacuation, creating the
ability to have an unpredictable number of patients at the
beginning of an evacuation.
Our efforts have been focused on the evacuation
process itself. First, each patient requires a nurse to move
from a hospital room to the staging area, and the time to
relocate the patients is a function of the patient’s acuity
level. Then, patients are divided into groups ready for
transporting to shelters. Each group requires a nurse/staff
member before proceeding to the loading area.
Depending on the patient’s or group’s condition, the
amount of time and resources required to move the patient
to the staging area and then load each patient onto a trans-
port vehicle may range from a few minutes to nearly an
hour.
As patients arrive to the staging area, they are as-
signed to travel to specific sheltering facilities. The cur-
rent model utilizes two types of transport vehicles: vans
for acuity levels 1 and 2 and ambulances for acuity level
3. Both types of vehicle types have the capacity to hold
one batch or group of patients. If capacity is no longer
available at the assigned shelter when the vehicle is ready
to depart, the remaining patients travel to an overflow
shelter with unlimited capacity. Once a van or ambulance
arrives and unloads at a sheltering facility, it returns to the
evacuation facility with similar transport delay times.
Once all patients are evacuated, the evacuation is de-
clared complete, and an overall evacuation completion
time is recorded. While this section does not present an
exhaustive list of issues in this area, these are certainly
among the most important. It is doubtful that any plan-
ning process will truly address all issues. However, the
robustness of the plan will depend on solid coverage of
the most essential issues.
4 EVACUATION MODEL RESULTS
In this section, we present the findings from the base
model described in Section 3. We consider the following
resource assignments in the base model:
10/10/25 nurses for acuity level 1/2/3 patients
3 shelters each with a capacity of 50 patients
1 overflow shelter with unlimited capacity
3 vans and 3 ambulances
For every test conducted, 20 simulation replications were
run, all producing fairly low variation (reported half-
width values in Arena were consistently less than one
hour). This is due in large part to some simplifying as-
sumptions that were made in the base model. Further
model development will remove such assumptions, result-
ing in an even more stochastic environment to consider.
First, we tested several initial patient counts when us-
ing the resources defined for the base model, and the av-
erage evacuation times per patient and average comple-
tion times are reported in Table 4.1.
Table 4.1: Base Model
Patient
Count
Avg.
Evacuation
Time (hrs)
Avg Com-
pletion
Time (hrs)
50/50/50 40.4 65.1
40/40/40 31.1 47.6
30/30/30 24.0 36.5
50/40/30 31.5 46.5
30/40/50 36.2 65.3
The correlation between patient counts and evacua-
tion times is not a major finding. Instead, the purpose of
running each test was to observe the magnitude of the
change in evacuation time across each test. Note that
these patient counts represent hospitals similar in size to
Cape Canaveral Hospital and Beaufort Memorial Hospital
(input data sources from Section 2). Also note that these
patient counts do not include the patients that could be re-
512
Taaffe, Johnson, Steinmann
leased early (and, thus, did not require evacuation). For a
hospital with 90 patients to evacuate, evenly split across
all three acuity levels, the average evacuation time per pa-
tient is 24 hours, with an overall completion time of 36.5
hours. However, when increased to 150 patients
(50/50/50 for acuity levels 1/2/3), the completion time
approached three full days. The travel time assumptions
are still quite liberal, which means that any additional de-
lays on the roadways would only further exacerbate the
delay in finishing the evacuation.
From the input data gathered, the critical task is
transportation and not building evacuation. In other
words, even if the hospital is evacuated more quickly, it
would not change the patient evacuation times or evacua-
tion completion time since transportation is the bottle-
neck. If, however, the transportation element can be re-
duced to requiring only a few hours, then the ability to
efficiently prepare patients for evacuation would become
increasingly important.
Note that the base model assumes only three ambu-
lances and three vans. Based on the results, the bottle-
neck operation is the evacuation of acuity level 3 (critical
care) patients. In order to gauge the effect of adding
transport vehicles, we perform additional tests on each pa-
tient count. The new tests, shown in Tables A-1 – A-5 in
the Appendix, report 12 combinations of initial vehicle
requirements (including the base model). Increasing the
number of ambulances has a greater positive impact than
increasing the number of vans. Also, with fewer acuity 3
patients, the system approaches its optimal (minimum)
evacuation time with a fewer number of vehicles. These
results support the idea that acuity 3 patients create the
bottleneck in the system and have the greatest effect on
the results.
This simulation model has the potential to test other
changes in resources. For instance, vehicle capacities may
be varied to allow more or less patients per vehicle. Dis-
tance and travel times between the hospital and shelters
can also be changed to demonstrate the impact of travel
obstacles and shelter location relative to the hospital. An-
other variable to consider is the number of nurses for each
type of patient. Holding the patient count constant, we
can test the effect of the number of nurses assigned to
each acuity level. In addition, we may vary the distribu-
tion of time required to prepare the patient for evacuation.
The purpose of these additional tests would be to deter-
mine the resources with the greatest impact on the success
of the evacuation, and these resources will be the focus
for improving the system.
5 CONCLUSIONS AND FUTURE RESEARCH
In this paper, we proposed a simulation model as a
method of understanding, analyzing, and improving hos-
pital evacuation plans. The model utilizes resource re-
quirement information to provide evacuation time data
that can aid risk managers in making decisions regarding
their hospital’s plans for evacuation.
The base model has great potential to simulate the
evacuation processes in an increasing amount of detail in
future research. For example, new patients may not arrive
to the hospital after the simulation run begins. However,
the model has the potential to check in new patients, as
long as the evacuation has not begun. New patients are
assigned information regarding acuity and expected re-
lease time. If a bed is available, the patient enters the hos-
pital; otherwise, the patient leaves the system.
In addition, a more advanced and accurate representa-
tion of staffing levels is under development at this time.
This alternative process allows staff members to enter the
system according to a hospital’s current staffing schedule.
The staff members are separated into different shifts and
specific floor assignments.
While this research focuses on hospital evacuation
due to hurricanes (where the evacuation can be planned),
no-notice evacuation of hospitals would be an extension
with great importance. Research on building evacuations
in non-hospital settings would likely be included in such
an extension.
APPENDIX: ADDITIONAL MODEL RESULTS
Table A-1: Alternative Test Set 1
TEST SET 1: Patient Count 50/50/50
Evacuation Time (hrs) Num. of
Vans
Num. of
Ambulances Average
per Patient
Average
Completion
Time (hrs)
3 3 40.4 65.1
3 6 37.6 56.4
3 9 36.8 56.2
3 12 36.4 56.0
Average 37.8 58.4
5 3 38.8 65.4
5 6 35.8 49.7
5 9 35.0 49.6
5 12 34.7 49.6
Average 36.1 53.6
8 3 38.1 65.3
8 6 35.2 48.3
8 9 34.3 48.0
8 12 34.0 48.0
Average 35.4 52.4
513
Taaffe, Johnson, Steinmann
Table A-2: Alternative Test Set 2
TEST SET 2: Patient Count 40/40/40
Evacuation Time (hrs) Num. of
Vans
Num. of
Ambulances Average
per Patient
Average
Completion
Time (hrs)
3 3 31.1 47.6
3 6 29.4 45.3
3 9 29.2 45.3
3 12 29.1 45.2
Average 29.7 45.9
5 3 29.7 47.1
5 6 28.1 41.6
5 9 27.9 41.1
5 12 27.8 41.1
Average 28.4 42.6
8 3 29.2 47.1
8 6 27.5 39.8
8 9 27.3 39.9
8 12 27.2 39.6
Average 27.8 41.6
Table A-3: Alternative Test Set 3
TEST SET 3: Patient Count 30/30/30
Evacuation Time (hrs) Num. of
Vans
Num. of
Ambulances Average
per Patient
Average
Completion
Time (hrs)
3 3 24 36.5
3 6 22.6 35.3
3 9 22.5 35.3
3 12 22.3 35.3
Average 22.9 35.6
5 3 23.1 36.4
5 6 21.8 35.3
5 9 21.6 35.4
5 12 21.4 35.4
Average 22.0 35.6
8 3 22.7 36.3
8 6 21.4 35.3
8 9 21.2 35.4
8 12 21.0 35.4
Average 21.6 35.6
Table A-4: Alternative Test Set 4
TEST SET 4: Patient Count 50/40/30
Evacuation Time (hrs) Num. of
Vans
Num. of
Ambulances Average
per Patient
Average
Completion
Time (hrs)
3 3 31.5 46.5
3 6 30.5 46.1
3 9 30.3 46.1
3 12 30.2 46.3
Average 30.6 46.3
5 3 29.2 41.1
5 6 28.1 40.9
5 9 28.0 41.1
5 12 27.9 41.1
Average 28.3 41.1
8 3 28.5 39.5
8 6 27.5 39.5
8 9 27.3 39.6
8 12 27.3 39.5
Average 27.7 39.5
Table A-5: Alternative Test Set 5
TEST SET 5: Patient Count 30/40/50
Evacuation Time (hrs) Num. of
Vans
Num. of
Ambulances Average
per Patient
Average
Completion
Time (hrs)
3 3 36.2 65.3
3 6 32.5 47.1
3 9 31.6 45.1
3 12 31.3 45.0
Average 32.9 50.6
5 3 35.3 65.1
5 6 31.6 46.9
5 9 30.7 41.7
5 12 30.3 41.4
Average 32.0 48.8
8 3 34.8 65.1
8 6 31.1 47.0
8 9 30.2 41.5
8 12 29.9 41.3
Average 31.5 48.7
514
Taaffe, Johnson, Steinmann
REFERENCES
Chang, E. C. 2003. Traffic Simulation for Effective Emer-
gency Evacuation [online]. March, 2003.
<http://www.ictpaweb.org/publicatio
n/NCA/conferean03/Proceeding/074.pd
f> [accessed July 6, 2006].
Cova, T.J. and J.P. Johnson. 2003. A network flow model
for lane-based evacuation routing. Transportation
Research Part A 37(7): 579–604.
Department of Health and Environmental Control
(DHEC). 2004. Emergency Order – Evacuation
Plans. State of South Carolina, August 30, 2004.
<http://www.scdhec.net/hr/pdfs/lice
n/licgen/hleo.pdf> [accessed July 6, 2006].
Frantzich, H. 1997. Fire Safety Risk Analysis of a Health
Care Facility. Department of Fire Safety Engineer-
ing, Lund University, October 11, 1997.
<http://www.brand.lth.se/bibl/3085.
pdf> [accessed July 6, 2006]
Franzese, O. and S. Joshi. 2002. Traffic Simulation Ap-
plication to Plan Real-Time Distribution Routes. Pro-
ceedings of the 2002 Winter Simulation Conference.
San Diego, CA.
Gladwin, C.H., H. Gladwin, and W.G. Peacock. 2001.
Modeling Hurricane Evacuation Decisions with Eth-
nographic Methods. International Journal of Mass
Emergencies and Disasters 19(2): 117–143.
U.S. Government Accountability Office (GAO). 2006.
GAO-06-443R: Preliminary observations on the
evacuation of hospitals and nursing homes due to
hurricanes. February 16, 2006.
Hobeika, A.G. and C. Kim. 1998. Comparison of traffic
assignments in evacuation modeling. IEEE Transac-
tions on Engineering Management 45(2): 192–198.
Hoffman, M. A. 2005. Unusual hurricane season drives
record tornado activity. Business Insurance 39: 4.
Law, A. and W. D. Kelton. 2000. Simulation Modeling
and Analysis. 3rd Ed. New York: McGraw-Hill.
Lindell, M.K. 2004. Hazardous Materials. American
Planning Association, Planning and Urban Design
Standards, New York: John Wiley and Sons.
McGlown, K.J. 1999. Determinants of the evacuation of
health care facilities. PhD Dissertation, University of
Alabama at Birmingham.
McGlown, K.J. 2001. Evacuation of Health Care Facili-
ties: A New Twist to a Classic Model. Natural Haz-
ards Review 2(2): 90–99.
Pidd, M., F.N. de Silva, and R.W. Eglese. 1996. A simu-
lation model for emergency evacuation. European
Journal of Operational Research 90(3): 413–419.
Pollak, E., M. Falash, L. Ingraham, and V. Gottesman.
2004. Operational Analysis Framework for Emer-
gency Operations Center Preparedness Training. In
Proceedings of the 2004 Winter Simulation Confer-
ence, ed. R. G. Ingalls, M. D. Rosetti, J. S. Smith,
and B. A. Peters, 839–848.
Rogers, G.O. 1994. The timing of emergency decisions:
Modelling decisions by community officials during
chemical accidents. Journal of Hazardous Materials
37: 353–373.
Sheffi, Y., H. Mahmassani, and W.B. Powell. 1982. A
transportation network evacuation model. Transpor-
tation Research Part A 16(3): 209–218.
Sorensen, J. H., B. L. Shumpert, and B. M. Vogt. 2004.
Planning for protective action decision making:
evacuate or shelter-in-place. Journal of Hazardous
Materials 109: 1–11.
Taaffe, K., R. Kohl, and D. Kimbler. 2005. Hospital
Evacuation: Issues and Complexities. In Proceedings
of the 2005 Winter Simulation Conference, ed. M. E.
Kuhl, N. M. Steiger, F. B. Armstrong, and J. A.
Joines, 943–950.
Taaffe, K., and E. Tayfur. 2006. Evaluating the Robust-
ness of Hospital Evacuation Plans. Forthcoming in
Proceedings of the 2006 Industrial Engineering Re-
search Conference. Orlando, FL.
Tufekci, S. 1995. An integrated emergency management
decision support system for hurricane emergencies.
Safety Science 20(1): 39–48.
Vogt, B.M. 1991. Issues in Nursing Home Evacuations.
International Journal of Mass Emergencies and Dis-
asters 9(2): 247–265.
AUTHOR BIOGRAPHIES
MATTHEW JOHNSON is an undergraduate student in
the Department of Industrial Engineering at Clemson Uni-
versity. He graduated in May of 2006. His email address
is <mbjohns@clemson.edu>.
DESIREE STEINMANN is an undergraduate student in
the Department of Industrial Engineering at Clemson Uni-
versity. She graduated in May of 2006. Her email ad-
dress is <dsteinm@clemson.edu>.
KEVIN M. TAAFFE is an assistant professor in the De-
partment of Industrial Engineering at Clemson University.
His research interests include transportation and logistics
systems analysis as well as production, inventory, and
demand management. He is a Senior Member of IIE, as
well as a member of INFORMS, Transportation Research
Board, and Decision Sciences Institute. His email address
is taaffe@clemson.edu and his Web address is
http://people.clemson.edu/~taaffe/.
515
... These articles are briefly summarised. Taaffe, Johnson, and Steinmann (2006) applied simulation to estimate the time required to evacuate patients from a hospital to receiving healthcare centres, assuming there are three different types of patients and patients could be moved to three receiving hospitals using a fixed number of ambulances. Tayfur and Taaffe (2009) developed a mixed integer programming formulation to find the scheduling and allocation of resources to minimise total operating and evacuation costs. ...
... Table 5 shows the majority of studies assumed the number of patients that should be evacuated is fixed and known at the beginning of the planning horizon, while Ra ani, Zhale hian, and Fa h a -e an a eh (201 ) and Kim et al. (2020) considered this parameter as a stochastic parameter. In addition, Taaffe, Johnson, and Steinmann (2006) assumed more patients could be added before starting the evacuation. In addition, in some studies, it is assumed that when transportation starts, all patients are already in the nearby staging area, while in real conditions, the preparation of patients is time-consuming, and they are gradually moved to pick up points like the study by Chen, Guinet, and Ruiz (2015). ...
... Human resources assignment and scheduling are still broadly missing from many hospital transport evacuation plans. In the literature, some studies assumed that there are several groups of human resources, for example, nurses with different levels of skills, that should be assigned in hospital evacuation planning including (Taaffe, Johnson, and Steinmann 2006;Taaffe 2009, 2009;Chen, Guinet, and Ruiz 2015), while others did not incorporate human resources in their studies , e.g. (Bish, Agca, and Glick 2014;Rabbani, Zhalechian, and Fa h a -e an a eh 201 ). ...
Article
Full-text available
The frequency and intensity of both human-made and natural disasters are predicted to increase, and hospitals play a critically important role in reducing injury and mortality rates. However, there is increasing evidence that many hospitals are vulnerable to disasters, and more effective strategies are needed to enable the safe evacuation of patients to alternative healthcare centres. Transport planning is central to this process, but there has been no systematic and critical review of the research on this critically important challenge. This means there is no collective synthesis of this literature for hospital managers, policymakers, and researchers to refer to in addressing these important vulnerabilities. This paper reports the findings of a critical literature review to contribute practical insights for health facilities planning and management decision-making in a context where both the likelihood and consequences of natural disasters are increasing in many countries.
... Despite substantial research assessing previous experiences in which natural disasters have influenced hospital healthcare delivery, there is a shortage of modelling research in the literature addressing transportation during hospital evacuation. Taaffe, Johnson and Steinmann [59] used simulation to assess how long it would take to transfer patients from a hospital to other healthcare facilities. To minimise operational and evacuation costs, Tayfur and Taaffe [41] suggested a mixed integer programming framework for determining resource scheduling and allocation. ...
Article
Full-text available
Extreme weather events such as floods are predicted to become increasingly common and severe as the climate changes. Effectively functioning hospitals are critical to a community's resilience to the adverse health impacts of such events. Yet many hospitals have not been designed with extreme weather risks in mind and are built in flood-prone areas, raising concerns about their ability to support community healthcare needs when they eventuate. While considerable research has been conducted on developing disaster responses to maintain healthcare services in the face of floods, there is a paucity of research on hospital evacuating planning. Addressing this critical gap in research, this paper explores current state-of-the-art hospital evacuation transportation models to identify key factors for developing more effective evacuation planning for future flooding risks. A new modelling framework is proposed, which addresses the limitations of existing hospital evacuation models and proposes factors that should be incorporated into future models to enhance community resilience to growing flood risks due to climate change.
... At present, plenty of researches are concentrated on crowd evacuation in hospitals, in terms of (1) human psychology and behaviors in emergency by questionnaires and interviews [1,18,23,24], (2) pedestrian movement characteristics [1,[25][26][27][28][29], and (3) optimization of evacuation strategies [7,26,[30][31][32][33][34][35][36][37][38][39]. Although the results from questionnaires and interviews could inform government agents that they should consider the evacuee's emotion and special behaviors when conducting fire safety management. ...
Conference Paper
Full-text available
Fire accident remains one of the most significant challenges facing the safety management of hospitals. Existing statistics have shown that hospital fires always cause severe casualties when occurred, and how to safely evacuate people from hospital fires is significant yet has not been systematically investigated by now. In this paper, state-of-the-art studies in hospital evacuation under fire are discussed. First, we identify the risks in hospitals that may lead to fires and hinder the evacuation process. Next, the whole evacuation process under hospital fires is presented and discussed, along with our future prospects. The disabled patients and those involving the use of wheelchairs/medical beds are crucial roles during the evacuation, and many efforts should be devoted to them,including elevator evacuation and assistance from the medical staff. Moreover, integrating Building Information Modeling and facilities management for establishing effective evacuation strategies may be worth studying. We expect that this work could bring a more comprehensive insight for the architects and engineers.
... Over the past decade, a number of researches have investigated transit-based and multimodal evacuation techniques that are based on the optimization of scheduling and operations [20][21][22][23]. Along with research concerning transit-based evacuation of the vulnerable population, studies focusing on persons with disabilities [24], availability of hospitals and special facilities [25,26], as well as those concerning the elderly [27,28] have drawn less attention in recent years in view of the challenges-lack of personnel, monetary resources, engagement, etc.-associated with their solutions. The biggest problem facing the evacuation of vulnerable populations involves identifying locations of these people, many of whom are not recognized by official systems. ...
Article
Full-text available
Under no-notice evacuation scenarios with limited time horizons, the effectiveness of evacuation can be negatively impacted by intermediate trips that are made by family members and the identification of vulnerable populations. The emergence of shared-mobility companies, such as Uber and DiDi, can be considered as a potential means to address above-mentioned concerns. The proposed study explores the utility of shared-mobility services under emergency-evacuation scenarios and makes recommendations to relevant bodies that are based on the obtained and they are discussed herein. The study investigates attitudes of the public, experts, and drivers towards the use of shared-mobility resources during emergency evacuations based on a stated preference survey. Results of questionnaires, driver interviews, and face-to-face expert interviews have been analyzed to validate the feasibility and identify potential problems of leveraging shared-mobility services during evacuation response, especially in metropolitan areas wherein such services are already ubiquitous. Numerical simulations have been performed to quantify potential improvements in the total trip distance and number of evacuees after incorporating the use of shared mobility into emergency-response operations. However, despite the observed improvement in emergency efficiency, certain realistic roadblocks must be overcome. Realization of the proposed objective heavily depends on actionable policy recommendations, provided herein as a reference for the government, emergency management agencies, and shared-mobility companies.
... Today, behavior representation is also used in entertainment, commercials and non-educational simulations. The field of application ranges from crowd control [17] and evacuation planning [18] to traffic density [8] and safety [3]. ...
Conference Paper
Full-text available
In this work, we present a GPU-based library, called BehaveRT, for the definition, real-time simulation, and visualization of large communities of individuals. We implemented a modular flexible and extensible architecture based on a plug-in infrastructure that enables the creation of a behavior engine system core. We used Compute Unified Device Architecture to perform parallel programming and specific memory optimization techniques to exploit the computational power of commodity graphics hardware, enabling developers to focus on the design and implementation of behavioral models. This paper illustrates the architecture of BehaveRT, the core plug-ins, and some case studies. In particular, we show two high-level behavioral models, picture and shape flocking, that generate images and shapes in 3D space by coordinating the positions and color-coding of individuals. We, then, present an environment discretization case study of the interaction of a community with generic virtual scenes such as irregular terrains and buildings.
Book
Full-text available
يساهم هذا الكتاب في وضع تصورات مستقبلية لصانعي القرار في المستشفيات الكبرى، حول المنهجية التى تتبع في مجال إخلاء وايواء وعزل المستشفيات، سواء من الناحية النظرية أو العملية بشكل مفصل، ويعالج القضايا المعقدة المرتبطة بوقوع أخطار مزدوجة تتطلب اتخاذ قرارات حادة في إدارة المستشفيات أثناء عمليات الإخلاء والايواء والعزل، ويقدم معلومات مهمة ومترابطة، تشكل أساسا علميا، لكل المهتمين والمختصين في إدارة الطوارئ، كما أنه يشكل منطلقا علميا هاما ، للمراجع البحثية المختلفة التي لها علاقة في موضوعات الكتاب.
Article
Full-text available
Background Resource allocation in patient care relies heavily on individual judgements of healthcare professionals. Such professionals perform coordinating functions by managing the timing and execution of a multitude of care processes for multiple patients. Based on advances in simulation, new technologies that could be used for establishing realistic representations have been developed. These simulations can be used to facilitate understanding of various situations, coordination training and education in logistics, decision-making processes, and design aspects of the healthcare system. However, no study in the literature has synthesized the types of simulations models available for non-technical skills training and coordination of care. Methods A systematic literature review, following the PRISMA guidelines, was performed to identify simulation models that could be used for training individuals in operative logistical coordination that occurs on a daily basis. This article reviewed papers of simulation in healthcare logistics presented in the Web of Science Core Collections, ACM digital library, and JSTOR databases. We conducted a screening process to gather relevant papers as the knowledge foundation of our literature study. The screening process involved a query-based identification of papers and an assessment of relevance and quality. Results Two hundred ninety-four papers met the inclusion criteria. The review showed that different types of simulation models can be used for constructing scenarios for addressing different types of problems, primarily for training and education sessions. The papers identified were classified according to their utilized paradigm and focus areas. (1) Discrete-event simulation in single-category and single-unit scenarios formed the most dominant approach to developing healthcare simulations and dominated all other categories by a large margin. (2) As we approached a systems perspective (cross-departmental and cross-institutional), discrete-event simulation became less popular and is complemented by system dynamics or hybrid modeling. (3) Agent-based simulations and participatory simulations have increased in absolute terms, but the share of these modeling techniques among all simulations in this field remains low. Conclusions An extensive study analyzing the literature on simulation in healthcare logistics indicates a growth in the number of examples demonstrating how simulation can be used in healthcare settings. Results show that the majority of studies create situations in which non-technical skills of managers, coordinators, and decision makers can be trained. However, more system-level and complex system-based approaches are limited and use methods other than discrete-event simulation.
Article
Full-text available
Tsunami is not only causing the raise of free sea surface but also strong horizontal water movement in a bay. Thus, an appropriate countermeasure is required for a passenger ship because the ship may be forced by tsunami to move unexpectedly. In this case, tsunami acts as a variable to limit evacuation time of passengers from a ship to shelter areas. For instance it is estimated that the time to tsunami arrival at Tempozan Passenger Terminal of Osaka Port, that is available time for passengers to evacuate is around 1 hours and 40 minutes (6000 sec). Passengers evacuation, which is caused by tsunami attacks has different problems since passengers not only have to evacuate from a ship, but also have to run to safe areas where they do not receive the impact of tsunami, such as in high land area. This paper describes the use of Discrete Event Simulation (DES) of passengers evacuation process due to tsunami attack considering several scenarios, such as in the morning, day and night scenario considering MSC.1/Circ.1033; therefore, it will be known that how many people are safe in the case of such incident. The outputs of analysis are prediction of time evacuation; as a result, they can be used as consideration of port safety management and evacuation procedure in case of tsunami disaster.
Article
This paper presents a review of highway-based evacuation modeling and simulation and its evolution over the past decade. The review includes the major components of roadway transportation planning and operations, including the current state of modeling in the forecasting of evacuation travel demand, distribution and assignment of evacuation demand to regional road networks to reach destinations, assignment of evacuees to various modes of transportation, and evaluation and testing of alternative management strategies to increase capacity of evacuation networks or manage demand. Although this discussion does not cover recent work in other modes used in evacuation such as air, rail, and pedestrian, this paper does highlight recent interdisciplinary modeling work in evacuation to help bridge the gap between the behavioral sciences and engineering and the application of emerging techniques for the verification, validation, and calibration of models. The manuscript also calls attention to special considerations and logistical difficulties, which have received limited attention to date. In addition to these concerns, the following future directions are discussed: further interdisciplinary efforts, including incorporating the medical community; using new technologies for communication of warnings and traffic condition information, data collection, and increased modeling resolution and confidence; using real-time information; and further model refinements and validation.
Article
Full-text available
This paper directly models individual and household hurricane evacuation behavior using ethnographic decision tree analysis...
Article
Full-text available
Protecting the public from airborne chemical releases is limited by the timing of the implementation of actions taken and the capacity of those actions to avoid or reduce exposure. This paper examines the community decision processes during emergencies to identify critical factors asso- ciated with the timing of emergency warning, and protective action recommendations. This research examines the decision process by tracing emer- gency response from the outset of the community decision process, through the decision to warn the public, including the communication of hazard to the public, and the all-clear at the end of the emergency period. Both community authorities and the public cycle through hazard detection, assessment communication, and behavioral response as they become aware of the hazard. A sample of emergency decisions during chemical emergen- cies was examined via post-emergency interviews with key community officials. Emergency responders in a systematic sample of events after 1984, but prior to 1990, were interviewed in the Fall of 1989. Finding that decisions in more recent events were more easily reconstructed, a randomly selected half of the significant chemical emergencies occurring during 1990 were interviewed within weeks of the chemical events. Previous work [1] shows that community decision processes are seldom immediate and often involve information seeking. This descriptive work is expanded herein to provide better models of the key factors effecting decision processes in chemical emergencies. Regression models of these data indicate that protective action and warning decisions occur more rapidly than all-clear decisions, and that each decision is influenced by different factors in the decision process. Moreover, these data indicate that the role of experts changes throughout the emergency response. When decisions lead to the active avoidance of exposure, officials seem to take evasive action more quickly, but when failure to decide results in passive avoidance of exposure and continued in convenience of the public, the decision process is often protracted.
Article
Full-text available
Most traffic delays in regional evacuations occur at intersections. Lane-based routing is one strategy for reducing these delays. This paper presents a network flow model for identifying optimal lane-based evacuation routing plans in a complex road network. The model is an integer extension of the minimum-cost flow problem. It can be used to generate routing plans that trade total vehicle travel-distance against merging, while preventing traffic crossing-conflicts at intersections. A mixed-integer programming solver is used to derive optimal routing plans for a sample network. Manual capacity analysis and microscopic traffic simulation are used to compare the relative efficiency of the plans. An application is presented for Salt Lake City, Utah.
Article
A methodology for deriving individual risk to people in a health care facility due to fire, is presented. The risk is expressed by the First Order Second Moment reliability index β. The evacuation time for the patients is compared to the time to untenable conditions arise. Subjecting variables to uncertainty results in a reliability index value for the escape time margin. The influence on reliability index of various number of staff is examined. The scenarios are presented with the event tree methodology.
Article
Disasters and their potential for damage are increasing in the United States and worldwide. Government officials warn that disasters will be occurring at an increasing rate and will worsen in the future. As these risks increase, the likelihood of evacuation of a health care facility grows. We know much about the role of hospitals in evacuation, yet little about how the decision to evacuate a facility is initially made. This study compares two models from studies related to the evacuation process: Vogt's adaptation of Quarantelli's classic model examining organizational ability to cope, and McGlown's model from variables considered critical to health care executives in the decision to evacuate a health care facility. This study contributes to the literature on evacuation behavior of health care organizations by examining the variables perceived by decision makers as critical in their decision to evacuate their facility compared to variables identified in the literature. These findings may form the first constructive step toward the development of an empirically based decision framework for health facility evacuation.
Article
This paper describes the development of a prototype spatial decision support system for use by emergency planners in developing contingency plans for evacuations from disaster areas. It links together a geographical information system (ARC/INFO) with a specially written object-oriented micro-simulator via a windowing computer operating system. The details of the system are described, its limitations are discussed and potential enhancements are identified.