Spatial and Spatio-temporal Epidemiology 30 (2019) 100285

Contents lists available at ScienceDirect

Spatial and Spatio-temporal Epidemiology

journal homepage: www.elsevier.com/locate/sste

Hospital surge capacity for an influenza pandemic in the triangle

region of North Carolina
Rachel L. Woodul a,∗, Paul L. Delamater a,b, Michael Emch a,b
a
Department of Geography, The University of North Carolina at Chapel Hill, NC United States
b
Carolina Population Center, NC United States

a r t i c l e i n f o a b s t r a c t

Article history: This research investigates the geographic aspects of health care delivery in the event of a sudden increase
Received 1 November 2018 in the need for care. We constructed an integrated disease outbreak and surge capacity model to evaluate
Revised 17 June 2019
the ability of a region’s healthcare system to provide care in the event of a pandemic. In a case study, we
Accepted 20 June 2019
implement the model to investigate how an influenza pandemic similar to the 1918 Spanish Flu pandemic
Available online 8 July 2019
would affect the population of the Raleigh-Durham-Chapel Hill metropolitan statistical area and the abil-
Keywords: ity of the region’s hospital system to respond to such an event. Under varying scenarios for hospital
Influenza capacity, we found that the population needing care would overwhelm the system’s ability to provide
Spatial model care in the case study. Our model is presented as a framework that can be augmented and expanded to
Surge capacity suit the needs of the particular event and healthcare system or services required. By integrating concepts
Hospital and models from epidemiology, geography, and health care services research, we provide a valuable tool
Pandemic
for potential use in disaster planning, hospital system evaluation, and pandemic preparedness.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction pandemic influenza (Monto et al., 2006). Given little or no exist-

Influenza, or flu, is a contagious respiratory illness caused by
infection of the nose, throat, and lungs with an influenza virus.
Infection can cause mild to severe illness, with serious cases
requiring hospitalization and potentially resulting in death. In-
fluenza is highly contagious; symptoms of infection typically de-
velop 1–4 days after exposure, and infected individuals are able
to spread the virus to others beginning at ∼1 day before the on-
set of symptoms and ∼5–7 days after becoming ill (CDC, 2017b).
In the US, seasonal influenza results in 9.2–35.6 million illnesses,
140,0 0 0–710,0 0 0 hospitalizations, and 12,0 0 0–56,0 0 0 deaths annu-
ally (CDC, 2017a). The impact of seasonal influenza, whether it be
the severity of the strain or the number of people who become ill,
varies greatly from year to year based on the strain of influenza
virus that has become epidemic during that season.
Sometimes, as a result of natural genetic variation, a different
type of influenza virus emerges that is unlike any other circulat-
ing strain. In these cases, yearly vaccines are ineffective and no
residual immunity exists in the human population. Unique varia-
tions in viral strain is the greatest contributor to occurrences of
Abbreviations: CDC, Centers for Disease Control and Prevention; ILI, Influenza-
like-illness; RTP, Research Triangle Park; BGs, Block Groups; OD, Origin Destination;
SIR, Susceptible, Infected, Recovered.
∗
Corresponding author.
E-mail address: rachel.woodul@unc.edu (R.L. Woodul).
ing immunity in the human population, more people will develop
an influenza infection than during a typical seasonal epidemic. Fur-
ther, because of the infection’s latent period, many people are able
to transmit the virus before they realize they are infected, and
some individuals will spread the virus despite experiencing very
few symptoms. Though rare, small outbreaks or local epidemics of
a novel influenza virus can quickly become a pandemic. Four in-
fluenza pandemics have taken place since 1900 (1918, 1957, 1968,
and 2009). The 1918 H1N1 influenza A pandemic, or the “Spanish
Flu” pandemic, is probably the most well-known. About one-third
of the world’s population was infected and more than 50 million
people died as a result of infection, making it the most severe in-
fluenza pandemic in recent years and one of the deadliest pub-
lic health crises in human history (CDC, 2017c; Morens and Fauci,
2007).
When compared to a typical seasonal influenza epidemic, a
novel influenza strain that becomes pandemic could increase death
rates by more than 56% (CDC, 2017d), quickly overwhelming health
care systems and having catastrophic consequences. Should a pan-
demic similar to the 1918 Spanish Flu pandemic occur today, the
number of both infected individuals and fatalities would greatly
exceed typical rates of seasonal influenza. Influenza-like-illness
(ILI) levels are indicative of the magnitude of individuals that will
demand care from healthcare facilities throughout the course of
an epidemic or pandemic, and unusually high ILI levels may re-
sult in unexpected increase in the need for care. The ability of the

https://doi.org/10.1016/j.sste.2019.100285
1877-5845/© 2019 Elsevier Ltd. All rights reserved.
2 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285
healthcare system to respond to this increased demand for care
is known as “surge capacity” (Barbisch and Koenig, 2006). During
such a competition for hospital resources, it is likely that many in-
dividuals in need of care may not be able to receive it due to the
finite amount of resources and space available. In large-scale crisis
events, an individual’s ability to utilize resources may be limited by
their physical distance from the health care facility. It is likely that
a novel influenza pandemic would quickly overwhelm the current
US health care system’s ability to provide care (Daugherty et al.,
2010; Osterholm, 2005).
Understanding surge capacity is an essential part of pandemic
preparedness. Currently, the US Centers for Disease Control and
Prevention (CDC) maintains a pandemic influenza simulation
model, FluSurge, that quantifies the potential impacts of a pan-
demic in the context of expected demand for hospital care. The
model is specifically designed to assess potential increase in de-
mand for care at a facility level and provide estimates of resource
usage in the event of an influenza pandemic. As a pandemic
model, the integration of surge capacity and potential population
case rates makes FluSurge a valuable tool to use for assessing
pandemic preparedness of health care facilities. Yet, while it
provides a thorough output of estimated impacts, FluSurge can
only be used to model increases in demand at a single facility,
as opposed to a system of facilities that provide health care for
the population within a larger region. There have been studies
in which FluSurge was used to evaluate the existing capacity of
a single hospital (Eyck, 2008; Lum et al., 2009), and there has
also been research investigating single hospital capacity for a
variety of disaster events, including pandemics (Kumar, 2011; Yi
et al., 2010). However, few models exist that integrate pandemic
influenza planning and hospital system capacity in an explicit
manner (Dugas et al., 2012; Sobieraj et al., 2007).
The aim of this research is to construct an integrated disease
outbreak and surge capacity model to evaluate the ability of a local
healthcare system to provide care in the event of a global influenza
pandemic. Using the Raleigh-Durham-Chapel Hill, North Carolina
metropolitan statistical area (the Triangle) as a case study, we in-
vestigate how an influenza pandemic similar to the 1918 Spanish
Flu pandemic would affect the population and the ability of the
region’s hospital system to respond to such an event. We use epi-
demiological and spatial modeling to estimate the number of peo-
ple that would become infected and require hospital care in an in-
fluenza pandemic, as well as their geographic distribution. Then,
we implement a capacitated shortest distance allocation model to
allocate those needing care to the nearest hospital with available
treatment space. This approach allows us to demonstrate the time
during an influenza pandemic that the hospital system would ex-
ceed its capacity and ability to provide care to infected individu-
als. Furthermore, we are able identify which individuals (by geo-
graphic location) would not receive care when capacity were to be
exceeded.
This study contributes to both the existing literature and the
currently evolving discussions regarding large-scale disaster plan-
ning, health care system response, and pandemic preparedness,
as it specifically investigates the geographic aspects of health
care delivery in the event of a sudden increase in the need for
care.
2. Methods and materials
2.1. Study area
The Raleigh-Durham-Chapel Hill Metropolitan Statistical Area of
North Carolina (the Triangle) is the second largest metropolitan
region in North Carolina. The region is made up of seven coun-
ties, which vary from highly urbanized to rural. Anchored by The
University of North Carolina at Chapel Hill, Duke University, and
North Carolina State University, the Triangle also contains fourteen
other universities and colleges, as well as the Research Triangle
Park (RTP), the largest research park in the United States, from
which the region takes its name. The Triangle presents an inter-
esting case study for this work because the population is highly
mobile and demographically and socioeconomically diverse; fur-
ther, because of the abundance of educational institutions and the
RTP, there are highly concentrated centers of young adults near the
urbanized areas. The health care infrastructure is well developed.
There are eleven hospitals within the region, two of which are ma-
jor research hospitals.
2.2. Data acquisition and pre-processing
2.2.1. Population and infrastructure data
A spatial data layer of the US Census Block Groups (BGs) for
North Carolina and tables containing the state’s 2010 population
by age group and sex were downloaded from the IPUMS National
Historical Geographic Information System database (Manson et al.,
2012). Total population counts by age group for each BG were cre-
ated by summing the male and female age group populations. The
tabular data was then joined to the spatial data layer and subset
to BGs within the study area.
The spatial data layer of hospital locations in North Carolina
was downloaded from the NC OneMap GeoSpatial Portal (http://
data.nconemap.gov/geoportal/catalog/main/home.page) and subset
to the study area. The number of licensed beds at each hospi-
tal was gathered from the North Carolina Department of Health
and Human Services – Division of Health Service Regulation (https:
//www2.ncdhhs.gov/dhsr/data/hllist.pdf). Fig. 1 shows the distribu-
tion of the total population by BG and the hospital locations (and
number of beds) in the Triangle region.
A roads data layer for North Carolina was downloaded from
the North Carolina Department of Transportation (https://connect.
ncdot.gov/resources/gis/pages/gis- data- layers.aspx). The roads layer
required a number of data cleaning steps prior to conversion to a
travel network dataset. First, connected road segments sharing the
same attributes were dissolved into a single segment. Second, road
connectivity at intersections was established by enforcing planar
topology in the layer. Third, minor digitizing errors (undershoots
and overshoots) in the data were corrected using the automated
editing tools in ArcGIS. After cleaning the roads layer, it was con-
verted to a network dataset.
Using the road network dataset, we calculated the road-based
shortest travel distance between all BGs and hospitals in the study
area. Geographic centroids of the BGs were used as the location
from which to measure distance. This resulted in an Origin Desti-
nation (OD) matrix containing the shortest road distance from ev-
ery BG to every hospital in the study area.
2.2.2. Influenza data
Age group-specific clinical case rates from the 1918 Spanish In-
fluenza were gathered from Brundage (2006), who calculated av-
erages from the original 1919 case surveillance data at local, re-
gional, and global scales (Frost, 1919, 1920; Frost and Sydenstricker,
1919). The total number of expected influenza cases requiring
hospitalization in the Triangle region was calculated by multiply-
ing the age group-specific clinical case rates by the correspond-
ing age group population counts in the region. Summing the age
group cases resulted in an estimate of 399,801 people that would
require hospitalization for the entire region. The age groups, clin-
ical case rates, population counts (and relative percent of the Tri-
angle population), and expected influenza cases are presented in
Table 1.
 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285 3

Fig. 1. Population of the Triangle by block group and hospital locations.

Table 1 with the same clinical case rate as the 1918 Spanish Influenza in
Age groups, clinical case rates (CCR, cases per 10 0 0 people), population (POP), rela-
the 2010 population of the Triangle region. First, we divided the
tive percent of the Triangle population (% Total POP), expected influenza cases (INF),
and relative percent of the Triangle’s expected influenza cases (% INF). count of total infected persons for the study region by the total
population of the study region (N) to calculate the proportion of
Age group CCR POP % Total POP INF % INF
the population infected by the end of the pandemic event (zƒ ):
0–4 320 112,873 7.24 36,119 9.03
5–9 375 108,391 6.96 40,646 10.17 count o f total in f ected
zf = ,
10–14 360 106,305 6.82 38,269 9.57 N
15–19 330 111,111 7.13 36,666 9.17
20–24 320 111,036 7.13 35,531 8.89 resulting in a value of zƒ = 0.26. We then used the zƒ value to esti-
25–29 325 116,692 7.49 37,924 9.49 mate the R0 value:
30–34 320 119,652 7.68 38,288 9.58  
35–39 280 124,254 7.97 34,791 8.70 log 1 − z f
40–44 225 122,255 7.85 27,507 6.88 R0 = − ,
zf
45–49 200 117,718 7.55 23,543 5.89
50–54 160 104,328 6.70 16,692 4.18 which produced an R0 = 1.15. We calculated the rate of recovery
55–59 130 86,590 5.56 11,256 2.82
(γ ) using seven days as the infectious period in which an infected
60–64 120 71,106 4.56 8532 2.13
65–69 115 48,346 3.10 5559 1.39 person is able to transmit the virus to a susceptible individual
70–74 100 33,354 2.14 3335 0.83 (CDC, 2017b).
≥ 75 80 64,191 4.12 5135 1.28
1
γ= ,
in f ectious period
producing a γ value of 0.14. Lastly, using the calculated rate
2.3. Pandemic simulation model of recovery (γ ) and R0 value, we estimated the rate at which
individuals will move from susceptible to infected (β ) (Liu and
A susceptible, infected, and recovered (SIR) model without vi- Stechlinski, 2017):
tal dynamics was used to calculate the daily counts of susceptible,
β = R0 ∗ γ .
infected, and recovered people in the region, given similar charac-
teristics to the 1918 Spanish Influenza and a modern population. For this population, β = 0.165. Other parameters required to
Using the total count of infected persons, we estimated the in- solve the SIR model were a pandemic time of 731 days, which is an
fectivity parameters necessary to produce an influenza pandemic estimation of the time of the 1918 Spanish Influenza (Barry, 2004),
4 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285
and a starting count of infected persons of 15 people (∼.001% of
the population). Setting the initial infected population at 15 peo-
ple, rather than a single index case, accounted for the potential
that multiple infected individuals could enter the Triangle dur-
ing a pandemic event due to the high mobility of the population.
The remainder of the population was assumed to be susceptible
to influenza infection. We used the deSolve library in R (Soetart
et al., 2010) to solve the differential equations and produced the
daily counts of people susceptible, infected, and recovered over the
study period.
The number of people infected each day (from the SIR model)
needed to be redistributed to the age groups and to the BGs to
produce the spatial distribution of infected people per day. The
daily counts of infected people were first distributed to age groups
based on the relative percent values found in Table 1 by multiply-
ing each day’s count by the percent values for each group. The age
group-specific daily counts were then distributed back to the BGs
based on the percent of the age group population in each block
group. For example, the number of people aged 0–5 in one of the
BGs was 177 and the total number of people aged 0–5 in the study
area was 112,873. As such, this BG contained 0.15% of all people in
this age group. If the daily count of infected people aged 0–5 was
74, this BG would be assigned 74 ∗ 0.15% = 0.11 infected people for
this age group. This process was implemented for each age group
and BG to calculate the age group-specific counts of infected peo-
ple per block group per day. Each BG’s age group-specific counts
were then summed to calculate the total count of infected people
per day.
2.4. Allocation model
2.4.1. General modeling approach and assumptions
A capacitated shortest distance allocation model was developed
for this study. The model assumes that the infected members of
the population will seek care at similar health care services, in
this case, acute care hospitals located within the boundaries of the
study area. The model assumes that individuals will attempt to ac-
cess the facility closest to them (as defined by road travel distance)
having available capacity. Hospital capacity is represented by num-
ber of hospital beds, which is used as a proxy for space, resources,
and providers. The model also assumes that all infected members
of the population are attempting to access healthcare facilities si-
multaneously each day; though this is not an entirely realistic sce-
nario, it allows for the highly nuanced influence of distance on
hospital utilization to be captured.
In the model, once an individual has been hospitalized,
they remain in the hospital and occupy a bed for three days.
Davis et al. (2005) assessed surge capacity and discharge rates for
mass-casualty events at multiple hospitals and found that, in a
large portion of mass-casualty events, patients can be discharged
within 24–72 hours (h) after admission. We chose to model the
hospitalization period for those infected with influenza as 72 h.
Once a hospital’s capacity is reached in the model, the hospital
will stop admitting patients and infected people needing hospital-
ization attempt to access their next closest facility. In the model, if
all the hospitals reach capacity, the remaining infected people will
not receive care.
2.4.2. Model implementation
We first ordered the OD matrix of BG to hospital driving dis-
tances from shortest to longest distance. This was used as the or-
der in which infected individuals would attempt to access hospi-
tals. The basic rules of the allocation model were as follows:
• If all infected people in a BG were allocated to a hospital, the
remaining distance pairs (BG to hospital) for that BG were re-
moved from the OD table.
• If the number of infected people in a BG was greater than the
remaining capacity at the nearest hospital, the number that
would bring the hospital to capacity were allocated to the hos-
pital. The remaining number infected people for the BG was
noted. Because the hospital’s capacity was reached (could no
longer accept people that day), the remaining distance pairs
(BG to hospital) for that hospital were removed from the OD
table.
For each day, the allocation of infected people proceeded by it-
erating through the distance OD table (which was updated after
each iteration to account for the number of infected people al-
located, the number of infected people remaining, and remaining
hospital capacity) until (1) all infected people were either hospi-
talized or (2) the system’s hospital capacity was reached. If the
system’s hospital capacity was reached, the remaining individuals
were considered to have been “turned away,” thus not receiving
care.
At the beginning of each successive day in the allocation
model, the patients that had been hospitalized for three days were
discharged (and removed from the hospital’s capacity), while the
patients that had been hospitalized for two days were shifted to
being hospitalized for a third day and the patients that had been
hospitalized the previous day were shifted to being hospitalized
for a second day. As such, each hospital’s remaining capacity
was updated prior to the allocation process of each day’s newly
infected people. This allocation process was repeated for each day
in the simulated pandemic.
2.4.3. Surge capacity scenarios
We implemented three hospital surge capacity scenarios. In the
first scenario, maximum hospital capacity was set as the number
of licensed beds for each hospital. This assumes that all licensed
beds are available for in-patient care of individuals infected with
influenza and that each licensed bed is accompanied by an ap-
propriate supply of medications, ventilators, medical profession-
als, and other necessary supplies. Though somewhat unrealistic,
this assumption was made to model a scenario in which the avail-
able hospital resources are directed towards treating pandemic in-
fluenza victims.
The second scenario represented hospital capacity as twice the
number of licensed beds. This is representative of a scenario in
which hospitals are able to provide one additional treatment space
(and necessary care) for each licensed bed. Provision of additional
patient care spaces may include setting up alternate care sites in
nearby public facilities, converting procedure suites into patient
care spaces, providing temporary beds and cots in open spaces like
hallways and lobbies, or erecting temporary care clinics outside the
hospital building in tents (Barbisch and Koenig, 2006; Davis et al.,
2005; Eastman et al., 2007).
The third scenario set maximum hospital capacity for people
with influenza infections at 75% of the licensed beds. This scenario
represents a situation wherein hospitals are required to provide in-
patient care for people having non-influenza acute care conditions
during the influenza pandemic. Therefore, not all resources will be
available to treat people with influenza. For the purposes of this
model, all hospitals were constrained to 75% of their licensed beds
for providing care for people with influenza.
3. Results
3.1. Pandemic simulation
The results of the SIR model with an outbreak period of 731
days and 15 initially infected individuals can be found Fig. 2. The
total number of infected people in the Triangle region over the
 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285
Fig. 2. SIR model results for each day in the simulated influenza pandemic. The
black line indicates percent of the population who are susceptible, the red line in-
dicates percent who are infected, and the green line indicates the percent of the
population who are recovered.
731-day period was 399,583, which nearly equals the number es-
timated from the age group-specific clinical case rates (399,801).
The number of new infections per day begins to rapidly increase
after about 200 days, which is about the length of typical influenza
Fig. 3. Expected count of individuals infected with influenza per block group.
5
season in the United States. The number of newly infected people
peaks at day 368. The geographic output of the infected population
for the entirety of the pandemic event is mapped by BG in Fig. 3.
Much of the variation in influenza cases across the study area is
due to the distribution of the raw number of people; however, the
distribution of cases was also influenced by the demographic com-
position of each BG and variations in clinical case rate among age
groups.
3.2. Hospital capacity
The capacitated shortest distance allocation model with maxi-
mum hospital capacity defined as the number of licensed beds re-
sulted in 317,141 individuals (79.4% of the infected population) be-
ing able to receive care. 82,485 individuals (20.6% of the infected
population) did not receive care. At its peak on Day 371 demand
for care exceeded available beds by roughly 126%. Maximum ca-
pacity for each hospital in the study area was reached at various
times throughout the study period (Fig. 4). Maximum capacity for
all hospitals in the system was reached on day 297 of the pan-
demic and maintained until Day 445. BGs located near the study
area boundary had higher rates of infected individuals not able to
receive care; however, BGs immediately bordering the urban re-
gions had the highest counts of individuals not receiving care due
to differences in the population counts (Fig. 5). The populations of
the more rural BGs are smaller than the populations of the BGs
located closer to urban centers.
The second scenario considered maximum hospital capacity as
double the number of licensed beds. By doubling the available
space, all infected individuals received care. Maximum capacity
6 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285

Fig. 4. Hospital usage for each day in the simulated influenza pandemic using the 100% capacity scenario. The dashed line shows the hospital’s maximum capacity. Hospital
ID’s are provided in the corner of each plot.

Fig. 5. The expected percent (left) and count (right) of individuals infected with influenza per block group unable to be hospitalized over the duration of simulated pandemic
in the 100% capacity scenario.
 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285 7

for all hospitals in the system peaks at 87% on day 380 of the havior. This understanding of hospital system capacity could po-
pandemic (Fig. S1). However, in this scenario, 10 of the 12 hos- tentially be integrated with existing location-allocation models to
pitals reached maximum capacity and one nearly reached capac- investigate where new facilities or temporary facilities could be lo-
ity (93.7%). The remaining capacity in this scenario resulted from cated within a study area to most effectively provide care for the
the available supply at the largest hospital in the region, which population in the event of a crisis or mass-casualty event.
could accommodate 1848 infected people in this scenario and only
reached 60.9% capacity at its peak.
5. Conclusion
When maximum hospital capacity for people with influenza in-
fections is set at 75% of the licensed beds in the third scenario,
The findings of this paper are relevant to large-scale disaster
266,932 individuals (66.8% of the infected population) received
planning and pandemic preparedness by simulating a pandemic
care, while 132,693 individuals (33.2% of the infected population)
and analyzing the ability of the health care system to provide
did not receive care. At its peak (on day 368), only 42.45% of that
care during epidemics. Our approach enhances the understanding
day’s newly infected people were able to receive care. Maximum
of surge capacity created by existing capacity models, such as
capacity for all hospitals was exceeded on day 278 of the pan-
FluSurge, by incorporating the location of patients and facilities
demic, and the system remained at maximum capacity until day
into the model. The integration of models from epidemiology,
465 (Figs. S2 and S3). The geographic distribution of people with
geography, and health care services research makes this model a
influenza that were unable to receive care is similar to the first
potentially valuable tool for disaster planning, hospital system
scenario.
evaluation, and pandemic preparedness. Our case study identified
vulnerable geographies within the Triangle region of North Car-
4. Discussion
olina, in which residents may not be able to receive care in the
event of an influenza pandemic, epidemic, or other large-scale
This research project integrated epidemiological and spatial
disaster.
modeling to simulate an infectious disease pandemic and a health
There is an urgent need for models such as this one, as
care delivery model to understand the ability of the health care
demonstrated by the 2017–2018 seasonal influenza epidemic in the
system to respond to the demand for care in the event of a large-
United States (CDC, 2018). Though a seasonal epidemic, not a pan-
scale outbreak. In doing so, our work offers a framework that can
demic, hospitals across the United States exceeded surge capac-
be augmented and expanded to suit the needs of the particular
ity and had to implement alternative care sites for influenza pa-
event and contributes to the evolving discussion of how to in-
tients (Karlamangla, 2017). This is demonstrative of the inability
corporate disease scenario modeling and health care system pre-
of the existing system to provide care for an increased number of
paredness and response models to better understand the effects of
patients associated with seasonally epidemic influenza, suggesting
disaster-level events. We aim to initiate new and better approaches
preparedness for pandemic influenza may be inadequate. Although
that consider not only the aspects of the affected population and
this research focused specifically on the Triangle region of North
the health care systems’ ability to respond, but also how the par-
Carolina, our approach can be adapted for application to a vari-
ticular geography of a region may promote or complicate these ef-
ety of diseases, health care services, and study regions. Further, our
forts.
approach allows for flexible outbreak and capacity parameters, en-
Additionally, this paper has highlighted vulnerable geographies
abling scenario modeling and making it easily adaptable to other
within the Triangle region wherein hospital capacity and the dis-
diseases and hospital systems.
tance from a hospital may result in many infected members popu-
lation not being able to receive care during an influenza pandemic.
For example, in the scenario with maximum capacity set to the Declarations of interest
number of licensed beds, each BG in Franklin County had at least
54% of infected individuals unable to be hospitalized, as there is no None.
hospital in this county and its population had to travel a greater
distance to reach a hospital with available capacity. The results of
the double capacity scenario showed that the ability of the system Funding source
(as a whole) to provide care hinged on a single hospital with re-
maining capacity. Funding for this project was provided by the Tom and Elizabeth
Our approach to construct an integrated disease outbreak and Long Research Award from Honors Carolina. Neither the Tom and
surge capacity model was reliant on several assumptions, and thus Elizabeth Long Research Award nor Honors Carolina played a role
is subject to several limitations. The pandemic simulation model in study design, the collection, analysis and interpretation of data,
did not consider the potential for vaccine development and distri- the writing of the report, or in the decision to submit this article
bution during a pandemic event and also assumed absolutely no for publication.
pre-existing immunity in the population, although other research
has suggested that individuals who have lived through pandemic
CRediT authorship contribution statement
influenza events may have residual immunity against a following
pandemic influenza event (Medina et al., 2010). Additionally, the
Rachel L. Woodul: Conceptualization, Methodology, Formal
allocation model assumed that all infected individuals would at-
analysis, Writing - original draft, Writing - review & editing, Vi-
tempt to seek care at a hospital; however, in an actual pandemic
sualization. Paul L. Delamater: Conceptualization, Methodology,
event, there will likely be alternate care options and infected in-
Writing - review & editing, Visualization, Supervision. Michael
dividuals that do not seek care. Further, the allocation model did
Emch: Writing - review & editing, Visualization, Supervision.
account for factors such as socioeconomic status, driver’s license
possession, car ownership, or other variables that may influence an
individual’s ability to seek care. However, these limitations provide Acknowledgments
opportunities for future research in this field, specifically research
that captures population-specific details such as vaccine availabil- Funding for this project was provided by the Tom and Elizabeth
ity and utilization, residual immunity, and health care access be- Long Research Award from Honors Carolina.
8 R.L. Woodul, P.L. Delamater and M. Emch / Spatial and Spatio-temporal Epidemiology 30 (2019) 100285

Supplementary materials Frost, W.H., 1919. The epidemiology of influenza. J. Am. Med. Assoc. 73 (5), 313–318.
doi:10.1001/jama.1919.02610310 0 070 03.
Frost, W.H., 1920. Statistics of influenza Morbidity: with special reference to certain
Supplementary material associated with this article can be factors in case incidence and case fatality. Public Health Rep. (1896–1970) 35
found, in the online version, at doi:10.1016/j.sste.2019.100285. (11), 584–597. doi:10.2307/4575511.
Frost, W.H., Sydenstricker, E., 1919. Influenza in Maryland: preliminary statistics of
certain localities. Public Health Rep. (1896–1970) 34 (11), 491–504. doi:10.2307/
References 4575056.
Karlamangla, S., (2017). Severe flu brings medicine shortages, packed ERs and a ris-
ing death toll in California. Retrieved April 2, 2018, from http://www.latimes.
Barbisch, D.F., Koenig, K.L., 2006. Understanding surge Capacity: essential elements.
com/local/california/la- me- ln- flu- surge- 20180106-htmlstory.html
Acad. Emerg. Med. 13 (11), 1098–1102. doi:10.1197/j.aem.2006.06.041.
Kumar, S., 2011. Modeling hospital surgical delivery process design using system
Barry, J.M., 2005. The great influenza: the epic story of the deadliest plague in his-
Simulation: optimizing patient flow and bed capacity as an illustration. Technol.
tory. Penguin.
Health Care 19 (1), 1–20.
Brundage, J.F., 2006. Cases and deaths during influenza pandemics in the United
Liu, X., Stechlinski, P., 2017. The switched SIR model. In: Infectious Disease Model-
States. Am. J. Prev. Med. 31 (3), 252–256. doi:10.1016/j.amepre.20 06.04.0 05.
ing. Springer, Cham, pp. 43–82. doi:10.1007/978- 3- 319- 53208- 0_3.
CDC. (2017a). Retrieved November 14, 2017, from https://www.cdc.gov/flu/about/
Lum, M.E., McMillan, A.J., Brook, C.W., Lester, R., Piers, L.S. (2009). Impact of pan-
disease/burden.htm
demic (H1N1) 2009 influenza on critical care capacity in victoria, 191(9), 5.
CDC. (2017b). Retrieved November 14, 2017, from https://www.cdc.gov/flu/about/
Medina, R.A., Manicassamy, B., Stertz, S., Seibert, C.W., Hai, R., Belshe, R.B., García-
disease/spread.htm
Sastre, A., 2010. Pandemic 2009 H1N1 vaccine protects against 1918 Spanish
CDC. (2017c). Retrieved November 14, 2017, from https://www.cdc.gov/flu/
influenza virus. Nat Commun 1 (3), 1–6. doi:10.1038/ncomms1026.
pandemic-resources/basics/past-pandemics.html
Manson, S., Schroeder, J., Van Riper, D., Ruggles, S., 2018. IPUMS National Histori-
CDC. (2017d). Retrieved November 5, 2017, from https://www.cdc.gov/flu/
cal Geographic Information System: Version 13.0 [Database]. University of Min-
pandemic-resources/basics/about.html
nesota, Minneapolis doi:10.18128/D050.V13.0.
CDC. (2018). Retrieved February 10, 2018; May 25, 2018 from https://www.cdc.gov/
Monto, A.S., Comanor, L., Shay, D.K., Thompson, W.W., 2006. Epidemiology of pan-
flu/weekly/summary.htm
demic Influenza: use of surveillance and modeling for pandemic preparedness.
Daugherty, E.L., Carlson, A.L., Perl, T.M., 2010. Planning for the Inevitable: preparing
J. Infect. Dis. 194 (Suppl._2), S92–S97. doi:10.1086/507559.
for epidemic and pandemic respiratory illness in the shadow of H1N1 influenza.
Morens, D.M., Fauci, A.S., 2007. The 1918 influenza Pandemic: insights for the 21st
Clin. Infect. Dis. 50 (8), 1145–1154. doi:10.1086/651272.
century. J. Infect. Dis. 195 (7), 1018–1028. doi:10.1086/511989.
Davis, D.P., Poste, J.C., Hicks, T., Polk, D., Rymer, T.E., Jacoby, I., 2005. Hospital bed
Osterholm, M.T., 2005. Preparing for the next pandemic. N. Engl. J. Med. 352 (18),
surge capacity in the event of a mass-casualty incident. Prehosp. Disaster Med.
1839–1842. doi:10.1056/NEJMp058068.
20 (3), 169–176. doi:10.1017/S1049023X0 0 0 02405.
Sobieraj, J.A., Reyes, J., Dunemn, K.N., Carty, I.H., Pennathur, A., Gutierrez, R.S., Har-
Dugas, A.F., Hsieh, Y.-H., Levin, S.R., Pines, J.M., Mareiniss, D.P., Mohareb, A., Roth-
ris, M.D., 2007. Modeling hospital response to mild and severe influenza pan-
man, R.E., 2012. Google flu trends: correlation with emergency department in-
demic scenarios under normal and expanded capacities. Mil Med. 172 (5), 486–
fluenza rates and crowding metrics. Clin. Infect. Dis. 54 (4), 463–469. doi:10.
490. doi:10.7205/MILMED.172.5.486.
1093/cid/cir883.
Soetaert, K., Petzoldt, T., Setzer, R., 2010. Solving Differential Equations in R: Package
Eastman, A.L., Rinnert, K.J., Nemeth, I.R., Fowler, R.L., Minei, J.P., 2007. Alternate
deSolve. J Statistical Software 33 (9), 1–25. doi:10.18637/jss.v033.i09.
site surge capacity in times of public health disaster maintains trauma center
Yi, P., George, S.K., Paul, J.A., Lin, L., 2010. Hospital capacity planning for disaster
and emergency department Integrity: hurricane Katrina. J. Trauma: Injury In-
emergency management. Socioecon. Plann. Sci. 44 (3), 151–160. doi:10.1016/j.
fect. Crit. Care 63 (2), 253–257. doi:10.1097/TA.0b013e3180d0a70e.
seps.20 09.11.0 02.
Eyck, R.P.T., 2008. Ability of regional hospitals to meet projected avian flu pan-
demic surge capacity requirements. Prehosp. Disaster Med. 23 (2), 103–112.
doi:10.1017/S1049023X0 0 0 05707.