ArticlePDF Available

Annual economic impacts of seasonal influenza on US counties: Spatial heterogeneity and patterns

Authors:

Abstract and Figures

Economic impacts of seasonal influenza vary across US counties, but little estimation has been conducted at the county level. This research computed annual economic costs of seasonal influenza for 3143 US counties based on Census 2010, identified inherent spatial patterns, and investigated cost-benefits of vaccination strategies. The computing model modified existing methods for national level estimation, and further emphasized spatial variations between counties, in terms of population size, age structure, influenza activity, and income level. Upon such a model, four vaccination strategies that prioritize different types of counties were simulated and their net returns were examined. The results indicate that the annual economic costs of influenza varied from 13.9thousandto13.9 thousand to 957.5 million across US counties, with a median of $2.47 million. Prioritizing vaccines to counties with high influenza attack rates produces the lowest influenza cases and highest net returns. This research fills the current knowledge gap by downscaling the estimation to a county level, and adds spatial variability into studies of influenza economics and interventions. Compared to the national estimates, the presented statistics and maps will offer detailed guidance for local health agencies to fight against influenza.
Content may be subject to copyright.
R E S E A R C H Open Access
Annual economic impacts of seasonal influenza
on US counties: Spatial heterogeneity and
patterns
Liang Mao
1*
, Yang Yang
1
, Youliang Qiu
1
and Yan Yang
2
Abstract
Economic impacts of seasonal influenza vary across US counties, but little estimation has been conducted at the
county level. This research computed annual economic costs of seasonal influenza for 3143 US counties based on
Census 2010, identified inherent spatial patterns, and investigated cost-benefits of vaccination strategies. The
computing model modified existing methods for national level estimation, and further emphasized spatial
variations between counties, in terms of population size, age structure, influenza activity, and income level. Upon
such a model, four vaccination strategies that prioritize different types of counties were simulated and their net
returns were examined. The results indicate that the annual economic costs of influenza varied from $13.9
thousand to $957.5 million across US counties, with a median of $2.47 million. Prioritizing vaccines to counties with
high influenza attack rates produces the lowest influenza cases and highest net returns. This research fills the
current knowledge gap by downscaling the estimation to a county level, and adds spatial variability into studies of
influenza economics and interventions. Compared to the national estimates, the presented statistics and maps will
offer detailed guidance for local health agencies to fight against influenza.
Keywords: Influenza, Economic costs, US Counties, Vaccination, Spatial heterogeneity
Introduction
Every year in the US, influenza viruses pose remarkable
impacts on socio-economy, such as costs of medical care,
loss of productivity, and deaths [1]. Since economic con-
siderations are essential for influenza control, decision
makers often need to examine following questions for
health interventions. How much will an influenza season
cost the US? Which states or counties bear high costs?
Where to distribute vaccines to achieve the maximum
returns? To date, only a small number of studies have esti-
mated the economic impacts of influenza in the US. The
Office of Technology Assessment reported that the influ-
enza accounts for $1 ~ 3 billion per year in medical costs
[2]. Meltzer, et al. argued that the annual economic bur-
den of pandemic influenza could range from $71.3 ~ 166.5
billion [3]. The latest estimation by Molinari et al. indi-
cated that the short-term costs and long-term burden of
seasonal influenza can be amounted to $26.8 ~ $87.1
billion a year [4]. These studies have established systematic
methods to analyze influenza economics and offered valu-
able guidance for interventions.
Previous studies, however, have focused on the national-
level estimation, while few have drilled down to a county
level and taken into account spatial heterogeneity between
counties. Many factors were assumed to be homogenous
across counties but in fact vary remarkably, such as the in-
fluenza activity, population size, age structure, income
level, and so on. Current estimates for the entire US fail to
differentiate influenza impacts between counties, and thus
offer little information for state/county-level health plan-
ning. For instance, the national estimates cannot inform
the design of county-based vaccination strategies, i.e.,
where (or which counties) should receive vaccines first for
a best cost-effectiveness. In addition, the lack of county-
level knowledge may cloud the identification of contribut-
ing factors to influenza costs, due to the modifiable areal
unit problem (MAUP). That is, different levels of aggrega-
tion, such as the county-, state-, and national levels, may
produce variation in statistical associations [5]. Although
* Correspondence: liangmao@ufl.edu
1
Department of Geography, University of Florida, Gainesville, FL 32611, USA
Full list of author information is available at the end of the article
INTERNATIONAL JOURNAL
OF HEALTH GEOGRAPHICS
© 2012 Mao et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.
Mao et al. International Journal of Health Geographics 2012, 11:16
http://www.ij-healthgeographics.com/content/11/1/16
the Centers for Disease Control and Prevention (CDC)
have offered a FluAid software to help estimate local eco-
nomic impacts [6], this tool still employs nationwide
homogeneous parameters, and cannot characterize any
inter-county variations except for demographics.
This article aims to estimate annual economic impact of
seasonal influenza for 3143 US counties based on US Cen-
sus 2010, characterize the inter-county variation of such
impact, and investigate cost-effectiveness of vaccination
strategies. The estimation modified existing methods in the
literature, and further emphasized spatial variations among
counties, in terms of population size, age structure, influ-
enza activity, and income level. Spatial and statistical ana-
lyses were conducted to identify spatial patterns of
influenza impacts across counties. Futhermore, four
county-based vaccination strategies were simulated and
their cost-effectiveness was compared to identify the
optimal.
Materials and methods
For a specific county, the economic impacts of influenza
are the sum of monetary costs to each influenza case in
this county (Equation 1).
Total economic impactsi¼X
n
j
Montery costsi;j
¼Pn
jDirect Costsi;jþIndirect Costsi;j

¼Pn
jMedical Costsi;jþLoss of Productivityi;j

ð1Þ
(where irepresents a county, ndenotes all influenza
cases in county i, and jindicates an influenza case in
county i).
The monetary costs of an influenza case can be further
divided into direct and indirect costs. The direct costs re-
sult from expenses of healthcare resources, e.g.,
hospitalization and antiviral treatment, while the indirect
costs come from the loss of productivity from school/work
absenteeism and death [3,7]. According to Equation 1, fol-
lowing three sub-sections describe the datasets and meth-
ods to estimate the county-level influenza cases, direct and
indirect costs, respectively.
County-level influenza cases and health outcomes
County population by age group
The population of a county is a basis for calculating the
number of influenza cases. Since influenza infections vary
by age, the county population was divided into five age
groups: under 5, 517, 1849, 5064, and 65 years. The
county population by age group was extracted from the US
Census 2010 Summary File 1 [8], the most recent demo-
graphic data available. For mapping purposes, the county
population was further geo-referenced to its administrative
boundary from the US Census Topologically Integrated
Geographic Encoding and Referencing system (TIGER)
Products [9].
County influenza attack rate by age group
For each age group in a county, the number of influenza
cases is a product of the age-group population and the age-
specific attack rate. Since no data has been published on
the influenza attack rates by county, three pieces of infor-
mation were used for estimation: the national attack rates
by age group, the national Influenza Like Illness (ILI) rates
(ILI visits per 100,000), and the ILI rates of major US cities.
As shown in Equation 2, the influenza attack rate in county
iat age group g
^
Attack ratei;g

is equal to the national at-
tack rate at the age group Nation attack rateg

gadjusted
by a ratio between the county ILI rate
^
CountyILIi

and
the national ILI rate (National ILI). Simiarly, the standard
devation of
^
Attack ratei;gis the adjusted standard devation
of Nation attack rateg.
^
Attack ratei;g¼
^
CountyILIi
NationILI Nation attack rateg
Std dev
^
Attack ratei;g

¼
^
CountyILIi
NationILI
Std dev Nation attack rateg

8
>
>
>
>
<
>
>
>
>
:ð2Þ
The national attack rates by age group Nation attack rate g

and the associated standard deviations were adopted
from the surveillance data and established literature [4],
with details listed in Supplementary file 1: Table S1. The
national ILI rate NationILIðÞwas the average of weekly
naitonal ILI rates from 2003 to 2010, published by the
Google Flu Trends [10]. To gain a representative ILI
rate for seasonal influenza, the data of Season 200910
was eliminated before averaging, due to the H1N1 flu
pandemic. The Flu Trends data could be a reasonable
proxy to influenza activities because recent literature
has reported its capability of predicting influenza activ-
ity [11,12]. The correlation between the data from the
Google Flu Trends and CDC Virus Surveillance can
achieve up to 82% [13].
The county specific ILI rate
^
CountyILIi(inEquation2)
was interpolated using the ILI rates of major cities near to
the county centroid. The ILI rates of 117 major US cities
were also averaged from the weekly ILI rates from Google
Flu Trends. Similar to the national ILI rate, the data of Sea-
son 200910 was removed before averaging. The ordinary
kriging, a sophisticated geostatistical method, was employed
for interpolation, because it has been commonly used for
predicting ILI rates [14,15]. In this research, a major advan-
tage of kriging lies in its ability to minimize the standard
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 2 of 8
http://www.ij-healthgeographics.com/content/11/1/16
error of an estimate and quantify this standard error
[16,17]. The kriging interploation was accomplished by four
steps. First, the 117 major cities were georeferenced to their
geographic locations. Their ILI rates were used to compute
semivariances between cities at different separations, and
consitute a sample semivariogram (squares in Figure 1a).
Second, a spherical model was fitted to the sample semivar-
iogram (curves in Figure 1a), using a least square method.
The sill, nugget, and major range of the spherical model
were estimated to be 34.21 km, 46,692, and 3405.5 km re-
spectively. Third, the searching neighbor was set to the
nearest 12 to 20 cities around the centriod of each county.
The fitted spehrical model was used to assign weights to
neigbhor cities, and the kriged county ILI was the weighted
sum of ILI rates at neighbor cities. Lastly, a cross-validation
was conducted and gave a mean error of 0.05 (close to
zero) and an average standard error of 310.1 (Figure 1b).
XThe kriged county ILI rates (Figure 2a) and asso-
ciated standard errors (Figure 2b) were used together to
simulate the
^
CountyILIi. Based upon the estimates of
Nation attack rateg

,NationILIðÞ,and
^
CountyILIi,the
county attack rates
^
Attack ratei;gwere cacluated for each
age group, and then multiplied by the age-group popula-
tions to obtain the number of influenza cases by age
group.
Risks and health outcomes by age group
Influenza cases may progress to different outcomes,
which costs distinctly. To refine the cost estimation, this
research classified influenza cases of each age group into
two types of risks, and then four health outcomes. The
two types of risks refer to non-high and high risks
of developing serious complications. An influenza case
was defined to be high risk if one or more medical condi-
tions were consistent with high-risk conditions identified
by the Advisory Committee on Immunization Practices
(ACIP), thus more likely to develop severe outcomes [18].
For each age-risk group, influenza cases were further sepa-
rated into four health outcomes, including: self-care (not
medically attended), outpatient visit, hospitalization, and
death. The likelihoods of becoming a high-risk case and the
probabilities of developing each health outcome were esti-
mated by Molinari et al. [4] based on the literature, medical
records, and reports. Mean values and statistical distribu-
tions of these parameters are shown in Additional file 1:
Table S1 by age and by risk group. Finally, each influenza
case in a county was assigned one of 40 categories (5 age
groups × 2 types of risks × 4 health outcomes). Each cat-
egory was associated with a direct cost and an indirect cost
discussed below.
Direct costs by county
The direct costs come from the medical expenditure in re-
sponse to influenza (e.g., hospitalizations, outpatient visits,
and drug purchases), and vary over the 40 age-risk-
outcome categories. Molinari et al. [4] had estimated the
national average medical cost (and distribution) for each of
the 40 categories according to a proprietary database that
contains health insurance claims data from 4 million
insured persons [19]. Details about the costs per category
are given in Additional file 1: Table S2, and these estimates
were used to parameterize the county model after an infla-
tion adjustment. Since the work of Molinari et al. was done
in 2003, this research inflated the medical costs from 2003
to 2010 (the year of census data) based on the consumer
Figure 1 The kriging model to estimate county ILI rates: (a) Sample semivariogram of city ILI rates (red squares) and the fitted
spherical model (blue curves): 46692*Nugget+342120*Spherical(3405500, 1599800, 328.8°); (b) Scatterplot of reported ILI rates vs.
kriging predicted ILI rates.The blue straightline summarizes the trend as: y=0.698x+465.73 (Figures were created with ESRI ArcGIS 10.0).
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 3 of 8
http://www.ij-healthgeographics.com/content/11/1/16
price index (CPI) of these two years. The CPI ratio between
year 2010 and 2003 was set to 1.185 as reported by the Bur-
eau of Labor Statistics [20]. Therefore, this research esti-
mated the economic impacts if a typical seasonal influenza
hit US in 2010 or years later on. By simulating the inflated
direct costs for each case, the direct costs of influenza to a
county were the sum of costs from all cases.
Indirect costs by county
The indirect costs include the loss of productivity due to
work/school absenteeism and death. The loss of product-
ivity due to work/school absenteeism was calculated by
multiplying the length of days absent from work by the
monetary value lost per day [21]. The national average
length of work/school absenteeism, along with its distribu-
tion, had been previously estimated by Molinari et al. [4]
for the 40 age-risk-outcome categories (Additional file 1:
Table S2). This research assumed that the length of absen-
teeism in a county follows the nationwide distribution,
while the monetary value lost per day varies between
counties. For each county, the monetary value lost per day
(Figure 2c) was equal to the average wage per job [22]
divided by the total working days per year (260 days).
For influenza cases ending with deaths, the productivity
loss was estimated as the present value of lost earnings
(PVLE), the projected earnings until retirement based on
apersons current salary [23]. The national average of
PVLE had been previously extrapolated by age group in
the work of Molinari et al. [4], and were inflated to 2010
in this research (Additional file 1: Table S2). In such a
way, the indirect cost of each influenza case was evaluated
and the sum of all cases gave the indirect costs of influenza
in a county.
Total economic costs by county
Upon the evaluation of influenza case number, associated
direct and indirect costs, the total economic costs for each
county were valued using Equation 1. To quantify the un-
certainties in estimation, this research employed an
individual-based stochastic approach. Each individual in a
county is a discrete modeling unit with properties, such as
the age group, attack rate, health outcome, direct and indir-
ect costs. Using Monte-Carlo simulation, these properties
were assigned random values from their adjusted probability
distributions according to (Additional file 1: Table S1 and
S2) and kriging estimates. The simulation was run by 1,000
realizations to establish 95% confidence intervals for esti-
mates of interest. Averaged from 1,000 realizations, the total
economic costs for each county were mapped in Figure 3a,
and the economic costs per capita were also computed by
prorating the total costs to the county population
(Figure 3b). Due to the word limits, the county estimates
and associated 95% confidence intervals were presented in
Additional file 2 and published on an interactive online
(a) Kriged influenza like illness (ILI) visits (b) Standard errors of Kriged ILI visits
(c) Daily average wage (d) Percentage of vulnerable population
Figure 2 Spatial heterogeneity between US counties, in terms of (a) Kriging estimated influenza like illness (ILI) visits (per 100,000
persons), (b) Standard errors of ILI estimation, (c)Daily average wage as an indicator of income level, (d)Percentage of vulnerable
population (age <5 and >50 years).
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 4 of 8
http://www.ij-healthgeographics.com/content/11/1/16
mapping system (http://heep.geog.ufl.edu/flucost). To cap-
ture the spatial clusters of economic impacts, a MoransI
statistic was applied the county cost map.
Net returns of vaccination against influenza
Vaccination is widely suggested to be a major strategy
for reducing the impacts of influenza. The use of vaccine
was assumed to avert outcomes of influenza cases, and
thus led to savings of costs that need to treat these out-
comes. The net return from vaccination is an important
economic measure in cost-benefit analysis of such inter-
vention. Following the method of Meltzer, et al. [3], the
total net return was calculated by Equation 3:
Total Net Return ¼PNet Returnage;risk group
¼X
ð
Savings from outcomes
averted in populationage:risk group
Vaccinated popoulationage;risk group
Cost of vaccine per person
Þ
ð3Þ
The savings from averted outcomes in a population
was determined by the effectiveness of vaccines
(reported in Additional file 1: Table S3). To consider the
match between vaccine and influenza virus, a probability
of good match 80% [24] was incorporated into the simu-
lation model. The vaccine effectiveness of a good match
was assumed to be twice as high as that of a poor match.
The vaccinated population was a function of vaccination
coverage, i.e., the proportion of people being inoculated.
The cost of vaccine per person was estimated to be $21
by Meltzer et al. in 1999 [3] and inflated to $27.50 in
2010. This cost includes the vaccine price, its distribu-
tion and administration fees (health-care worker time,
supplies), patient travel, time lost from work and other
activities; and cost of side effects.
By explicitly considering the county differences, this
research was capable of investigating four county-based
vaccination strategies. The first strategy, referred to as
the random strategy, randomly vaccinated US popula-
tion to a predefined coverage (in percentage), regardless
of the age group and the county they live in. The second
strategy vaccinated the same amount of people, but
prioritized those who live in the counties of high influ-
enza attack rates, thus called High-Attack-Firststrategy.
The third strategy is similar to the second, but prioritiz-
ing those who live in the counties with high proportion
of vulnerable population, named as a High-
Vulnerability-Firststrategy. According to the recommen-
dation by CDC [25], the population under 5 years and
over 50 years was defined as the most vulnerable popula-
tion (Figure 2d). The last strategy aims to vaccinate
people who live in the counties with high income level,
and thus called High-Income-Firststrategy. Of the four
strategies, each was simulated based on Equation 3 at a
range of vaccination coverage from 10% to 90% with a
10% increment. The average influenza case number and
the net return ($) were estimated for each strategy-
coverage combination after 1,000 simulation runs
(Figure 4). The four strategies were compared to identify
the optimal one that produces the fewest influenza cases
and highest net return.
Results and discussion
At the national level, the seasonal influenza resulted in
25.34 million cases a year (95%CI: 24.83 25.86 mil-
lion), 8.1% of the total population in 2010 (Table 1).
The annual economic costs were estimated to be
$29.12 billion (95% CI: $28.44 $29.87 billion), ap-
proximately 0.2% of the gross domestic product of US
in 2010. About 65% of the economic costs comes from
the indirect cost, i.e., loss of productivity due to work
absenteeism and death, while the rest of 35% is from
the direct medical cost.
The economic impacts of influenza varied dramatically
across US counties (Table 2). Kalawao County, Hawaii,
had the smallest number of influenza cases (7 cases, 95%
(a) Economic costs by county (b) Economic cost per capita by county
Figure 3 Spatial heterogeneity between US counties, in terms of (a) Total economic costs of influenza, and (b) Economic cost per
capita. The detailed county estimates and associated 95% confidence intervals were available at http://heep.geog.ulf.edu/flucost and Additional
file 2.
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 5 of 8
http://www.ij-healthgeographics.com/content/11/1/16
CI: 213) and the lowest economic costs ($13,883, 95%
CI: $281$97,037), because its population was only 90 by
2010. On the other hand, Los Angeles County, California,
is the most populous county in US (9,818,605 persons in
2010). Not surprisingly, this county possessed the largest
number of cases (868,587, 95% CI: 610,091 1,118,763)
and the highest economic costs ($957.5 million, 95%CI:
$664.71248.2 million). Due to the wide variability among
counties, the median value would be a more reasonable
statistic to indicate the centrality than the mean value. On
average, a US county would have 2009 influenza cases per
year based on US census 2010. The economic costs for a
county averaged $2.47 million, and a quarter of counties
(above the 75% percentile) may experience an economic
loss greater than $6.42 million.
The geographic distribution of economic costs is of
interest (Figure 3a). Note that each color shade represents
differences orders of magnitude. The distribution of eco-
nomic costs strongly coincides with the population distri-
bution, in that a large population often has more
influenza cases and requires more money to alleviate the
disease. In general, counties with high costs were concen-
trated in the coastal areas, such as the Pacific region
(Washington, Oregon, California) and the Middle Atlantic
region (New York, Pennsylvania, Connecticut, New
Jersey). The inland areas in the Mid-West and Mountain
regions, such as Montana, Idaho, Minnesota, and
Oklahoma, had relatively lower costs. The MoransI,
indicated that the county economic costs were signifi-
cantly clustered over space (MoransI=0.28, Zscore =
27.95, p-value< 0.05). Two spatial clusters of high costs
can be easily identified: one was centered at Los Angeles
County, California along the West Coast, while the other
encompassed the Boston-New York City metropolitan
areas at the East Coast. Other scattered metropolitan
areas also showed high levels of costs, such as Miami
(Florida), Seattle (Washington), and Houston (Texas), etc.
It is also interesting to examine what if the economic
costs were prorated to every person in a county
(Figure 3b). The economic cost per capita by county
exhibited a distinct spatial pattern. The annual cost per
capita ranged from $32.5 (Ziebach County, South Dakota)
to $272.4 (Llano County, Texas). One spatial cluster of
high cost per capita was the South Central region (Texas,
Oklahoma, Arkansas, and Louisiana), where every resident
needed to spend more than $150 to combat influenza.
The other cluster can be found in the mid-Atlantic and
south Atlantic regions, with a personal cost between
$100 ~ 150 per capita. Residents in the Mountain region
bore the lightest burden (under $50 per capita) except for
southern counties in Arizona and New Mexico. This
spatial pattern can be explained by the fact that the eco-
nomic cost per capita is independent of county population
size. Counties with high cost per capita were associated
with high levels of influenza attack rates. A person in a
high attack-rate county is more likely to be infected and
develop high-risk complications, and thus costs more than
a person in a low attack rate county.
With regarding to the vaccination strategies, the
High-Attack-Firststrategy significantly outperforms any
other strategy due to the lowest influenza cases and
highest net return (Figure 4). In other words, it would be
an optimal strategy to first vaccinating people living in
the West-South-Central region, including Texas, Okla-
homa, Arkansas, and New Mexico (dark red regions in
Figure 2a). A possible reason is that this strategy directly
prohibited the transmission of influenza, while other
strategies address influenza indirectly through popula-
tion vulnerability and income level. The effect of High-
Table 1 Estimated annual influenza impacts on the entire
US using population and costs of 2010
Total Lower 95%CI Upper 95%CI
Number of cases
(Million)
25.34 24.83 25.86
Direct costs
(Millions $)
10,262.98 10,046.60 10,482.80
Indirect costs
(Millions $)
18,853.66 18,321.81 19,439.80
Total economic costs
(Millions $)
29,116.65 28,441.24 29,870.84
Figure 4 Comparison of cost-benefits between four county-based strategies: (a) Total influenza cases and (b) Net returns ($) as a
function of vaccination coverage from 10% to 90% of the population.
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 6 of 8
http://www.ij-healthgeographics.com/content/11/1/16
Attack-Firstvaccination is, thus, more straightforward
than other strategies. The random, High-Vulnerability-
Firstand High-Income-Firststrategies reduce the
influenza cases to a similar extent, but the High-
Vulnerability-Firststrategy returns more monetary bene-
fits, and is therefore the second cost-effective strategy.
This is because the High-Vulnerability-Firststrategy
greatly decreases the people who may otherwise develop
severe outcomes, such as hospitalizations and death, and
thus curtails the major source of influenza costs. In this
sense, it is also a wise strategy to prioritize vaccines to the
North West region and North Midwest region (including
Idaho, Wyoming, North Dakota, South Dakota, Minne-
sota and Wisconsin), where the proportion of vulnerable
population is pretty high (dark red regions in Figure 2c).
There are several issues that require special attention as
aresultofthisstudy.First,although many parameters were
estimated for individual counties, a few parameters were
still assumed to be nationwide homogeneous, such as the
length of absenteeism and the present value of lost
earnings. The major reason is the lack of relevant data for
each of US counties. These homogenous parameters may
reduce the variability of costs between counties and lead to
a smoother map than the reality. This may also mislead
our understandings on counties with extremely low and
high costs. Second, the use of Google Flu Trends data
would introduce biases into the estimation, even though
its correlation to the viral surveillance can reach 82%. A
finer scale influenza ILI data at the county level would be
helpful to improve estimation. The authors had tried to
directly obtain county flu infection data, instead of using
Google flu data. Unfortunately, among the 51 US states,
only 6 have explicitly published their county-level flu attack
rates. The US CDC only releases regional flu data, with
each region covering several states. Therefore, the Google
fludatacouldbethebestdatathatispubliclyavailableand
contains sufficient details. Third, the influenza attack rates,
direct and indirect costs were assumed to be homogeneous
within a county, regardless of urban and rural areas. But in
reality, these parameters may vary between urban and rural
areas because of different population density, land use pat-
terns, accessibility to healthcare, etc. It remains unclear
whether or not the modeling of urbanrural difference
wouldsignificantlychangethecurrent estimation, but this
question warrants a future study.
Conclusions
The contributions of this research have two folds. First,
the current estimation of influenza impacts uses nation-
wide homogenous parameters, which flatten the spatial
variations among US counties. This research is the first
attempt to calculate and map the county-level economic
impacts of seasonal influenza in the US, thereby filling
the current gap that only national estimates are avail-
able. The economic costs of influenza range from $13.9
thousand to $957.5 million among US counties, varying
over demographics, economy, and epidemics. There are
two spatial clusters of high costs: one centered at Los
Angeles County, California along the West Coast, while
the other embracing the Boston-New York City metro-
politan areas on the East Coast. Secondly, before this re-
search, most studies investigate vaccination strategies
only at a national level, but few have considered the
county differences, i.e., distributing vaccine resources
based on county characteristics. This research has
explored four county-based strategies and suggested that
vaccination prioritizing counties with high attack rates
would produce the greatest cost-benefits. This research
adds a county/spatial perspective into the design of
health interventions, and sheds insight on new cost-
effective health policies.
It is argued that any estimation model can only produce
crude approximations to reality. The key of this research
is not to look for the absolute numeric predictions, but for
differences in outcomes between different counties or be-
tween different scenarios. In this sense, this research lays a
Table 2 Annual influenza impacts on US counties using population and costs of 2010
Minimum Maximum Mean Median Standard
deviation
25%
percentile
75%
percentile
Number of
cases
7 868,588 8,061 2009 27,950 829 5530
Direct costs
(Millions $)
0.004 327.11 3.26 0.89 10.56 0.38 2.35
Indirect costs
(Millions $)
0.008 630.38 6.00 1.57 20.17 0.65 4.10
Total costs
(Millions $)
0.014 957.49 9.26 2.47 30.72 1.02 6.42
Costs/Capita ($) 32.51 272.36 97.50 89.27 32.77 75.84 113.16
*Statistics and 95% confidence intervals by county are available at: http://heep.geog.u fl.edu/flucost/.
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 7 of 8
http://www.ij-healthgeographics.com/content/11/1/16
foundation for county-level study of influenza economics
and interventions, and can be easily expanded to other in-
fectious diseases. The numerical estimates and maps pre-
sented here not only inform general health planning for
the entire US, but also offer detailed guidance for state or
county level interventions to fight future influenza
outbreaks.
Additional files
Additional file 1: Supplementary file 1. Annual Economic Impacts of
Seasonal Influenza and Vaccination on US Counties: Spatial Heterogeneity
and Patterns [3,4].
Additional file 2: Supplementary file 2. Annual Economic Impacts of
Seasonal Influenza and Vaccination on US Counties: Spatial Heterogeneity
and Patterns.
Competing interests
The authors declare that they have no competing interests.
Authorscontributions
LM designed the work, performed all coding and simulation, and drafted the
manuscript. YY carried out part of data collection and analyses. YLQ and YY
established the GIS website. All authors read and approved the final
manuscript.
Authorsinformation
LM and YLQ are Assistant professors of Geography in the University of
Florida. YY is a PhD candiate of Geography in the University of Florida. The
other YY is a PhD candiate of Geography in the University at Buffalo, State
University of New York.
Acknowledgements
The authors are thankful for the valuable comments from the editor and two
reviewers. Publication of this article was partially funded by the University of
Florida Open-Access Publishing Fund.
Author details
1
Department of Geography, University of Florida, Gainesville, FL 32611, USA.
2
Department of Geography, University at Buffalo, State University of New
York, Amherst, NY 14261, USA.
Received: 17 January 2012 Accepted: 19 April 2012
Published: 17 May 2012
References
1. Schoenbaum SC: Economic impact of influenza: The individual's
perspective. Am J Med 1987, 82(6):2630.
2. US Congress: Office of Technology Assessment: Cost-effectivness of Influenza
Vaccination. Washington, DC: GPO; 1981.
3. Meltzer MI, Cox NJ, Fukuda K: The economic impact of pandemic
influenza in the United States: priorities for intervention. Emerg Infect Dis
1999, 5(5):659671.
4. Molinari NAM, Ortega-Sanchez IR, Messonnier ML, Thompson WW, Wortley PM,
Weintraub E, Bridges CB: The annual impact of seasonal influenza in the US:
measuring disease burden and costs. Vaccine 2007, 25(27):50865096.
5. Fotheringham AS, Wong DWS: The modifiable areal unit problem in
multivariate statistical analysis. Environ Plann A 1991, 23(7):10251044.
6. Centers for Disease Control and Prevention: FluAid software. In Center for
Diseae Control and Prevention; 2000. http://www.cdc.gov/flu/tools/fluaid/.
7. Szucs T: The socio-economic burden of influenza. J Antimicrob Chemother
1999, 44(suppl 2):1115.
8. US Census: Census 2010 Summary File 1 (SF 1) 100-Percent Data. US
Census Bureau; 2010.
9. US Census: TIGER
W
, TIGER/Line
W
and TIGER
W
-Related Products. US
Census Bureau; 2010.
10. Google, Inc: Google Flu Trend - United States: Google 2011. http://www.
google.org/flutrends/.
11. Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, Brilliant L:
Detecting influenza epidemics using search engine query data. Nature
2008, 457(7232):10121014.
12. Carneiro HA, Mylonakis E: Google trends: a web-based tool for real-time
surveillance of disease outbreaks. Clin Infect Dis 2009, 49(10):15571564.
13. Ortiz J, Zhou H, Shay D, Neuzil K, Goss C: Does Google Influenza Tracking
Correlate With Laboratory Tests Positive For Influenza. Am J Respir Crit
Care Med 2010, 181:A2626.
14. Carrat F, Valleron AJ: Epidemiologic mapping using the krigingmethod:
application to an influenza-like epidemic in France. Am J Epidemiol 1992,
135(11):12931300.
15. Sakai T, Suzuki H, Sasaki A, Saito R, Tanabe N, Taniguchi K: Geographic and
temporal trends in influenzalike illness, Japan, 19921999. Emerg Infect
Dis 2004, 10(10):18221826.
16. Oliver MA, Webster R: Kriging: a method of interpolation for geographical
information systems. Int J Geogr Inform Syst 1990, 4(3):313332.
17. Isaaks EH, Srivastava RM: Applied geostatistics, Volume 2. New York: Oxford
University Press; 1989.
18. Fiore AE, Shay DK, Haber P, Iskander J, Uyeki T, Mootrey G, Bresee JS, Cox
NJ: Prevention and control of influenza: Recommendations of the
Advisory Committee on Immunization Practices (ACIP). Morbidity Mortality
Wkly Rep 2007, 56:154.
19. MarketScan: Medstat MarketScan: Thomson Reuters; 2003.
20. US Bureau of Labor Statistics: Consumer Price Index Bureau of Labor
Statistics; 2010.
21. Akazawa M, Sindelar JL, Paltiel AD: Economic Costs of Influenza-Related
Work Absenteeism. Value Health 2003, 6(2):107115.
22. US Bureau of Economic Analysis (BEA): Local Area Personal Income
(Average wage per job). U.S. Department of Commerce 2009.
23. Haddix AC, Teutsch SM, Corso PS: Prevention effectiveness: a guide to decision
analysis and economic evaluation. USA: Oxford University Press 2003.
24. Nichol K, Margolis K, Wuorenma J, Von Sternberg T: The efficacy and cost
effectiveness of vaccination against influenza among elderly persons
living in the community. New Engl J Med 1994, 331(12):778784.
25. Centers for Disease Control and Prevention: Key Facts about Influenza
(Flu) & Flu Vaccine. [http://www.cdc.gov/flu/keyfacts.htm].
doi:10.1186/1476-072X-11-16
Cite this article as: Mao et al.:Annual economic impacts of seasonal
influenza on US counties: Spatial heterogeneity and patterns.
International Journal of Health Geographics 2012 11:16.
Submit your next manuscript to BioMed Central
and take full advantage of:
Convenient online submission
Thorough peer review
No space constraints or color figure charges
Immediate publication on acceptance
Inclusion in PubMed, CAS, Scopus and Google Scholar
Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit
Mao et al. International Journal of Health Geographics 2012, 11:16 Page 8 of 8
http://www.ij-healthgeographics.com/content/11/1/16
... Apart from the high mortality rate, influenza also provokes financial damages worldwide. Indicatively, the total financial loss of annual influenza epidemics in the United States is approximately $87.1 billion (5). ...
Article
Infectious disease models are regularly used by epidemiologists to predict the course of diseases. However, the accuracy of such models remains hotly debated. In this study, we developed an algorithm to estimate the contact rate and the average infectious period of influenza using a Susceptible, Infected, and Recovered (SIR) epidemic model. The parameters in this model were estimated using data on infected Greek individuals collected from the National Public Health Organization. The data interval investigated spans 13 weeks of 2019 and the first 8 weeks of 2020. After estimating the nonlinear SIR epidemic model’s parameter values, we computed the theoretical influenza reproduction number and simulated the evolution of the three SIR variables over time. We hypothesized that the reproduction number would be greater than one, which means that influenza is not like other simple diseases but is an epidemic. Our model labeled influenza as an epidemic with a basic reproduction value greater than one.
... Furthermore, empirical evidence on seasonal epidemics such as influenza, Dengue virus (DENV), Zika virus (ZIKV), Ebola virus (EBOV), have also a considerable economic impact in the affected areas. In the United States (US), the annual economic costs of influenza vary from US$13,900 to US$957.5 million across US counties, with a median of US$2.47 million (Liang, Yang, Youliang, & Yan, 2012). In the case of Dengue fever, the overall annual cost in 2013 amounted approximately up to US$8.9 billion spread across 141 countries (Shepard, Undurraga, Halasa, & Stanaway, 2016). ...
Article
Full-text available
This paper establishes conceptual foundations for analyzing the economic dimensions of regional or global emerging and endemic infectious disease events, such as the case of Covid-19. The Covid-19 Global Economic Impact Simulator attempts to identify the Covid-19 transmission parameters and forecast its trajectories. The model introduces seven basic indicators - (i) the Covid-19 contagious spread intensity rate (S.I.), (ii) the treatment level for Covid-19 infected cases rate (T); (iii) the number of Covid-19 causalities rate (-C); (iv) the economic wear from the Covid-19 epidemic rate (-Π); (v) the Covid-19 contagious cases multiplier rate (M); (vi) the total economic leaking from the Covid-19 epidemic rate (-Ltotal); and (vii) the economic desgrowth from the Covid-19 epidemic rate (-δ2019-nCoV). Findings show that Covid-19 exhibits parallel spatial and temporal conditions with the related R.N.A. virus family but carries distinct infection signatures and magnitude of virus replication. Covid-19 ferocity can trigger a severe public health emergency in China with significant impacts on the domestic and world economies.
... Many avian and mammalian species such as chicken, duck, quail, and crow, and human, swine, horse, and cat, respectively, might be infected with IAV (Taubenberger and Kash, 2010;Hussain et al., 2017). Every year, millions of people are infected with influenza worldwide, with increasing economic losses due to diagnosis, treatment, and vaccine development costs (Molinari et al., 2007;Mao et al., 2012;Palekar et al., 2019). Likewise, some strains of the IAV with a high morbidity and mortality rate severely impact the poultry industries (Thomas and Noppenberger, 2007;Yoo et al., 2018). ...
Article
Full-text available
The subtype prevalence, drug resistance- and pathogenicity-associated mutations, and the distribution of the influenza A virus (IAV) isolates identified in Bangladesh from 2002 to 2019 were analyzed using bioinformatic tools. A total of 30 IAV subtypes have been identified in humans (4), avian species (29), and environment (5) in Bangladesh. The predominant subtypes in human and avian species are H1N1/H3N2 and H5N1/H9N2, respectively. However, the subtypes H5N1/H9N2 infecting humans and H3N2/H1N1 infecting avian species have also been identified. Among the avian species, the maximum number of subtypes (27) have been identified in ducks. A 3.56% of the isolates showed neuraminidase inhibitor (NAI) resistance with a prevalence of 8.50, 1.33, and 2.67% in avian species, humans, and the environment, respectively, the following mutations were detected: V116A, I117V, D198N, I223R, S247N, H275Y, and N295S. Prevalence of adamantane-resistant IAVs was 100, 50, and 30.54% in humans, the environment, and avian species, respectively, the subtypes H3N2, H1N1, H9N2, and H5N2 were highly prevalent, with the subtype H5N1 showing a comparatively lower prevalence. Important PB2 mutations such D9N, K526R, A588V, A588I, G590S, Q591R, E627K, K702R, and S714R were identified. A wide range of IAV subtypes have been identified in Bangladesh with a diversified genetic variation in the NA, M2, and PB2 proteins providing drug resistance and enhanced pathogenicity. This study provides a detailed analysis of the subtypes, and the host range of the IAV isolates and the genetic variations related to their proteins, which may aid in the prevention, treatment, and control of IAV infections in Bangladesh, and would serve as a basis for future investigations.
... Furthermore, empirical evidence on seasonal epidemics such as influenza, dengue virus (DENV), zika virus (ZIKV), and Ebola virus (EBOV) also have a considerable economic impact in the affected areas. In the United States, the annual economic costs of influenza varied from 13.9 thousand dollars to 957.5 million dollars across U.S. counties, with a median of 2.47 million dollars (Liang et al., 2012) (-20.5 percent) (L. He, 2020). ...
Article
Full-text available
This paper formulates an analytical framework to understand the spatiotemporal patterns of epidemic disease occurrence, its relevance, and implications to financial markets activity. The paper suggests a paradigm shift: a new multi-dimensional geometric approach to capture all symmetrical and asymmetrical strategic graphical movement. Furthermore, it introduces the concept of stagpression, a new economic phenomenon to explain the uncharted territory the world economies and financial markets are getting into. The Massive Pandemic Contagious Diseases Damage on Stock Markets Simulator (φ-Simulator) to evaluate the determinants of capital markets behavior in the presence of an infectious disease outbreak. The model investigates the impact of Covid-19 on the performance of ten stock markets, including S&P 500, TWSE, Shanghai Stock Exchange, Nikkei 225, DAX, Hang Seng, U.K.-FTSE, KRX, SGX, and Malaysia-FTSE.
Article
Aralık 2019'un sonlarında, yeni ve bulaşıcı bir hastalık (COVID-19), Çin'in en kalabalık şehri olan Wuhan'da ortaya çıktı. Çin hükümetinin 2020’nin başlarında virüsle ilgili bilgileri halka açık olarak paylaşmasının hemen ardından, topluluk karantinaları ve geçici iş yerleri kapatma gibi daha katı önleyici tedbirler alındı. Bu çalışmada COVİD-19’un Borsa İstanbul Alt Sektörleri üzerinde etkisini incelemek amacıyla Basit Doğrusal Regresyon analizi kullanılmıştır. Analizde bağımsız değişken olarak COVID-19 günlük vaka sayısı, bağımlı değişkenler olarak ise Borsa İstanbul Alt Sektör (Turizm, Ulaştırma, Tekstil-Deri, Gıda-İçecek, Sınai) endeks verileri kullanılmıştır. Analiz sonucunda COVID-19 salgının Turizm, Tekstil ve Sınai Sektör Endeksleri üzerinde nispeten yüksek düzeyde anlamlı ve olumsuz etkisinin olduğuna dair bulgulara erişilmiştir. Gıda-İçecek ve Ulaştırma Endekseleri üzerinde ise daha düşük düzeyde anlamlı ve olumsuz etkisinin olduğu gözlemlenmiştir.
Article
Full-text available
Influenza viruses lead to substantial morbidity and mortality including ~3-5 million cases of severe illness and ~290,000-650,000 deaths annually. One of the major hurdles regarding influenza vaccine efficacy is generating a durable, robust cellular immune response. Appropriate stimulation of the innate immune system is key to generating cellular immunity. Cross-talk between innate dendritic cells (DC) and natural killer (NK) cells plays a key role in activating virus-specific T cells, yet the mechanisms used by influenza A viruses (IAV) to govern this process remain incompletely understood. Here, we used an ex vivo autologous human primary immune cell culture system to evaluate the impact of DC-NK cell cross-talk and subsequent naïve T cell activation at steady-state and after exposure to genetically distinct IAV strains–A/California/07/2009 (H1N1) and A/Victoria/361/2011 (H3N2). Using flow cytometry, we found that exposure of DCs to IAV in co-culture with NK cells led to a decreased frequency of CD83⁺ and CD86⁺ cells on DCs and an increased frequency of HLA-DR⁺ on both DCs and NK cells. We then assessed the outcome of DC-NK cell cross-talk on T cell activation. At steady-state, DC-NK cell cross-talk increased pan T cell CD69 and CD25 expression while exposure to either IAV strain reduced pan T cell CD25 expression and suppressed CD4⁺ and CD8⁺ T cell IFN-γ and TNF production, following chemical stimulation with PMA/Ionomycin. Moreover, exposure to A/Victoria/361/2011 elicited lower IFN-γ production by CD4⁺ and CD8⁺ T cells compared with A/California/07/2009. Overall, our results indicate a role for DC-NK cell cross-talk in T cell priming in the context of influenza infection, informing the immunological mechanisms that could be manipulated for the next generation of influenza vaccines or immunotherapeutics.
Article
Full-text available
We quantified the effects of adherence to various non-pharmaceutical interventions (NPIs) on the seasonal influenza epidemic dynamics in Japan during 2020. The total monthly number of seasonal influenza cases per sentinel site (seasonal influenza activity) reported to the National Epidemiological Surveillance of Infectious Diseases and alternative NPI indicators (retail sales of hand hygiene products and number of airline passenger arrivals) from 2014-2020 were collected. The average number of monthly seasonal influenza cases in 2020 had decreased by approximately 66.0% (p < 0.001) compared to those in the preceding six years. An increase in retail sales of hand hygiene products of ¥1 billion over a 3-month period led to a 15.5% (95% confidence interval [CI]: 10.9-20.0%; p < 0.001) reduction in seasonal influenza activity. An increase in the average of one million domestic and international airline passenger arrivals had a significant association with seasonal influenza activity by 11.6% at lag 0-2 months (95% CI: 6.70-16.5%; p < 0.001) and 30.9% at lag 0-2 months (95% CI: 20.9-40.9%; p < 0.001). NPI adherence was associated with decreased seasonal influenza activity during the COVID-19 pandemic in Japan, which has crucial implications for planning public health interventions to minimize the health consequences of adverse seasonal influenza epidemics.
Article
Full-text available
The COVID‐19 global pandemic has triggered one of the greatest economic shocks in a century. Effective COVID‐19 vaccines have been developed, but a proportion of people either are hesitant or refuse to be vaccinated, facilitated by a global misinformation campaign. If ‘herd immunity’ cannot be achieved, there is potential not only for ongoing surges in infection, but also for development of new strains of the virus that could evade vaccines and precipitate further health and economic crises. We review the economics of vaccination and of vaccine hesitancy and refusal, and their potential effects on the recovery from the COVID‐19 pandemic.
Article
Full-text available
Influenza vaccines have been available for over 80 years. They have contributed to significant reductions in influenza morbidity and mortality. However, there have been limitations in their effectiveness, in part due to the continuous antigenic evolution of seasonal influenza viruses, but also due to the predominant use of embryonated chicken eggs for their production. The latter furthermore limits their worldwide production timelines and scale. Therefore today, alternative approaches for their design and production are increasingly pursued, with already licensed quadrivalent seasonal influenza vaccines produced in cell cultures, including based on a baculovirus expression system. Next-generation influenza vaccines aim at inducing broader and longer-lasting immune responses to overcome seasonal influenza virus antigenic drift and to timely address the emergence of a new pandemic influenza virus. Tailored approaches target mechanisms to improve vaccine-induced immune responses in individuals with a weakened immune system, in particular older adults.
Article
Full-text available
Current seasonal influenza virus vaccines target regions of the hemagglutinin (HA) head domain that undergo constant antigenic change, forcing the painstaking annual reformulation of vaccines. The development of broadly protective or universal influenza virus vaccines that induce cross-reactive, protective immune responses could circumvent the need to reformulate current seasonal vaccines. Many of these vaccine candidates target the HA stalk domain, which displays epitopes conserved within and across influenza virus subtypes, including those with pandemic potential. While HA head-mediated antigenic drift is well understood, the potential for antigenic drift in the stalk domain is understudied. Using a panel of HA stalk-specific monoclonal antibodies (MAbs), we applied selection pressure to the stalk domain of A/Netherlands/602/2009 (pdmH1N1) to determine fitness and phenotypes of escape mutant viruses (EMVs). We found that HA stalk MAbs with lower cross-reactivity caused single HA stalk escape mutations, whereas MAbs with broader cross-reactivity forced multiple mutations in the HA. Each escape mutant virus greatly decreased mAb neutralizing activity, but escape mutations did not always ablate MAb binding or Fc-Fc receptor-based effector functions. Escape mutant viruses were not attenuated in vitro but showed attenuation in an in vivo mouse model. Importantly, mice vaccinated with a chimeric HA universal vaccine candidate were protected from lethal challenge with EMVs despite these challenge viruses containing escape mutations in the stalk domain. Our study indicates that while the HA stalk domain can mutate under strong MAb selection pressure, mutant viruses may have attenuated phenotypes and do not evade a polyclonal, stalk-based vaccine-induced response. IMPORTANCE Broadly protective or universal influenza virus vaccines target viral epitopes that appear to be conserved. However, it is unclear whether the virus will be able to escape once immunological pressure is applied to these epitopes through vaccination of large proportions of the population. Studies that investigate the fitness and antigenic characteristics of viruses that escape immunological pressure on these conserved epitopes are therefore urgently needed.
Chapter
In Applied Geostatistics the authors demonstrate how simple statistical methods can be used to analyse earth science data. In clear language, they explain how various forms of the estimation method called kriging can be employed for specific problems. A case study of a simulated deposit is the focus for the book. This model helps the student develop an understanding of how statistical tools work, serving as a tutorial to guide readers through their first independent geostatistical study.
Article
Geographical information systems could be improved by adding procedures for geostatistical spatial analysis to existing facilities. Most traditional methods of interpolation are based on mathematical as distinct from stochastic models of spatial variation. Spatially distributed data behave more like random variables, however, and regionalized variable theory provides a set of stochastic methods for analysing them. Kriging is the method of interpolation deriving from regionalized variable theory. It depends on expressing spatial variation of the property in terms of the variogram, and it minimizes the prediction errors which are themselves estimated. We describe the procedures and the way we link them using standard operating systems. We illustrate them using examples from case studies, one involving the mapping and control of soil salinity in the Jordan Valley of Israel, the other in semi-arid Botswana where the herbaceous cover was estimated and mapped from aerial photographic survey.
Article
Google Flu Trends can detect regional outbreaks of influenza 7–10 days before conventional Centers for Disease Control and Prevention surveillance systems. We describe the Google Trends tool, explain how the data are processed, present examples, and discuss its strengths and limitations. Google Trends shows great promise as a timely, robust, and sensitive surveillance system. It is best used for surveillance of epidemics and diseases with high prevalences and is currently better suited to track disease activity in developed countries, because to be most effective, it requires large populations of Web search users. Spikes in search volume are currently hard to interpret but have the benefit of increasing vigilance. Google should work with public health care practitioners to develop specialized tools, using Google Flu Trends as a blueprint, to track infectious diseases. Suitable Web search query proxies for diseases need to be established for specialized tools or syndromic surveillance. This unique and innovative technology takes us one step closer to true real-time outbreak surveillance.
Article
In this paper the examination of the modifiable areal unit problem is extended into multivariate statistical analysis. In an investigation of the parameter estimates from a multiple linear regression model and a multiple logit regression model, conclusions are drawn about the sensitivity of such estimates to variations in scale and zoning systems. The modifiable areal unit problem is shown to be essentially unpredictable in its intensity and effects in multivariate statistical analysis and is therefore a much greater problem than in univariate or bivariate analysis. The results of this analysis are rather depressing in that they provide strong evidence of the unreliability of any multivariate analysis undertaken with data from areal units. Given that such analyses can only be expected to increase with the imminent availability of new census data both in the United Kingdom and in the USA, and the current proliferation of GIS (geographical information system) technology which permits even more access to aggregated data, this paper serves as a topical warning.