Preventive Veterinary Medicine 111 (2013) 10–16

Contents lists available at SciVerse ScienceDirect

Preventive Veterinary Medicine

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

Spatial accessibility to vaccination sites in a campaign against

rabies in São Paulo city, Brazil
Gina Polo a,∗ , Carlos Mera Acosta b , Ricardo Augusto Dias a
a
Laboratory of Epidemiology and Biostatistics, Department of Preventive Veterinary Medicine and Animal Health, School of Veterinary
Medicine, University of São Paulo, Av. Prof. Dr. Orlando Marques de Paiva, 87, CEP: 05508270, Cidade Universitária, São Paulo, Brazil
b
Institute of Physics, University of São Paulo, Rua do Matão Travessa R Nr. 187, CEP: 05508090, Cidade Universitária, São Paulo, Brazil

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

Article history: It is estimated that the city of São Paulo has over 2.5 million dogs and 560 thousand cats.
Received 12 December 2012 These populations are irregularly distributed throughout the territory, making it difficult
Received in revised form 9 March 2013
to appropriately allocate health services focused on these species. To reasonably allocate
Accepted 24 March 2013
vaccination sites, it is necessary to identify social groups and their access to the referred
service. Rabies in dogs and cats has been an important zoonotic health issue in São Paulo
Keywords:
and the key component of rabies control is vaccination. The present study aims to intro-
Vaccination against rabies
Spatial accessibility duce an approach to quantify the potential spatial accessibility to the vaccination sites of the
Gaussian based two-step floating 2009 campaign against rabies in the city of São Paulo and solve the overestimation associ-
catchment method ated with the classic methodology that applies buffer zones around vaccination sites based
Dogs and cats on Euclidean (straight-line) distance. To achieve this, a Gaussian-based two-step floating
catchment area method with a travel-friction coefficient was adapted in a geographic infor-
mation system environment, using distances along a street network based on Dijkstra’s
algorithm (short path method). The choice of the distance calculation method affected the
results in terms of the population covered. In general, areas with low accessibility for both
dogs and cats were observed, especially in densely populated areas. The eastern zone of
the city had higher accessibility values compared with peripheral and central zones. The
Gaussian-based two-step floating catchment method with a travel-friction coefficient was
used to assess the overestimation of the straight-line distance method, which is the most
widely used method for coverage analysis. We conclude that this approach has the potential
to improve the efficiency of resource use when planning rabies control programs in large
urban environments such as São Paulo. The findings emphasize the need for surveillance
and intervention in isolated areas.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction and cats since 1968. However, following several adverse

In Latin America, rabies is transmitted mainly by mam-
malian carnivores and vampire bats (Diaz et al., 1994). In
the city of São Paulo, Brazil, urban rabies has been effec-
tively controlled by annual vaccination campaigns for dogs
∗ Corresponding author. Tel.: +55 11 3091 1393; fax: +55 11 3091 7928.
E-mail addresses: gina@vps.fmvz.usp.br, imaginapolo@gmail.com
(G. Polo).
reactions to a rabies vaccine, officials of the government
of São Paulo decided to suspend the rabies vaccination
campaign in 2010, making it available again in 2011. Nev-
ertheless, outbreaks of rabies transmitted by vampire bats
have menaced not only livestock but also humans in rural
and urban areas (Sato et al., 2005). In 2011, after 28 years
without detected cases of dog or cat rabies, a cat was diag-
nosed and confirmed positive for rabies variant AgV3. This
probably shows that it was not long for the disease to
be re-introduced after the interruption of vaccination that

0167-5877/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.prevetmed.2013.03.010
 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16
took place in 2010. Promoting access to rabies vaccination
campaigns across all population groups is a long-standing
priority for the public health service in São Paulo. To rea-
sonably allocate the vaccination sites, the public service
first needs to identify the social groups and areas that have
poor access to the service (Wan et al., 2012).
Access to health care services refers to the relative ease
with which healthcare can be reached from a given location
(Guagliardo, 2004; Wang, 2007; Luo and Whippo, 2012).
Many factors may influence access to vaccination sites.
These include, but are not limited to, the availability of vac-
cination sites in the area (supplies), the number of dogs
and cats in this area (demands), the geographic barriers
between supplies and demands, and people’s awareness
of the benefit of program adherence, among others (Dai,
2011).
Various methods have been proposed to estimate spa-
tial access. These methods include the regional availability
model (Khan, 1992), kernel density models (Guagliardo,
2004), the gravity model (Joseph and Bantock, 1982) and
floating catchment methods (Luo and Wang, 2003). The
two-step floating catchment 2SFCA method, first proposed
by Radke and Mu (2000), modified by Luo and Wang (2003),
and enhanced by Luo and Qi (2009), is suitable for mea-
suring accessibility to vaccination sites. The basic 2SFCA
model works in two steps. It first estimates the demand
for each service site and calculates the site-to-population
ratio according to the site’s capacity and the local demand.
The second step sums up the site-to-population ratios of
the nearby sites for each population. The 2SFCA method
has been employed in a number of studies to estimate spa-
tial access to human healthcare services (Guagliardo, 2004;
Langford and Higgs, 2006; Yang et al., 2006; Wang, 2007;
Wang et al., 2008) but there are no reports or studies using
this methodology in veterinary medicine. This method is
limited because it assumes that all the population locations
within the catchment have equal access while disregard-
ing the distance impedance within the catchment (Luo
and Wang, 2003). The most recent version of 2SFCA is the
E2SFCA method (Luo and Qi, 2009), which was designed
to overcome the limitation of the basic 2SFCA method. The
E2SFCA method also works in two steps. In the two steps,
a 30-min travel time catchment area is generated for each
point, and the catchment area is divided into three sub-
zones of 10-, 20-, and 30-min travel intervals. However, this
method is limited because it also assumes that all the pop-
ulation locations within the sub-zones have equal access
and disregards the distance impedance within each travel
time zone.
This article provides an approach to measuring the spa-
tial accessibility to rabies vaccination sites in the city of São
Paulo, and potentially in other cities, and alerts the pub-
lic services to take measures in areas with limited spatial
access.
2. Methods
2.1. Study area and data sources
The study area was the city of São Paulo, located
in southeastern Brazil, which has 11,244,369 inhabitants
11
(IBGE, 2010) and more than 2.5 million dogs and 560 thou-
sand cats (Canatto, 2010). In 2009, the rabies vaccination
program had 1742 vaccination sites distributed across the
city, with the capacity to administer 1000 daily doses at
each site. The address of each vaccination site was obtained
from the website of the prefecture of the city (São Paulo,
2009). The geocoding process was performed using the
most precise method according to Bell et al. (2012), by man-
ually converting the address information to geographical
coordinates using a geographical information system envi-
ronment. To represent the population of humans, 18,953
centroids were obtained from every census tract in the city
(IBGE, 2010) and for each centroid was calculated the popu-
lation of dogs and cats (Canatto, 2010). These census tracts
are meaningful to residents, planners and decision makers.
The distribution of the vaccination sites and the population
centroids is shown in Fig. 1.
2.2. Data analysis
2.2.1. Distance calculation
The overestimates of straight-line distances with regard
to street-network distances were also evaluated. The spa-
tial coverage of service areas were obtained by GIS for
buffer zones around the vaccination sites, based on the
Euclidean distance (straight-line distance) and on the
distances along a street network based on Dijkstra’s algo-
rithm (short-path method) (Dijkstra, 1959). The procedure
for calculating Euclidean distances has the advantage of
simplicity, although it is somewhat inaccurate because
the pedestrian follows the layout of the street network,
not a straight line. To avoid the overestimation of the
aforementioned procedure, it is possible to use analysis
capabilities of the GIS to calculate the distance along the
street network, simulating the real routes based on the
shortest path method (Gutierrez and Garcia-Palomares,
2008).
2.2.2. The model
To generalize the assumption of previous method-
ologies, instead of considering sub-zones which are
discriminated with different weights, we integrated a
Gaussian function into the 2SFCA method, as proposed by
Dai (2010), to continuously discount the access within a
catchment, considering that each population centroid k,
has a different Gaussian weight G related to the distance
dkj to a vaccination site j. Thus, if two points are close,
they will have different accessibilities, which cannot be
observed with sub-zones. A fact that should be empha-
sized is that each population location has the ability to
access more than one vaccination site, which also avoids
overestimation. Furthermore, a friction coefficient ˇ was
incorporated into the Gaussian weight to represent a value
for walking under road conditions. The first step was to
generate a catchment with a threshold travel distance of
800 m d0 for each vaccination site j, and search all popu-
lation locations k within the d0 . The population at k will
be weighed using a Gaussian function G. The weighted
populations within the catchment for j represent the poten-
tial users for the vaccination sites. Computing the ratio Rj
12 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16

Fig. 1. Spatial distribution of vaccination sites and population centroids. São Paulo, 2012.

of the vaccination site to a population location is given population location k and a vaccination site j and G is the
by: Gaussian weight:
⎧ 2
⎨ e−ˇ(dkj /d0 ) − e−ˇ
Sj , if dkj ≤ 0
Rj =  (1) G(dkj , d0 ) = 1 − e−ˇ (2)
P G(dkj , d0 ) ⎩
k∈(dkj ≤d0 ) k 0, if dkj > 0
where ˇ is the friction coefficient. The impedance function
where Pk is the population at location k within the catch- for the model used in this paper, with three different values
ment d0 ; Sj is the capacity (i.e., vaccination doses) of the of friction coefficient ˇ (0.75, 0.5 and 0.25), is presented in
vaccination site j; dkj is the travel distance between a Fig. 2. The coefficient of friction during walking represents
 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16 13

Fig. 2. Impedance function for the Gaussian-based two-step floating catchment area method with different values for the travel-friction coefficient ˇ. São
Paulo, 2012.

a value between the foot and the road surface and deter-
mines a measure of the ease with which a person can move
on a certain surface (Marpet, 1996). The coefficient of fric-
tion of 0.5 is usually used as a safe value for a pedestrian
walking on different types of soil (Bell et al., 2012).
The second step was to search all vaccination sites l
within the threshold distance d0 for each population loca-
tion i, thus formulating the catchment for the population
at i. After subtracting each Rl using the Gaussian function
G, the vaccination-to-population ratios were summed to
obtain the spatial accessibility at population location i:

Ai = Rl G(dil , d0 ) (3)
k∈(dkj ≤d0 )

where l denotes all vaccination sites within the catchment

of population location i, and all other annotations are the
same as in Eq. (1). It should be emphasized that each pop-
ulation location has access to more than one vaccination
site, which avoids the overestimation problem.
This study uses the Network Analyst Tool in ArcGis 10.1
(Environmental System Research Institute, Inc., Redlands,
CA). Using the road network of São Paulo, the calculation of
the distances between census tract centroids and the vac-
cination sites were computed using the origin–destination
cost matrix function.

3. Results

Given a walking threshold distance of 800 m, the size

of the buffer zone is constant (d02 ), but the size of the
network distance area is variable. The service areas deter-
mined using the Euclidean distances are wider than the
street-network distances. The mean area of the buffer
Fig. 3. Calculation of service area in straight line (circle), and through
the street network (irregular polygon), based on one vaccination site. The
two contours do not coincide at any point. The part of the buffer zone not
covered by the network distance coverage area is the overestimated area.
14 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16

Fig. 4. Comparative maps of the population covered by the vaccination program using (a) circular buffer zones and (b) the network distance zones.

Fig. 5. Spatial accessibility for (a) dogs and (b) cats to vaccination sites in the city of São Paulo using the Gaussian-based two-step floating catchment area
method with a travel friction coefficient.
 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16
zones computed using Euclidean distances was calculated
in 2,010,285.2 m2 and the mean value of the irregular
polygons determined using the shortest path method
was calculated in 87,256.9 m2 , such that the first method
overestimates the size of the population service areas in
113,508.3 m2 (Fig. 3). The straight-line distance and street-
network distance service areas coincide only partially. Part
of the service area using a straight line is not included in the
service area calculated using the network distance. Special
attention should be paid to this area because the straight-
line method tends to overestimate the population covered
by the vaccination program (Fig. 4).
The spatial accessibility to rabies vaccination sites for
dogs or for cats determined using the Gaussian-based
2SFCA method with the travel-friction coefficient is pre-
sented in Fig. 5. This method identified a large area of the
city with low spatial accessibility. The eastern zone has a
higher spatial access, whereas the peripheral and central
areas have lower spatial access. The vaccination sites that
remained in one location for more than one day during
the campaign (2.41%) coincided with the areas with better
spatial accessibility. Areas with low accessibility for both
dog and cat vaccination sites were observed, especially
in densely populated areas. Furthermore, the methodol-
ogy identified the area where the 2011 feline rabies case
occurred as a zone with poor accessibility for cat vaccina-
tion during the campaign.
4. Discussion and conclusion
This study advances the measure of spatial accessi-
bility to vaccination sites using a Gaussian-based 2SFCA
method that incorporates a travel-friction coefficient. This
approach allows for more realistic analysis of spatial acces-
sibility compared to other measures widely used in spatial
coverage studies. The major advantage of this method lies
in its more reasonable assumption of distance, consider-
ing the impedance for each individual population location
and the possibility that each population location may have
access to more than one vaccination site. People in dif-
ferent locations are likely to be surrounded by different
sets of vaccination site opportunities and transportation
networks; thus, they may exhibit different travel patterns
in accessing different vaccination sites. Furthermore, the
use of the friction coefficient value generates a spatial
accessibility model that more realistically represents walk-
ing conditions (Marpet, 1996; Blanchette et al., 2011). This
work used a friction coefficient value of 0.5 because this
represents a conservative value for an ordinary pedes-
trian walking under many foreseeable conditions (Marpet,
1996).
As indicated by this study of the city of São Paulo, this
article proposes a methodological approach that provides
a biologically plausible improvement over existing meth-
ods and is useful for identifying low vaccination coverage
sites. Provision of vaccination sites in these problem areas
could improve vaccination coverage and human and ani-
mal health. The travel distance along the road network
used in this study reveals a more realistic measure than
that found using the Euclidean distance. The buffer zone
method is widely used in accessibility studies due to its ease
15
of calculation (Gutierrez and Garcia-Palomares, 2008) from
assuming that the walking distance for a person accessing
a vaccination site is the same as the Euclidean distance.
However, the actual walking distance is longer due to the
design of the streets. As a result, the buffer zone (straight-
line distance) method tended to overestimate the area and
the population covered by the vaccination program. For
this reason, we used distances along a street network based
on Dijkstra’s algorithm (network-distance) in the present
study. A travel distance d0 of 800 m was used in previ-
ous research and is a reasonable walking distance based
on empirical studies (Lovett et al., 2002; Bell et al., 2012).
In the present work, this travel distance was used because
there is limited information on the distance traveled by a
person in search of a vaccination service for their pet in the
city.
The geographical patterns of the results for São Paulo
show that the central area is a zone with low accessibility.
This is consistent with studies that have shown that the
majority of the population in this area uses private vac-
cination services and little effort was made by the public
service to catch this area (Canatto, 2010). This method also
identified a large proportion of the peripheral zone of the
city as a lower access area, making this area ‘isolated’ from
public services. This peripheral area is a very critical zone
because it is the urban–rural interface at risk of reintro-
ducing the rabies virus. The cat population showed greater
accessibility values most likely due to the smaller number
of individuals of this species. This means that a single point
would have a larger catchment, serving a larger number of
cats in the coverage area.
This study used a walking model. Considering other
travel models, such as private vehicles and public trans-
portation, may be necessary to achieve a complete
understanding of accessibility to vaccination sites. Addi-
tionally, this study focused on the potential access to
vaccination sites for the entire population, not taking
into account how access may differ based on non-spatial
characteristics such as age, socioeconomic status, gender,
ethnicity or other variables.
The GIS-based accessibility model allows for the detec-
tion of shortage areas that would not otherwise have been
suspected. It provides an effective tool for potentially mea-
suring access to different health programs and has an
important policy implication for rabies program planning
in the city of São Paulo.
References
Bell, S., Wilson, K., Ikram Shah, T., Gersher, S., Elliot, T., 2012]. Investigat-
ing impacts of positional error on potential health care accessibility.
Spatial Spatio-Temporal Epidemiol. 3, 17–29.
Blanchette, M., Brault, J., Powers, C., 2011]. The influence of heel height
on utilized coefficient of friction during walking. Gait Posture 34,
107–110.
Canatto, B.D., 2010. Caracterização das populações de cães e gatos domicil-
iadas no município de São Paulo. Dissertação de mestrado. University
of São Paulo.
Dai, D., 2010]. Black residential segregation, disparities in spatial access
to health care facilities, and late-stage breast cancer diagnosis in
metropolitan Detroit. Health Place 16, 1038–1052.
Dai, D., 2011]. Racial/ethnic and socioeconomic disparities in urban green
space accessibility: where to intervene? Landscape Urban Plan. 102,
234–244.
16 G. Polo et al. / Preventive Veterinary Medicine 111 (2013) 10–16
Diaz, A., Papo, S., Rodriguez, A., Smith, J.S., 1994]. Antigenic analysis of
rabies-virus isolates from Latin America and the Caribbean. Zbl. Vet.
41, 153–160.
Dijkstra, E., 1959]. A note on two problems in connection with graphs.
Numer. Math. l, 269–27I.
Guagliardo, M.F., 2004]. Spatial accessibility of primary care: concept,
methods and challenges. Int. J. Health Geogr. 3, 3.
Gutierrez, J., Garcia-Palomares, 2008]. Distance measure impacts on the
calculation of transport service areas using GIS. Environ. Plan. B: Plan.
Des. 35, 480–503.
IBGE. Instituto Brasileiro de Geografia e Estatística, 2010. Censo popula-
cional. São Paulo.
Joseph, A., Bantock, R., 1982]. Measuring potential physical accessibility
to general practitioners in rural areas: a method and case study. Soc.
Sci. Med. 16 (1), 85e90.
Khan, A., 1992]. An integrated approach to measuring potential spa-
tial access to health care services. Socio-Econ. Plan. Sci. 26 (4),
275e287.
Langford, M., Higgs, G., 2006]. Measuring potential access to primary
health care services: the influence of alternative spatial representa-
tions of population. Prof. Geogr. 58, 294–306.
Lovett, A., Haynes, R., Sunnenberg, G., Gale, S., 2002]. Car travel time
and accessibility by bus to general practitioner services: a study using
patient registers and GIS. Soc. Sci. Med. 55, 97–111.
Luo, W., Qi, Y., 2009]. An enhanced two-step floating catchment area
(E2SFCA) method for measuring spatial accessibility to primary care
physicians. Health Place 15, 1100–1107.
Luo, W., Wang, F., 2003]. Measures of spatial accessibility to health care in
a GIS environment: synthesis and a case study in the Chicago region.
Environ. Plan. B: Plan. Des. 30, 865–884.
Luo, W., Whippo, T., 2012]. Variable catchment sizes for the two-step
floating catchment area (2SFCA) method. Health Place 18, 189–795.
Marpet, M., 1996]. On threshold values that separate pedestrian walkways
that are slip resistant form those that are not. J. Foren. Sci. 5, 747–755.
Radke, J., Mu, L., 2000]. Spatial decomposition, modeling and mapping
service regions to predict access to social programs. Geogr. Inform.
Sci. 6, 105–112.
São Paulo, 2009. http://www.prefeitura.sp.gov.br/cidade/secretarias/
saude/vigilanciaemsaude/controledezoonoses/raivaanimal/index.
php?p=5435
Sato, G., Tanabe, H., Shoji, Y., Itou, T., Ito, F., Sato, T., Sakai, T., 2005]. Rapid
discrimination of rabies viruses isolated from various host species in
Brazil by multiplex reverse transcription polymerase chain reaction.
J. Clin. Virol. 33, 267–273.
Wan, N., Benjamin Zhan, F., Zou, B., Chow, E., 2012]. A relative spatial
access assessment approach for analyzing potential spatial access to
colorectal cancer services in Texas. Appl. Geogr. 32, 191–299.
Wang, L., 2007]. Immigration, ethnicity, and accessibility to culturally
diverse family physicians. Health Place 13, 656–671.
Wang, F., McLafferty, S., Escamilla, V., Luo, L., 2008]. Late-stage breast
cancer diagnosis and health care access in Illionois. Prof. Geogr. 60,
54–69.
Yang, D., George, R., Mullner, R., 2006]. Comparing GIS-based methods
of measuring spatial accessibility to health services. J. Med. Syst. 30,
23–32.