Sa

DGa

Rb

Mc

Md

a

ARRAA

KMSEVN

1

moa2t

FR

0h

Acta Tropica 131 (2014) 56–62

Contents lists available at ScienceDirect

Acta Tropica

jo u r n al homep age: www.elsev ier .com/ locate /ac ta t ropica

patial analysis for identification of priority areas for surveillancend control in a visceral leishmaniasis endemic area in Brazil

avid Soeiro Barbosaa,∗, Vinícius Silva Beloa,b, Maurício Eduardo Salgado Rangelc,uilherme Loureiro Wernecka,d

Escola Nacional de Saúde Pública Sergio Arouca (ENSP), Fundac ão Oswaldo Cruz (FIOCRUZ), Rua Leopoldo Bulhões no. 1480, 21041-210 Rio de Janeiro,J, BrazilDepartamento de Ciências Básicas da Saúde, Universidade Federal de Juiz de Fora – UFJF , Governador Valadares,G, Brazil

DEGEO/GEOPRO – Departamento de Geociências/Laboratório de Geoprocessamento e Análises Espaciais/Universidade Federal do Maranhão, São Luís,A, BrazilInstituto de Medicina Social, Universidade do Estado do Rio de Janeiro, Rio de Janeiro, RJ, Brazil

r t i c l e i n f o

rticle history:eceived 25 March 2013eceived in revised form 10 October 2013ccepted 4 December 2013vailable online 14 December 2013

eywords:edical geography

patial data analysispidemiological studiesisceral leishmaniasisortheast Brazil

a b s t r a c t

Spatial analysis of epidemiological data may be used to assist in the implementation of surveillance andcontrol measures against visceral leishmaniasis (VL) in endemic areas. This ecological study aimed toidentify priority areas for surveillance and control of VL in São Luís, the capital of the state of Maranhãoin northeast Brazil, a highly endemic area for the disease. We evaluated the spatial structure of the inci-dence rates of human VL and of the mean number of human and canine cases occurring between 2005 and2007 in 355 neighborhoods (aggregated into 203 geographical analytical units) within the municipality.The presence of spatial autocorrelation was explored using global and local Moran’s I statistics. A localindicator of spatial autocorrelation was used to generate maps for the identification of VL clusters. Theglobal Moran’s I index revealed a weak, but statistically significant spatial autocorrelation for human VLincidence rates (I = 0.138). A total of 43 geographical analytical units, encompassing 121 neighborhoods,were identified as priority areas for implementing surveillance and control actions. For the purpose ofdefining an action plan for the delivery of these measures, those 16 geographical analytical units (encom-passing 54 neighborhoods) identified as clusters with high incidence rates of human VL should receive thehighest priority. An additional nine geographical analytical units (comprising 28 neighborhoods) showed

non-significant clustering of high rates of human, and might be considered as the next priority for VLmanagement. Finally, a further 18 geographical analytical units (covering 39 neighborhoods) had recordsof coexisting human and canine VL cases during the study period, and these should receive priority atten-tion when resources become available. Spatial data analysis is a valuable tool for defining priority areasfor VL surveillance in high transmission areas contributing to a more effective management of financialand technical resources, increasing the sustainability and efficiency of control efforts.

. Introduction

Visceral leishmaniasis (VL) is a major public health problem inany parts of the world. Approximately 0.2–0.4 million VL cases

ccur every year (Alvar et al., 2012) in 12 countries in the Americas

nd in 65 in Europe, Asia and Africa (World Health Organization,010). In Brazil, VL is a zoonotic vector-borne disease widely dis-ributed throughout Brazil with a high mean annual incidence rate

∗ Corresponding author at: Escola Nacional de Saúde Pública Sergio Arouca,undac ão Oswlado Cruz, Rua Leopoldo Bulhões no. 1480, Sala 603, 21041-210io de Janeiro, RJ, Brazil. Tel.: +55 21 971427710.

E-mail address: davidsoeiro@gmail.com (D.S. Barbosa).

001-706X/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.actatropica.2013.12.002

© 2013 Elsevier B.V. All rights reserved.

of around two cases per 100,000 inhabitants (Ministério da Saúdedo Brasil, 2006). VL constitutes an important public health issueowing to the increasing number of new cases notified and the con-tinuous expansion of the disease in state capitals and adjacentmetropolitan regions (Barreto et al., 2011; Costa, 2008; Oliveiraet al., 2008; Werneck, 2008).

A program to monitor and control VL (Programa de Vigilânciae Controle da Leishmaniose Viscera – PVCLV) has been created bythe Brazilian Ministry of Health with the objective of reducingtransmission rates and the mortality and morbidity burden asso-

ciated with the disease. The strategies of this program are (i) earlydiagnosis and treatment of human VL cases, and (ii) control ofthe insect vector population (phlebotomines) and elimination ofinfected reservoirs (dogs, in urban areas) (Gontijo and Melo, 2004;

a Tropica 131 (2014) 56–62 57

M2re2bmfnotdtm

upateoat2

r22aDNilTa2tb

pbic1Hs(d

S2iWmvah

2

2

CA

Fig. 1. Health districts (defined by numbers) and geographical analytical units (sub-divisions) of the municipality of São Luís, state of Maranhão, Brazil.

D.S. Barbosa et al. / Act

inistério da Saúde do Brasil, 2006; Savani et al., 2003; Silva et al.,007). Although PVCLV guidelines emphasize the importance ofisk reduction measures, the strategies employed have not beenffective in reducing the incidence of VL (Romero and Boelaert,010). Indeed, in comparison with the many advances that haveeen made in the control of other infectious diseases in Brazil, theanagement of VL must be considered one of the most striking

ailures (Barreto et al., 2011). In order to augment the effective-ess of the strategies proposed by PVCLV, the Brazilian Ministryf Health recommends that all municipalities classified as intenseransmission zones (mean annual incidence rate ≥4.4 notified casesuring the previous three years) should identify priority areas forhe implementation of the recommended surveillance and control

easures (Ministério da Saúde do Brasil, 2006).Geographic information systems (GIS) are important tools in

nderstanding the epidemiology of infectious diseases, and canrovide valuable contributions to the determination of resourcellocation and implementation of control measures. The combina-ion of GIS with reliable data generated by surveillance activitiesnables the routine construction of maps showing the distributionf vectors, reservoirs and cases of human VL, through which pastnd present situations can be compared and changes in the pat-erns of disease occurrence identified (World Health Organization,010).

Spatial data analyses have been employed in epidemiologicalesearch of VL in Brazil at the municipal (Camargo-Neves et al.,001; Carneiro et al., 2004; Margonari et al., 2006; Oliveira et al.,001; Vigilato, 2004; Werneck et al., 2002, 2007; Lopes et al., 2010)nd state levels (Bavia et al., 2005; Correa-Antonialli et al., 2007;antas-Torres and Brandão-Filho, 2006; Mestre and Fontes, 2007;ieto et al., 2006; Tavares and Tavares, 1999). Some of these stud-

es were performed in Northeast Brazil, with different approaches,evels of analysis and geostatistical techniques used (Tavares andavares, 1999; Bavia et al., 2005; Nieto et al., 2006; Dantas-Torresnd Brandão-Filho, 2006; Werneck et al., 2007; Carneiro et al.,004). However, there is still a need to explore the potential ofhese techniques to generate more practical information to be usedy the health services.

The state of Maranhão in the northeast region of Brazil is ofarticular epidemiological importance with respect to human VLecause the reported incidence of the disease (7.2 cases per 100,000

nhabitants for 2009) is much higher than the average of the wholeountry. Moreover, in São Luís, the state capital, the incidence is.7 cases per 100,000 inhabitants, and the Brazilian Ministry ofealth has classified the municipality as an intense VL transmis-

ion zone because of the elevated average annual number of cases29.6) reported during the period 2007–2009 (Ministério da Saúdeo Brasil, 2011).

While various epidemiological studies have been carried out inão Luís (Caldas et al., 2001, 2002; Costa et al., 1995; Mendes et al.,002; Silva et al., 2008), new approaches are required in order to

mprove the effectiveness of VL surveillance and control measures.ithin this context, the municipality of São Luís was selected as aodel area for a detailed investigation aimed at determining the

alue and limitations of spatial data analysis in identifying priorityreas and developing an action plan for surveillance and control ofuman VL.

. Materials and methods

.1. Ethical considerations

The research project was approved by the Research Ethicsommittee of the National School of Public Health Sergiorouca/Oswaldo Cruz Foundation (ENSP/FIOCRUZ) under protocol

Source: Adapted from Instituto da Cidade (INCID, Prefeitura de São Luís – Maranhão),Instituto Brasileiro de Geografia e Estatística (IBGE) and Secretaria Municipal deSaude (SEMUS)/Prefeitura Municipal de São Luís databases.

number 208/10. No human and animal populations were dam-aged or harmed by the collection, storage, and analysis of the dataemployed in the present study or by the publication of the findingsderived there from.

2.2. Study area

São Luís is located in northern Maranhão in the metropoli-tan region of the Ilha do Maranhão (S 2◦31′47′′; W 44◦18′10′′;altitude 24.391 m). The municipal territory comprises an area ofapproximately 835 km2 with a population of 1,014,837 inhabitants(Instituto Brasileiro de Geografia e Estatística, 2011). Accordingto the City Health Department (Secretaria Municipal de Saude –SEMUS) of São Luís, the city is divided into seven health districts,namely Centro, Itaqui-bacanga, Coroadinho, Cohab, Bequimão,Tirirical and Vila Esperanc a (Prefeitura Municipal de São Luís,2012). The municipality of São Luís encompasses 355 urbanand rural neighborhoods (villages, farms, homesteads, industrialareas, ports, etc.) which, for the purposes of this investigation,were assembled into 203 geographical analytical units (Fig. 1).The geographic analytical units were defined by the researchteam considering the need to make different databases fromvarious sources compatible. The different sources of geographicinformation considered were: (a) the geographic units used by theInstitute for Research and Urban and Rural Planning of São Luís(INCID) (urban neighborhoods); (b) the geographic data availablefrom the Brazilian Institute of Geography and Statistics (InstitutoBrasileiro de Geografia e Estatística – IBGE) (urban and rural censustracts) and (c) field maps used by health workers from SEMUS.

The geographic analytical units were then defined consideringthe established limits in the INCID geographic database, plusnew geographic units created based on census tracts (IBGE) andconsidering the locations defined in the field maps from SEMUS.

5 a Trop

Wa

2

astAw

2

wmArcdtbgnw

utcbma(B

2

stfnegw

tartdpid

IsawliMiai

8 D.S. Barbosa et al. / Act

e considered as rural-urban transition areas those geographicnalytical units described as rural but adjacent to urban areas.

.3. Study design

This ecological study was based on confirmed cases of humannd canine VL notified to SEMUS among residents of São Luís. Thepatial distribution of the incidence of human VL was evaluated forhe whole period of 2005–2007 in order to minimize variability.dditionally, an analysis of coexisting human and canine VL casesithin single geographical analytical units was performed.

.4. Variables and indicators

Records of cases of human and canine VL reported in São Luísere supplied by the Zoonosis Control Center (ZCC) and the Depart-ent of Sanitary and Epidemiological Surveillance of the SEMUS.ll notifications received during the period 2005–2007 involvingesidents of São Luís presenting confirmed symptoms of VL wereonsidered for inclusion in the study. In each case, the area of resi-ence of the individual was taken as an indication of autochthony ofhe disease in the neighborhood where VL cases are usually located,ut for this study VL cases were aggregated according to definedeographic analytical units. Human and canine VL cases that couldot be georeferenced to any of the geographical analytical unitsere excluded from the study.

The incidence rate of human VL in each geographical analyticalnit was determined as the ratio of the number of cases notified andhe population estimate for that analytical unit. Data concerninganine VL came from programmed serological surveys performedy the ZCC in all seven health districts, and on spontaneous com-unity requests for the examination of an animal. Dogs presenting

ntibody titers ≥1:40 in the indirect fluorescent antibody testIFAT) were regarded as positive according to the guidelines of therazilian Ministry of Health (Ministério da Saúde do Brasil, 2006).

.5. Global and local indicators of spatial autocorrelation

Analysis of the spatial patterns of the disease indicators (repre-ented by the incidences of human VL per 100,000 inhabitants) inhe period 2005–2007 was carried out using the Moran’s I statisticollowing the methodology previously described for the determi-ation of malaria risk areas (Atanaka-Santos et al., 2007; Rodriguest al., 2008). For assessing spatial patterns in the VL data over theeographical analytical units we used a contiguity-based spatialeights matrix.

The global Moran’s I index was employed as a measure of spa-ial autocorrelation in order to test whether neighbor geographicalnalytical units showed greater similarity with respect to VL occur-ence than would be expected for a random pattern. In this context,he Moran’s I index tests the null hypothesis of spatial indepen-ence in the entire area (Moran’s I = 0), generating a global I valueotentially varying from −1 to +1: negative values indicating an

nverse correlation (dispersion) while positive values suggesting airect correlation (clustering).

In order to identify localized spatial clusters, the local Moran’s index was calculated, generating normalized values that werepecific for each unit. The values of the local Moran’s I index arerranged in four quadrants of the so-called Moran scatter plothere: (i) quadrant 1, or Q(+/+), indicates those geographical ana-

ytical units with normalized positive values of the local Moran’s Index and a positive average of the normalized values of the local

oran’s I index in its neighboring units; (ii) quadrant 2, or Q(−/−),ndicates the presence of units with normalized negative valuesnd a negative average of the normalized values in the neighbor-ng units; (iii) quadrant 3, or Q(+/−), indicates the presence of a

ica 131 (2014) 56–62

unit with a normalized positive value and a negative average ofthe normalized values in the neighboring units and (iv) quadrant4, or Q(−/+), indicates the presence of a unit with a normalizednegative value and a positive average of the normalized valuesin the neighboring units. The first and second quadrants indicateareas with positive spatial autocorrelation since the geographicalanalytical units and their adjacent neighbors possess normalizedvalues with the same sign. These four quadrants are depicted inthe so-called Box map. Moran maps for the VL indicators wereconstructed in order to highlight the geographical analytical unitsexhibiting statistically significant (P ≤ 0.01) spatial autocorrelation(Atanaka-Santos et al., 2007; Rodrigues et al., 2008).

Neighborhoods encompassing small populations with few VLcases notified in the three year study period exhibited large varia-tions in VL incidence rates. In order to reduce random fluctuationand to facilitate subsequent spatial data analysis, the calculatedincidence rates for each of the geographical analytical units werereestimated using an empirical Bayesian smoothing approach. Thecalculations were based on the first order neighborhood criterion,which considers as neighbor areas those geographical analyticalunits that have common boundaries (Bailey and Gatrell, 1995;Braga, 2007). To evaluate the spatial variability of the data, a spa-tial proximity matrix using contiguity-based spatial weights wasbuilt. The elements of this matrix can take values 1 (if geographicalanalytical units are adjacent) or 0 (otherwise) (Druck et al., 2004).Analyses were performed using TerraView software version 3.6.0(Instituto Nacional de Pesquisas Espaciais, 2010). In addition, a mapindicating the coexistence of human and canine VL cases during thethree year period was constructed.

2.6. Definition of priority areas for surveillance and control of VL

Priority areas for the implementation of VL surveillance andcontrol programs in São Luís were defined according to three dif-ferent proposals: (i) geographical analytical units identified on theMoran map as of the type Q(+/+) showing statistically significant(P ≤ 0.01) spatial autocorrelation according to the Box map, (ii) geo-graphical analytical units included in (i) together with those of thetype Q(+/+) showing non-significant spatial autocorrelation, thatrepresents similar areas with high incidence and (iii) geographi-cal analytical units included in (i) and (ii) plus those units withcoexisting human and canine VL cases.

3. Results

The incidence of human VL in São Luís decreased during theperiod 2005 to 2007 from 5.94 to 4.13 cases per 100,000 inhabi-tants, but the highest case fatality rate associated with the disease(7.3%) was recorded in 2007. During the period under study, theprevalence of seropositive dogs was approximately 7%, with a max-imum of 7.8% reported in 2006. Rural-urban transition zones of themunicipality were most affected by canine infection.

According to the global Moran’s I index, there was a weak butstatistically significant spatial autocorrelation regarding the inci-dence of human VL (I = 0.138; P = 0.04). The Moran scatter plot(Box map) revealed that most geographical analytical units (68.4%;139/203) were distributed in the low priority Q(−/−) quadrant,while quadrants Q(+/−) and Q(−/+) accounted for a further 19.2%(n = 39) of the units (Table 1). The Q(+/+) quadrant contained25 (12.3%) geographical analytical units and, of these, 16 (64%)were in areas with statistically significant spatial autocorrelation

according to the Moran map (Fig. 2A) and were distributed in thehealth districts of Vila Esperanc a and Itaqui-bacanga. The Fig. 2Bshows these geographical analytical units together with the non-significant units identified with positive spatial autocorrelation

D.S. Barbosa et al. / Acta Tropica 131 (2014) 56–62 59

Table 1Priority areas for surveillance and control of human visceral leishmaniasis (VL) in the municipality of São Luís, state of Maranhão, Brazil, stratified according to the incidenceof the disease during the period 2005–2007.

Priority areas Total numberof analyticalunits (N)

Indicatora,b Populationvariabilityb

No. of analytical units stratifiedaccording to positive spatialautocorrelation Q(+/+) in Box map (n)

Significantc Non-significant

Quadrant 1 (+/+)d 25 302.1 (59.4–1538.4) 2418.6 (91–8137) 16 9Quadrant 2 (−/−)e 139 0.67 (0–27.7) 6128.3 (0–51136) 55 84Quadrants 3 (+/−) and 4(−/+)f 39 33.4 (0–472.2) 2064 (3–16416) 15 24

a Incidence of human visceral leishmaniasis per 100,000 inhabitants (number of notified cases/population of the geographical analytical unit).b mean (maximum – minimum).c Moran map (P ≤ 0.01).

Qis2utmt(ch

dPi1wuino

FiBt

d High priority.e Low priority.f Intermediate priority.

(+/+) according to the Box map including priority areas locatedn Vila Esperanc a, Itaqui-bacanga and Tirirical. Analysis of mapshowing human and canine VL cases during the period 2005 to007 (Fig. 3) revealed that the 41 (20.2%) geographical analyticalnits with records of coexisting cases of the disease followed a dis-ribution pattern that was similar to those of the Moran and Box

aps, but with considerably more units located in the Tirirical dis-rict and some in Bequimão (n = 3), Cohab (n = 1) and Coroadinhon = 1). Interestingly, 125 (61.6%) of the 203 analytical units hadases of canine VL only, 2 (1%) had only human VL and 35 (17.2%)ad been free of human and canine VL.

The proposals outlined in Section 2.6 were applied in order toefine priority areas for surveillance and control of VL in São Luís.roposal (i), which considered only statistically significant analyt-cal units identified in the Moran map, defined a cluster containing6 critical areas covering 54 neighborhoods (Fig. 2A). Proposal (ii),hich combined the units of proposal (i) with the non-significant

nits identified with positive spatial autocorrelation Q(+/+) accord-

ng to the Box map (n = 9), containing 25 analytical units covering 82eighborhoods (Fig. 2B). Proposal (iii), which considered the unitsf proposal (i and ii) and additionally those identified as presenting

ig. 2. Priority areas for the implementation of a surveillance and control program for vncluding geographical analytical units identified on the Moran map as of the type Q(+/+)ox map; and (B) geographical analytical units included in (A) together with those of thehe period 2005–2007.

coexisting human and canine VL (n = 18), defined a cluster contain-ing 43 analytical units covering 121 neighborhoods, i.e. 34% of allneighborhoods in the municipality) (Fig. 4).

4. Discussion

In the municipality of São Luís, which is located in the northeastof Maranhão, Northeast Brazil, the maximum incidence of humanVL within the study period occurred in 2005 with 5.94 cases per100,000 inhabitants, whereas the maximum case-fatality rate of7.3% was observed in 2007. In comparison, the maximum incidenceof VL in Campo Grande, the capital of Mato Grosso do Sul (central-west Brazil), was 20.98 cases per 100,000 inhabitants with a case-fatality rate of 11% in 2003 (Correa-Antonialli et al., 2007), whilein Belo Horizonte, the capital of Minas Gerais (southeastern Brazil),the reported incidence for 2004 was 5.64 per 100,000 inhabitantsand the case-fatality rate was 18.6% (Margonari et al., 2006). Datashows that São Luís has high incidence levels of VL in comparison

to other Brazilian cities also considered as high transmission areas.

Dogs have long been considered as the most important reser-voir hosts of Leishmania in urban areas (Amóra et al., 2006). Within

isceral leishmaniasis (VL) in São Luís, state of Maranhão, Brazil. (A) Priority areas showing statistically significant (P ≤ 0.01) spatial autocorrelation according to the

type Q(+/+) showing non-significant spatial autocorrelation. Data collected during

60 D.S. Barbosa et al. / Acta Tropica 131 (2014) 56–62

Fig. 3. Stratification of human and canine cases of visceral leishmaniasis (VL) in SãoLuís, state of Maranhão, Brazil, showing: � areas with absence of human and canineVL; areas with presence of human VL only; areas with presence of canine VLonly; � and areas with cases of human and canine VL. Data collected during thep

tfLtorcame2rcl

aaic2wsccMcLria

Fig. 4. Priority areas for the implementation of a surveillance and control programfor visceral leishmaniasis (VL) in São Luís, state of Maranhão, Brazil, including thosegeographical analytical units identified on the Moran map as of the type Q(+/+)showing statistically significant (P ≤ 0.01) spatial autocorrelation according to theBox map, with those of the type Q(+/+) showing non-significant spatial autocorre-lation, plus those units with coexisting human and canine VL cases. Data collected

eriod 2005–2007.

he period of the present study, the maximum frequency of dogsound to be seropositive for Leishmania in the municipality of Sãouís as a whole was 7.8% (in 2006), although in some rural-urbanransition zones the percentage was even higher. High prevalencesf infection in canine reservoirs have been reported in areas ofecent occupation characterized by dwellings with poor sanitaryonditions and adjacent to woods in which there is generally anbundance of vectors and reservoirs that can contribute to theaintenance and dissemination of VL (Barbosa et al., 2010; Felipe

t al., 2011; Guimarães et al., 2005; Oliveira et al., 2006; Silva et al.,005). In this context, the present study demonstrated that theural-urban transition zones of São Luís were most affected byanine VL with importance for the implementation of VL surveil-ance and control programs.

It was not possible to analyze the correlation between humannd canine VL in the present study because insufficient data werevailable for the calculation of the prevalence of seropositive dogsn the analytical units. However, a number of researchers havelaimed the existence of such an association (Camargo-Neves et al.,001; Oliveira et al., 2001; Vigilato, 2004; Margonari et al., 2006)hile others have voiced dissent (Dietze et al., 1997). While the

ignificance of canine VL for the occurrence of human VL remainsontroversial, studies of urban epidemics have concluded thatanine VL generally precedes human VL (Gontijo and Melo, 2004).uch of the current knowledge concerning the VL transmission

ycle sustains the hypothesis that the dog is the main reservoir ofeishmania within the urban environment, but some aspects of the

elationship between human and canine VL remain to be elucidatedncluding, for example, the level of canine infection that represents

real risk to humans (Gontijo and Melo, 2004).

during the period 2005–2007.

A number of researchers have employed spatial analysis tools inthe study of various infectious diseases, including leishmaniasis, inBrazil (Camargo-Neves et al., 2001; Correa-Antonialli et al., 2007;Dantas-Torres and Brandão-Filho, 2006; Margonari et al., 2006;Mestre and Fontes, 2007; Oliveira et al., 2001; Tavares and Tavares,1999; Vigilato, 2004; Werneck et al., 2002, 2007). Some of themhave shown that the main foci of VL tend to remain unchangedover time (Oliveira et al., 2001; Tavares and Tavares, 1999). Asimilar pattern was also observed in São Luis, where some fociidentified in this study had already been observed in other studies,suggesting that disease incidence is sustained over time in someneighborhoods (Silva et al., 2008). The spatial data analysis of VLpatterns also suggests that disease outbreaks have been observedmore frequently in specific neighborhoods in some health districtsof São Luís, and these areas should be prioritized for continuoussurveillance and control in order to improve the effectiveness ofthe management program. Werneck et al. (2002) employed geo-processing and spatial data analysis to evaluate the distributionof human VL in Teresina, state of Piauí (northeast Brazil), duringthe period 1993–1996, and was able to identify risk areas in whichefforts should be focused in order to augment the effectiveness ofthe control measures. Similar mapping techniques were employedin Arac atuba, state of São Paulo (southeastern Brazil) to facilitatethe identification of areas with elevated potential risk of humanVL (Camargo-Neves et al., 2001). Additionally, Margonari et al.(2006) suggested that control strategies should involve all possi-

ble outbreak areas independent of artificial administrative barriers.However, none of these studies proposed an operational plan ofaction for the health services.

a Trop

mootarlrtsnth

tosashfa

easfiyosttrafcapisiasoedIsihmi

D

A

((ma

D.S. Barbosa et al. / Act

A previous study pointed out that the weaknesses in the imple-entation of the PVCLV program in São Luís included the absence

f a sustainable plan of action, the shortage of staff, and the lackf funds, vehicles, materials and equipment (Silva et al., 2008). Inhe present study, we have not only defined the priority VL riskreas in São Luís, but we have also developed a plan of action. Weecommend that the areas presenting significant spatial autocorre-ation regarding disease occurrence, as assigned in the Moran map,eceive the highest priority for continuous control action. As fur-her financial, technical and human resources become available, theurveillance and management measures should be extended to theon-significant areas identified with positive spatial autocorrela-ion in the Box map and, subsequently, to the areas with coexistinguman and canine VL cases.

To be useful in practical settings, the approach presented inhis paper requires accurate, complete, and timely informationn canine and human cases. Additionally, health professionalshould undergo basic training in geographic information systemsnd spatial analysis in order to produce maps of priority areas forurveillance and control. The spatial analysis techniques appliedere are relatively simple and intuitive and can be developed using

ree software, which facilitates the diffusion and application of suchpproach in most of the highly endemic municipalities in Brazil.

A number of limitations in the present study need to be acknowl-dged, including (i) the use of secondary data that were notcquired exclusively for this study but were obtained from healthervice sources, (ii) the restricted number of human VL cases noti-ed per year, which necessitated their aggregation over the threeear study period, (iii) the lack of data concerning the numberf dogs examined for the presence of VL, which made it impos-ible to calculate the percentage of seropositive dogs in each ofhe analytical units, (iv) the use of IFAT as the standard diagnosticest, even though this method is known to produce false-positiveesults, and (v) the organization of municipal neighborhoods intonalytical units, which could obscure the actual site of an outbreakollowing amalgamation with other neighborhoods. Quality andonsistency of data across years need to be improved in order tollow prompt updates in the identification of priority areas and toermit the development of more formal predictive modeling stud-

es. In general, exploratory techniques such as Moran indices andcatter maps are very useful for identifying priority areas for themplementation of public policies relating to disease monitoringnd control (Druck et al., 2004). The use by health professionals ofpatial data analysis tools combined with conventional method-logies, would contribute to decision making by facilitating thevaluation of suitability and applicability of data in estimating, pre-icting, interpreting and understanding epidemiological evidence.

n the case of human VL, it is clearly essential that priority areashould be more accurately defined. The plan of action developeds a valuable tool for defining such areas for VL surveillance inigh transmission areas in order to improve the management ofaterial, financial and human resources in the context of PVCLV

ncreasing the sustainability and efficiency of control efforts.

isclosure statement

The authors declare that they have no conflicts of interest.

cknowledgments

The authors are grateful to Reinaldo S. Santos and André Perissé

ENSP/FIOCRUZ), Gustavo A.S. Romero (UNB) and José U. BragaIMS/UERJ) for advice during the investigation and helpful com-

ents concerning the preparation of the manuscript. The authorscknowledge the support of FAPEMA and FIOCRUZ for the grants

ica 131 (2014) 56–62 61

awarded to Barbosa, D. S. G.L.W. was partially funded by the Brazil-ian Research Council (CNPq #306267/2010-1). In addition, wethank the technicians and staff from the ZCC (especially to Mariado Carmo Dias Magalhães), SEMUS and INCID for assistance withthe collection of data.

References

Alvar, J., Vélez, I.D., Bern, C., Herrero, M., Desjeux, P., et al., 2012. Leishmaniasisworldwide and global estimates of its incidence. PLoS ONE 7 (5), e35671.

Amóra, S.S.A., Santos, M.J.P., Alves, N.D., Costa, S.C.G., Calabrese, K.S., Monteiro, A.J.,Rocha, M.F.G., 2006. Fatores relacionados com a positividade de cães para leish-maniose visceral em área endêmica do estado do Rio Grande do Norte. Brasil.Cienc. Rural 36, 1854–1859.

Atanaka-Santos, M., Souza-Santos, R., Czeresnia, D., 2007. Spatial analysis for strat-ification of priority malaria control areas, Mato Grosso State. Brazil. Cad. SaúdePública 23, 1099–1112.

Bavia, M.E., Carneiro, D.D., Gurgel, H.C., Madureira-Filho, C., Barbosa, M.G., 2005.Remote sensing and geographic information systems and risk of American vis-ceral leishmaniasis in Bahia. Braz. Parassitol. 47, 165–169.

Bailey, T.C., Gatrell, A.C., 1995. Interactive Spatial Data Analysis. Longman, NY.Barbosa, D.S., Rocha, A.L., Santana, A.A., Souza, C.S.F., Dias, R.A., Costa-Júnior, L.M.,

Abreu-Silva, A.L., 2010. Soroprevalência e variáveis epidemiológicas associadasà leishmaniose visceral canina em área endêmica no município de São Luís,Maranhão. Brasil. Cienc. Anim. Bras. 11, 653–659.

Barreto, M.L., Teixeira, M.G., Bastos, F.I., Ximenes, R.A.A., Barata, R.B., Rodrigues,L.C., 2011. Sucessos e Fracassos no Controle de Doenc as Infecciosas no Brasil:O Contexto Social e Ambiental, Políticas, Intervenc ões e Necessidades dePesquisa. Lancet – Saúde no Brasil 3, 47–60, Available at: http://download.thelancet.com/flatcontentassets/pdfs/brazil/brazilpor3.pdf (accessed 20.08.12).

Braga, E.B., (M.Sc. Dissertation) 2007. Análise Espac o-Temporal da Incidência daLeishmaniose Visceral no Período de 1996 a 2005, Teresina, Piauí. UniversidadeEstadual do Rio de Janeiro, Rio de Janeiro, RJ, Brazil.

Caldas, A.J.M., Costa, J.M., Silva, A.A., Vinhas, V., Barral, A., 2002. Risk factors asso-ciated with asymptomatic infection by Leishmania chagasi in north-east Brazil.Trans. R. Soc. Trop. Med. Hyg. 96, 21–28.

Caldas, A.J.M., Silva, D.R.C., Pereira, C.C.R., Nunes, P.M.S., Silva, B.P., Silva, A.A.M.,Barral, A., Costa, J.M.L., 2001. Infecc ão por Leishmania (Leishmania) chagasi emcriancas de uma área endêmica de leishmaniose visceral americana na ilha deSão Luís-MA. Brasil. Rev. Soc. Bras. Med. Trop. 34, 445–451.

Camargo-Neves, V.L., Katz, F.G., Rodas, L.A.C., Poletto, D.W., Lage, L.C., Spinola, R.M.F.,Cruz, O.G., 2001. Utilizac ão de ferramentas de análise espacial na vigilância epi-demiológica de leishmaniose visceral americana – Arac atuba, São Paulo, Brasil,1998–1999. Cad. Saúde Pública 15, 1263–1267.

Carneiro, D., Bavia, M., Rocha, W., Lobão, J., Madureira-Filho, C., et al., 2004.Identificac ão de áreas de risco para a leishmaniose visceral americana, através deestudos epidemiológicos e sensoriamento remoto orbital, em Feira de Santana,Bahia, Brasil (2000–2002). RBSP 8, 19–32.

Correa-Antonialli, S.A., Torres, T.G., Paranhos-Filho, A.C., Tolezano, J.E., 2007. Spatialanalysis of American visceral leishmaniasis in Mato Grosso do Sul, central Brazil.J. Infect. 54, 509–514.

Costa, C.H.N., 2008. Characterization and speculations on the urbanization of visceralleishmaniasis in Brazil. Cad. Saúde Pública 24, 2959–2963.

Costa, J.M.L., Viana, G.M.C., Saldanha, A.C.R., Nascimento, M.D.S.B., Alvim, A.C., Burat-tini, M.N., Silva, A.R., 1995. Leishmaniose visceral no Estado do Maranhão: aevoluc ão de uma epidemia. Cad. Saúde Pública 11, 321–324.

Dantas-Torres, F., Brandão-Filho, S.P., 2006. Expansão geográfica da leishmaniosevisceral no estado de Pernambuco. Rev. Soc. Bras. Med. Trop. 39, 352–356.

Dietze, R., Barros, G.B., Teixeira, L., Harris, J., Michaelson, K., Falqueto, A., Corey, R.,1997. Effect of eliminating seropositive canines on the transmission of visceralleishmaniasis in Brasil. Clin. Infect. Dis. 25, 1240–1242.

Druck, S., Carvalho, M.S., Câmara, G., Monteiro, A.V.M., 2004. Análise Espacial deDados Geográficos. EMBRAPA, Brasília.

Felipe, I.M.A., Aquino, D.M.C., Kuppinger, O., Santos, M.D.C., Rangel, M.E.S., Bar-bosa, D.S., Barral, A., Werneck, G.L., Caldas, A.J.M., 2011. Leishmania infection inhumans, dogs and sandflies in a visceral leishmaniasis endemic area in Maran-hão, Brazil. Mem. Inst. Oswaldo Cruz. 106, 207–211.

Gontijo, C.M.F., Melo, M.N., 2004. Leishmaniose visceral no Brasil: quadro atual,desafios e perspectivas. Rev. Bras. Epidemiol. 7, 338–349.

Guimarães, K.S., Batista, Z.S., Dias, E.L., Guerra, R.M.N.C., Costa, A.D.C.C., Oliveira, A.S.,Calabrese, K.S., Cardoso, F.O., Souza, C.S.F., Zaverucha-do-Vale, T., Gonc alves-da-Costa, S.C., Abreu-Silva, A.L., 2005. Canine visceral leishmaniasis in São José deRibamar, Maranhão State. Brazil. Vet. Parasitol. 131, 305–309.

Instituto Brasileiro de Geografia e Estatística, 2011. Cities@. Available at:http://www.ibge.gov.br/cidadesat/link.php?uf=ma (accessed20.08.12).

Instituto Nacional de Pesquisas Espaciais, 2010. TerraView 4.1.0. INPE, São José dosCampos, SP, Brazil. Available at: www.dpi.inpe.br/terraview (accessed 20.08.12).

Lopes, E.G.P., Magalhães, D.F., Silva, J.A., Haddad, J.P.A., Moreira, E.C., 2010.Distribuic ão temporal e espacial da leishmaniose visceral em humanos e cãesem Belo Horizonte-MG, 1993 a 2007. Arq Bras. Med. Vet. Zootec 62, 1062–1071.

Margonari, C., Freitas, C.R., Ribeiro, R.C., Moura, A.C.M., Timbó, M., Gripp, A.H.,Pessanha, J.E., Dias, E.S., 2006. Epidemiology of visceral leishmaniasis through

6 a Trop

M

M

M

M

O

O

O

P

N

R

Epidemiol. Infect. 135, 195–201.World Health Organization, 2010. Control of the Leishmaniases. Report of a Meet-

2 D.S. Barbosa et al. / Act

spatial analysis, in Belo Horizonte municipality, state of Minas Gerais, Brazil.Mem. Inst. Oswaldo Cruz 101, 31–38.

endes, W.S., Silva, A.A.M., Trovão, J.R., Silva, A.R., Costa, J.M.L., 2002. Expansãoespacial da leishmaniose visceral Americana em São Luís do Maranhão. Brasil.Rev. Soc. Bras. Med. Trop. 35, 227–231.

estre, G.L.C., Fontes, C.J.F., 2007. A expansão da epidemia da leishmaniose vis-ceral no estado de Mato Grosso, 1998–2005. Rev. Soc. Bras. Med. Trop. 40,42–48.

inistério da Saúde do Brasil, 2006. Manual de Vigilância e Controle da LeishmanioseVisceral. Série A. Normas e Manuais Técnicos. Secretaria de Vigilância em Saúde.Departamento de Vigilância Epidemiológica, Brasília, DF, Brasil, Available at:http://portal.saude.gov.br/portal/arquivos/pdf/manual leish visceral2006.pdf(accessed 20.08.12).

inistério da Saúde do Brasil, 2011. Datasus: informac ões de saúde. Departamentode Vigilância Epidemiológica, Brasília, DF, Brasil, Available at: http://www2.datasus.gov.br/DATASUS/index.php?area=02 (accessed 04.10.11).

liveira, A.C., Abreu-Silva, A.L., Lima, T.B., Silva, A.P.C., Reis, L.F., Barbosa, D.S., Guerra,R.M.S.N.C., 2006. Soroprevalência da leishmaniose visceral canina no bairroJardim São Raimundo em São Luís. Rev. Soc. Bras. Med. Trop. 1, 140–143.

liveira, C.D.L., Assunc ão, R.M., Reis, I.A., Proietti, F.A., 2001. Spatial distribution ofhuman and canine visceral leishmaniasis in Belo Horizonte, Minas Gerais State,Brazil, 1994–1997. Cad. Saúde Pública 17, 1231–1239.

liveira, C.D.L., Morais, M.H.F., Machado-Coelho, G.L.L., 2008. Visceral leishman-iasis in large Brazilian cities: challenges for control. Cad. Saúde Pública 24,2953–2958.

refeitura Municipal de São Luís Secretaria Municipal de Saúde (SEMUS). Guia deDistritos. Available at: http://www.saoluis.ma.gov.br/semus/frmPagina.aspx?id pagina web=520 (accessed 20.08.12).

ieto, P., Malone, J.B., Bavia, M.E., 2006. Ecological niche modeling for visceralleishmaniasis in the state of Bahia, Brazil, using genetic algorithm for rule-set

prediction and growing degree day-water budget analysis. Geospat. Health 1,115–126.

odrigues, A.F., Escobar, A.L., Souza-Santos, R., 2008. Análise espacial e determinac ãode áreas para o controle da malária no Estado de Rondônia. Rev. Soc. Bras. Med.Trop. 41, 55–64.

ica 131 (2014) 56–62

Romero, G.A., Boelaert, M., 2010. Control of visceral leishmaniasis in Latin America– a systematic review. Plos Negl. Trop. Dis. 4, e584.

Savani, E.S.M.M., Schimonsky, B.V., Camargo, M.C.G.O., Dauria, S.R.N., 2003. Vigilân-cia de leishmaniose visceral americana em cães de área não endêmica, São Paulo.Rev. Saúde Pública 37, 260–262.

Silva, A.R., Tauil, P.L., Cavalcante, M.N.S., Medeiros, M.N., Pires, B.N., Gonc alves, E.G.R.,2008. Situac ão epidemiológica da leishmaniose visceral, na Ilha de São Luís,Estado do Maranhão. Rev. Soc. Bras. Med. Trop. 41, 358–364.

Silva, A.V.M., de Paula, A.A., Cabrera, M.A.A., Carreira, J.C.A., 2005. Leishman-iose em cães domésticos: aspectos epidemiológicos. Cad. Saúde Pública 21,324–328.

Silva, E.A., Andreotti, R., Honer, M.R., 2007. Comportamento de Lutzomyia longipalpis,vetor principal da lesihmaniose visceral Americana, em Campo Grande, Estadode Mato Grosso do Sul. Rev. Soc. Bras. Med. Trop. 40, 420–425.

Tavares, L.M.S.A., Tavares, E.D., 1999. Distribuic ão geográfica e aspectos ambientaisdas áreas endêmicas da leishmaniose visceral em Sergipe. Informe Epidemi-ológico do SUS 8, 47–52, Available at: http://scielo.iec.pa.gov.br/pdf/iesus/v8n1/v8n1a06.pdf (accessed 20.08.12).

Vigilato, M.A.N., (M.Sc. Dissertation) 2004. Distribuic ão Espacial da LeishmanioseVisceral Canina e Humana no Município de Birigui. Universidade EstadualPaulista, Botucatu, SP, Brazil.

Werneck, G.L., 2008. Forum: geographic spread and urbanization of visceral leish-maniasis in Brazil. Introduction. Cad. Saúde Pública 24, 2937–2940.

Werneck, G.L., Costa, C.H.N., Walker, A.M., David, J.R., Wand, M., Maguire, J.H., 2002.The urban spread of visceral leishmaniasis: clues from spatial analysis. Epidemi-ology 13, 364–367.

Werneck, G.L., Costa, C.H.N., Walker, A.M., David, J.R., Wand, M., Maguire, J.H., 2007.Multilevel modeling of the incidence of visceral leishmaniasis in Teresina. Brazil.

ing of the WHO Expert Committee on the Control of Leishmaniases. WHOTechnical Report Series no. 949, Geneva, Available at: http://apps.who.int/iris/bitstream/10665/44412/1/WHO TRS 949 eng.pdf (accessed 20.08.12).