SciELO - Scientific Electronic Library Online

 
vol.29 número2Distribución de la sífilis congénita en el estado de Tocantins, Brasil, 2007-2015 índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados

Revista

Articulo

Indicadores

  • No hay articulos citadosCitado por SciELO

Links relacionados

  • No hay articulos similaresSimilares en SciELO

Compartir


Epidemiologia e Serviços de Saúde

versión impresa ISSN 1679-4974versión On-line ISSN 2237-9622

Epidemiol. Serv. Saúde vol.29 no.2 Brasília  2020  Epub 17-Abr-2020

http://dx.doi.org/10.5123/s1679-49742020000200004 

Health Surveillance Management Tools

Automatic stratification of priority areas for Dengue control using the QGIS Model Builder in multicriteria analysis

Edmar Franco de Paiva Júnior (orcid: 0000-0002-0836-3151)1  , Tânia da Silva Vaz (orcid: 0000-0001-8532-0668)1  , Marcello Rosa (orcid: 0000-0003-3019-7628)1  , Ivana Lúcia Borges e Garcia (orcid: 0000-0003-2396-1149)1 

1Secretaria de Estado da Saúde de Goiás, Superintendência de Vigilância em Saúde, Goiânia GO, Brazil

Abstract

Objective:

to present a methodological sequence resulting from multicriteria analysis indicating areas with different Aedes aegypti intervention priorities.

Methods:

a Female Aedes Displacement Index (IDFAedes) was created, consolidated according to urban blocks, representing interaction between population densities, Aedes aegypti oviposition sites and dengue case notifications; a graphical model (Model Builder) was developed with QGIS software using the Kernel mapping algorithm and IDFAedes as the weighting factor.

Results:

stratification for the evaluated example - Anápolis, GO, Brazil - indicated intervention priority levels for urban blocks - 17.5%, very low priority; 37.3%, low; 33.6%, medium; 10.2%, high; 1.4%, very high -; blocks with medium, high, and very high priority accounted for 22.53,% of the territory in the area.

Conclusion:

the spatial block method proposed in this article can be included in health surveillance programs for intensified targeting and planning of control actions.

Keywords: Spatial Analysis; Vector Control; Dengue

Introduction

Brazilian public practices for addressing Aedes aegypti have not achieved a satisfactory level of control. A literature review indicated dengue epidemic peaks in Brazil in 2002, 2008 and 2010.1 Data retrieved from the Notifiable Health Conditions Information System (SINAN)2 show that in 2013, 2015 and 2016 the threshold of 1.5 million cases nationwide was exceeded, resulting in incidence levels greatly above those considered to be risk levels by the World Health Organization (WHO), namely 300 cases/100,000 inhabitants.

Repeated epidemics and marked seasonality require specific action strategies. Notwithstanding, much of the lack of success of these actions arises from the methodology involving universal and indiscriminate coverage of buildings, based on cycles, with the aim of controlling the immature form of the Aedes aegypti vector. Surveillance programs are completed by using the ‘notified dengue cases’ marker for adopting measures to combat the winged form of the vector. However, reality has shown that even when coverage targets are met, municipal environmental surveillance teams frequently find themselves overwhelmed by dengue transmission. As such, changes to current strategies are clearly needed and surveillance and control actions need to be enhanced.3

In view of the weaknesses of the current dengue combat model, studies aimed at defining strategic intervention areas have gained importance, using cluster analysis techniques from a spatial perspective.4-8 The common stem of these studies consists of categorizing areas with known risk factors and identifying clusters, in order to target intensified control actions.

Structuring the planning of efficient health surveillance public policies is fundamental for compiling and analyzing a large amount of reliable and up to date information. Within this context, evaluation routines can be managed and automated in a competent manner using geoprocessing techniques.

By using Geographic Information System (GIS) software, the objective of this study was to present a practical and direct methodological sequence resulting from multicriteria analysis for automatic indication of priority areas for intervention and planning actions to combat Aedes aegypti.

Methods

When using multicriteria analysis to establish priority areas, factors were used that have already been scientifically explored and which interfere,9 decisively, with biological behavior and with the prevalence of the female of the Aedes aegypti species in certain locations. These factors are population density and presence of oviposition sites. Dengue notification geoespatialization was also used to express the convergence of situations that lead to the occurrence of the disease.

An auxiliary index was created, referred to as the Female Aedes Displacement Index (IDFAedes). This was based on the premise that areas with higher density of hosts and higher records of oviposition sites and dengue case notifications are most propitious to least displacement of Aedes aegypti females. Consequently, these regions are more vulnerable to dengue transmission, given that they are open to greater biological prevalence of the vector. A similar understanding was obtained in mark-release-recapture experiments in different urban arrangements in the state of Rio de Janeiro.10

Urban blocks were used as a primary unit of data inclusion and analysis. The urban block shapefile was imported from the State of Goiás Zero Aedes Monitoring System (SIMAZ).11 When creating IDFAedes, population density, history of Aedes aegypti breeding sites and dengue notifications were combined. The parameter values were grouped into quintiles and individually given scores ranging from 1 to 5. Each block was represented by the sum of its individual scores, totaling at least 3 and 15 at the most. The ranges were defined with the following segmentation: 3 to < 6 (very low priority), 6 to < 8 (low priority), 8 to <10 (medium priority), 10 to <12 (high priority) and above 12 (very high priority), as shown in Table 1. The levels of priority for intervention were divided into quintiles, according to the sum of the individual scores (Table 1). The analyses were performed using SIG QGIS 2.14 Essen and Excel® 2013. The database of the municipality of Anápolis, located in the state of Goiás (GO), for the years 2016 and 2017, was used to exemplify the priority area automatic selection model.

Table 1 - Matrix of the composition of the Female Aedes Displacement Index, for intervention and planning actions against the Aedes aegypti vector 

Percentile class Parameters IDFAedes a
Population density History of vector breeding sites History of dengue case notification
Score Range Intervention priority
quintile 1 5 5 5 >12 Very high
quintile 2 4 4 4 10 to 12 High
quintile 3 3 3 3 8 to <10 Medium
quintile 4 2 2 2 6 to <8 Low
quintile 5 1 1 1 3 to <6 Very low

a) IDFAedes: Female Aedes Displacement Index.

Population information, by urban block, was obtained from the Ministry of Health’s Primary Care Information System. This information is generated by Family Health Strategy teams and community health agents. History of Aedes aegypti breeding sites, also by urban block, was imported from the SIMAZ program. Dengue notifications were retrieved from the SINAN system and tabulated using Excel®. The QGIS Web Service Geocode algorithm was used to geocode 8,737 records of addresses of dengue notifications. In this stage, owing to erratic positions produced by the algorithm, some manual corrections and data removal were necessary (manual geocoding) and to this end support was provided by the municipal epidemiological surveillance team. Taking the ‘count points in polygon’ native algorithm, the sum of dengue notifications per georeferenced urban block was consolidated.

Excel® was used to calculate mean breeding sites between 2016 and 2017, segregated into two annual periods: October to March and May to September, which are the rainy season and the dry season in Goiás, respectively. In order to create IDFAedes categories, each individual parameter was divided into quintiles, giving a score of 1 for least female Aedes aegypti displacement, reaching a score of 5 for greatest displacement. The sum of the scores of the individual parameters represented the IDFAedes per urban block, whereby the IDFAedes was greater when the trend of the female Aedes aegypti to displace itself in search of conditions favorable to its survival was greater. Ultimately, supported by a composite index, the IDFAedes seeks to express the different levels of interaction between the parameters used, consolidated according to urban blocks.

A model was developed to be run on QGIS, along with the GRASS, GDAL/OGR and SAGA tools (Figure 1). This model is comprised of a series of computing routines, logically structured to incorporate, process and provide data. Basically, it is built on the territorial base and associated tabular data. These data are processed and prepared in order to run the ‘kernel density curve’ algorithm. Finally, class and style formatting is done, thus making the result more user-friendly. The detailed sequencing of the actions and the description of the algorithms and products obtained are listed in Table 2.

Figure 1 - Model built using the QGIS Model Builder 

Table 2 - Action sequencing, algorithm description and products obtained 

Stage Description Algorithm/library used Product obtained
1 Load municipal urban block layer -/GDALb Polygon vector layer added to project
2 Load previously formatted table, indicating IDFAedes a per urban block -/GDALb Table added to project
3 Table union to block layer Union/GDAL b Polygon vector layer linked to attribute table
4 Generate block layer centroid Polygon centroids/GDALb Point vector layer linked polygon layer
5 Reproject centroid layer to UTM projection appropriate for the region (DATUM 31982) Reproject Layer/GRASSc Centroid layer with SRC 31982
6 Enveloping polygon on the block layer Polygon from layer extent/GRAS c Enveloping polygon created
7 Reproject enveloping polygon to UTM project appropriate for the region (DATUM 31982) Reproject Layer/GRASSc Centroid layer with SRC 31982
8 Boundaries of the enveloping polygon vector layer Vector layer boundaries/GRASSc Spatial delimitation of the reprojected enveloping polygon coordinates
9 Multi-weighted heatmap using the weight of the IDFAedes column and data of the delimitation produced in stage 8 Heatmap/SAGAd Raster layer created
10 Migration of ‘Mean’ information on raster values consolidated per urban block Zonal Statistical/GDALb Creation of ‘mean’ column on the block layer
11 Style applied to the layer produced in stage 10, categorizing 5 classes of the ‘mean’ column in equal intervals Set style for vector layer/GDALb Blocks stratified in 5 categories

a) IDFAedes: Female Aedes Displacement Index .

b) GDAL: Geospatial Data Abstraction Library .

c) GRASS: Geographic Resources Analysis Support System.

d) SAGA: System for Automated Geoscientific Analyses.

Results

The direct result of applying the graphical model that was built can be seen in Figure 2. Stratification indicated the following priority levels for intervention in the urban blocks: 17.5% very low priority; 37.3% low priority; 33.6% medium priority; 10.2% high priority; and 1.4% very high priority. The priority classes were shown in a cluster pattern, distributed over all regions of the municipality.

Figure 2 - Product obtained by applying the graphical model of automatic selection of priority areas for intervention and planning actions against the Aedes aegypti vector, Anápolis, Goiás, 2016-2017 

The technique used distinguished microregions with different dengue transmission predispositions. The classes with the highest priority for intervention had the highest values, among all the parameters used. According to Table 3, breeding site density (Aedes aegypti breeding sites per hectare [ha.], where 1ha. = 10,000 m2]) increased 1600% between the very low priority class and the very high priority class. Population density and dengue case density also showed grading compatible with the level of priority proposed: variations of 13.2 inhab./ha. to 203.6 inhab./ha., and 0.03 dengue cases/ha to 3.96 dengue cases/ha., respectively.

Table 3 - Attributes of the priority areas for intervention and planning actions against the Aedes aegypti vector, selected for the annual dry period (May-September), Anápolis, Goiás, 2016-2017 

Parameters Intervention priority categories
Very low Low Medium High Very high
Quantity (number of blocks) 1,341 2,848 2,569 778 106
Area (ha.) 5,713 2,875 1,986 471 40
Breeding site density (number/ha.) 0.003 0.020 0.029 0.055 0.051
Population density (inhab./ha.) 13.2 89.4 128.6 144.8 203.6
Dengue case density (number/ha.) 0.03 0.42 1.67 2.56 3.96
Incidence (cases/100,000 inhab.) 359.9 749.9 2,045.5 2,785.5 3,054.6

Discussion

The model presented assumes accurate geocoding of the parameters used. Dengue breeding sites are input on SIMAZ, enabling geographic correspondence at the urban block level. However, dengue notifications can have certain conversion problems. The algorithm used has limitations with regard to interpretation of hierarchical levels, depending on input address variation. This fact is most serious for small cities where, generally, the addressing system is incomplete and/or out of date.

Detailed exploration of address geocoding nuances is beyond the objectives of this study, this being why the study stuck to commonly used systems (Web Service Geocode and Google Maps baseline). Notwithstanding, automatic geocoding error correction is possible, including correction of non-standardized input and correction of the geographic certainty indicator.12

It should be noted that address geocoding difficulties could be easily overcome on the national level, if the information input system (SINAN) migrated to or synchronized with the form of urban addressing for geographic addresses. Geocoding problems on the SINAN database may affect up to 16% of data, owing to database shortcomings.13 On the state level, given the pre-existence of the SIMAZ system, the solution would be to integrate the notifying sector with this geographic information system.

The spatial block method proposed for this study differs from usual aggregation levels. They generally use political/administrative divisions (neighborhood, census tract, health district, municipality). Use of this latter method persists owing to inertia, tradition or even because of indexation with current information systems. The processes, both environmental and social, that promote or restrict health risk situations are not limited to political/administrative boundaries.4 It should be emphasized that the evolution of spatial analysis aggregation tools enables good interpretation of the environment-disease system, and breaking away from such divisions is fundamental for enhancing understanding of the modulation of this process.

Table 3 shows information segregated according to the level of stratification proposed in this study. It is important to emphasize that the areas taken to be a higher priority for intervention (very high and high) represented 11.6% of the blocks in the territory and just 4.61% of the extent of the urban area. However, in these clusters there was more intense convergence of interaction of the parameters that make them more vulnerable and, consequently, make them a priority as well. This interpretation of the territory represents an alternative for targeted intensification, which is promising for the model currently in force (universal and indistinct coverage).

The spatial block method used in this study is in accordance with the premises of the heterogeneous model of dengue transmission, according to which the probability of the vector contaminating a host differs between the different regions of the municipality.14 This difference emanates from the population structure, and the graphical model proposed seeks to make this distinction.

Based on analysis of secondary data and application of the k-means and the kernel density curve methods for evaluating clusters, an instrument was proposed for planning Aedes aegypti control actions in Niterói, RJ,4 and Natal, RN,7 respectively.

Using kernel density curves in the multicriteria evaluation provided territorial stratification applied to health surveillance with neighborhood analysis and its influence on the production of risk factors in the different parts of the municipality. This is an interpretative gain, to the extent that each cell of the grid pre-established in the input algorithm receives a score resulting from the weighted influence of the factors evaluated in the neighborhood (dengue cases, population density and presence of Aedes aegypti breeding sites).

Neighborhood analysis techniques and spatial dependence in the dynamics of dengue transmission have been investigated. A study conducted in the year 2006 in 157 of the 160 neighborhoods existing in the municipality of Rio de Janeiro, RJ, evaluated the spatial correlation between different indicators: the Gini index, the rainfall index and the Breteau index, and total dengue cases. Positive spatial correlation was found for all cases indicated by the global Moran’s index in a temporal / spatial sample.14 This signifies that there was a general spatial dependence pattern in the distribution of these indicators, i.e. adjacent neighborhoods had greater similarity than neighborhoods distant from each other.13 It must be emphasized that positive spatial correlation of dengue notifications, as demonstrated by the authors of the cited article, as well as of the other parameters assessed, reveals the importance of including neighborhood analysis in area stratification.

Various different control protocols (‘National Guidelines for Dengue Epidemic Prevention and Control’, at federal level; and ‘Goiás against Aedes’, at state level) adopt household visits in 100% of the urban grid with the same grading. In this work model, regardless of the social, environmental, entomological and epidemiological profile of the intervention areas, they receive the same treatment. The product proposed in this article (stratification in risk categories) can be included as a parameter for intensifying control actions in given areas, or for altering current intervention dynamics.

This article presented a proposal for parameters and for interaction between them, when building the matrix that informed the base table for the spatial areas considered. Notwithstanding, other criteria can be used but this in-depth discussion does not fall within the overriding objectives of the study in question, i.e., to present and discuss the results of an automatic area selection mechanism using GIS software.

We suggest that other area stratification methods be tested, such as including secondary data,7-8 especially socio-economic data.8 Other forms of series sequencing (percentile) can be tested, such as the use of natural breaks, standard deviation or equal interval. Other temporal clusters than the cluster used (2016 and 2017 dry season) can also be assessed. It should be emphasized that, regardless of the matrix that is built, the resulting table should be included in the graphic model proposed.

Including other entomological attributes, especially adult Aedes aegypti indices, will enable the model’s sensitivity to be refined. Indicators such as the human development index (HDI) and the Gini index, schooling and income brackets, and degree of sanitation, can also be tested and validated. Moreover, it is of fundamental importance, to the extent that control actions are based on the stratification model proposed here, to modulate responses in stages,7 in addition to mechanisms for evaluating the effectiveness and planning of feedback on the parameters used.

A graphical model was prepared using GIS open source software (QGIS 2.14 Essen), which involves simple inputs (table in .csv format, with usual current program parameters and shapefile of any inframunicipal geographic database). Despite the analytical complexity it includes (kernel density curves), the product of the model is intuitive and has a ramp user interface, with different colorimetric intensities. These are characteristics that make it easy for the model to integrate with current health information systems and, as such, they can assist with the generation of new health surveillance work routines.

Referências

1. Bohm AW, Costa CS, Neves RG, Flores TR, Nunes BP. Tendência da incidência de dengue no Brasil, 2002-2012. Epidemiol Serv Saúde [Internet]. 2016 out-dez [citado 2020 fev 7];25(4):725-33. Disponível em: Disponível em: https://doi.org/10.5123/s1679-49742016000400006Links ]

2. Ministério da Saúde (BR). Secretaria de Vigilância à Saúde. Sistema de informação de agravos de notificação - Sinan: normas e rotinas. Brasília: Ministério da Saúde; 2002. [ Links ]

3. Pessanha JEM, Caiaffa WT, César CC, Proietti FA. Avaliação do plano nacional de controle da dengue. Cad Saúde Pública [Internet]. 2009 jul [citado 2020 fev 7];25(7):1637-41. Disponível em: Disponível em: https://doi.org/10.1590/S0102-311X2009000700024Links ]

4. Resendes APC, Silveira NAPR, Sabroza PC, Souza-Santos R. Determinação de áreas prioritárias para ações de controle da dengue. Rev Saúde Pública [Internet]. 2010 abr [citado 2020 fev 7];44(2):274-82. [ Links ]

5. Cavalcante MPR, Oliveira C, Simão FB, Lima PR, Monteiro PS. Análise geoespacial: um estudo sobre a dengue. Acta Paul Enferm [Internet]. 2013 [citado 2020 fev 7];26(4):360-8. Disponível em: Disponível em: https://doi.org/10.1590/S0103-21002013000400010Links ]

6. Defavari ER, Fonseca EP, Silva RP, Silveira Moreira R, Pereira AC, Batista MJ. Análise espacial da incidência da dengue em um município de médio porte do estado de São Paulo de 2008 a 2015. Rev Saúde Coletiva UEFS [Internet]. 2017 [citado 2020 fev 7];7(3):10-7. Disponível em: Disponível em: http://periodicos.uefs.br/index.php/saudecoletiva/article/view/2560/2175 . doi: 10.13102/rscdauefs.v7i3.2560 [ Links ]

7. Barbosa IR, Tavares ADM, Torres UPDS, Nascimento CAD, Moura MCBM, et al. Identificação de áreas prioritárias para a vigilância e controle de dengue e outras arboviroses transmitidas pelo Aedes aegypti no município de Natal-RN: relato de experiência. Epidemiol Serv Saúde [Internet]. 2017 jul-set [citado 2020 fev 7];26(3);629-38. Disponível em: Disponível em: https://doi.org/10.5123/s1679-49742017000300020Links ]

8. Araújo JRD, Ferreira EF, Abreu MHNG. Revisão sistemática sobre estudos de espacialização da dengue no Brasil. Rev Bras Epidemiol [Internet]. 2008 dez [citado 2020 fev 7];11(4);696-708. Disponível em: Disponível em: https://doi.org/10.1590/S1415-790X2008000400016Links ]

9. Zara ALSA, Santos SM, Fernandes-Oliveira ES, Carvalho RG, Coelho GE. Estratégias de controle do Aedes aegypti: uma revisão. Epidemiol Serv Saúde [Internet]. 2016 [citado 2020 fev 7];25(2);391-404. Disponível em: Disponível em: https://doi.org/10.5123/s1679-49742016000200017Links ]

10. Maciel-de-Freitas R, Lourenço-de-Oliveira R. Presumed unconstrained dispersal of Aedes aegypti in the city of Rio de Janeiro, Brazil. Rev Saúde Pública [Internet]. 2009 [citado 2020 fev7];43(1):8-12. Disponível em: Disponível em: https://doi.org/10.1590/S0034-89102009000100002Links ]

11. Secretaria de Estado da Saúde de Goiás. Sistema de monitoramento Aedes Zero -SIMAZ [Internet]. Goiânia: Secretaria de Estado da Saúde de Goiás; 201- [citado 2018 dez]. Disponível em: Disponível em: https://extranet.saude.go.gov.br/sacd/EstatisticaQuadrasVisitadas.jsfLinks ]

12. Martins D, Davis Jr CA, Fonseca FT. Geocodificação de endereços urbanos com indicação de qualidade [Internet]. Proceedings XIII GEOINFO; 2012 nov 25-27; Campos do Jordão (SP). Campos do Jordão (SP): MCTI/INPE; 2012 [citado 2020 fev 7]. p. 36-41. Disponível em: Disponível em: http://www.geoinfo.info/proceedings_geoinfo2012.split/proceedings_geoinfo2012.43-48.pdfLinks ]

13. Galli B, Chiaravalloti Neto F. Temporal-spatial risk model to identify areas at high-risk for occurrenceof dengue fever. Rev Saúde Pública [Internet]. 2008 Aug [cited 2020 Feb 7];42(4);656-63. Available from: Available from: https://doi.org/10.1590/S0034-89102008005000032Links ]

14. Teixeira TRA, Cruz OG. Spatial modeling of dengue and socio-environmental indicators in the city of Rio de Janeiro, Brazil. Cad Saúde Pública [Internet]. 2011 Mar [cited 2020 Feb 7];27(3):591-602. Available from: Available from: https://doi.org/10.1590/S0102-311X2011000300019Links ]

Received: April 01, 2019; Accepted: November 15, 2019

Correspondence: Edmar Franco de Paiva Júnior - Avenida 136, Quadra F44, Lotes 22 e 24, Ed. Cesar Sebba, Setor Sul, Goiânia, GO, Brazil. Postcode: 74093250 E-mail: edmar.junior@saude.go.gov.br

Authors’ contributions

Paiva-Júnior EF contributed to the conception and design of the study, data analysis and interpretation and writing the manuscript. Vaz TS, Rosa M and Garcia ILB contributed to data analysis and interpretation and relevant critical revision of the intellectual content of the manuscript. All the authors have approved the final version of the manuscript and are responsible for all aspects of the work, including guaranteeing its accuracy and integrity.

Associate Editor: Lúcia Rolim Santana de Freitas - orcid.org/0000-0003-0080-2858

Creative Commons License Este é um artigo publicado em acesso aberto sob uma licença Creative Commons