Spatial point analysis based on dengue surveys at household level in central Brazil
1 Institute of Tropical Pathology and Public Health, Federal University of Goias, Department of Collective Health, Goias, Brazil
2 Oswaldo Cruz Foundation, DIS/CICT, Rio de Janeiro, Brazil
3 Oswaldo Cruz Foundation, Centro de Pesquisas Aggeu Magalhães, Pernambuco, Brazil
4 Scientific Computation Program, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil
BMC Public Health 2008, 8:361 doi:10.1186/1471-2458-8-361Published: 20 October 2008
Dengue virus (DENV) affects nonimunne human populations in tropical and subtropical regions. In the Americas, dengue has drastically increased in the last two decades and Brazil is considered one of the most affected countries. The high frequency of asymptomatic infection makes difficult to estimate prevalence of infection using registered cases and to locate high risk intra-urban area at population level. The goal of this spatial point analysis was to identify potential high-risk intra-urban areas of dengue, using data collected at household level from surveys.
Two household surveys took place in the city of Goiania (~1.1 million population), Central Brazil in the year 2001 and 2002. First survey screened 1,586 asymptomatic individuals older than 5 years of age. Second survey 2,906 asymptomatic volunteers, same age-groups, were selected by multistage sampling (census tracts; blocks; households) using available digital maps. Sera from participants were tested by dengue virus-specific IgM/IgG by EIA. A Generalized Additive Model (GAM) was used to detect the spatial varying risk over the region. Initially without any fixed covariates, to depict the overall risk map, followed by a model including the main covariates and the year, where the resulting maps show the risk associated with living place, controlled for the individual risk factors. This method has the advantage to generate smoothed risk factors maps, adjusted by socio-demographic covariates.
The prevalence of antibody against dengue infection was 37.3% (95%CI [35.5–39.1]) in the year 2002; 7.8% increase in one-year interval. The spatial variation in risk of dengue infection significantly changed when comparing 2001 with 2002, (ORadjusted = 1.35; p < 0.001), while controlling for potential confounders using GAM model. Also increasing age and low education levels were associated with dengue infection.
This study showed spatial heterogeneity in the risk areas of dengue when using a spatial multivariate approach in a short time interval. Data from household surveys pointed out that low prevalence areas in 2001 surveys shifted to high-risk area in consecutive year. This mapping of dengue risks should give insights for control interventions in urban areas.