Spatial prediction of canine visceral leishmaniasis in an endemic urban area of Brazil

Abstract

Canine visceral leishmaniasis (CVL) is a widespread zoonotic disease in Brazil. This study aimed to identify and predict spatial patterns of CVL in an endemic city, Votuporanga, and examine disease associations with key environmental and anthropogenic factors at a fine spatial scale. First, we estimated the spatial clustering of CVL cases relative to non-cases from 8,146 dogs. Second, we assessed CVL density using a Kernel density ratio map. Third, we analyzed associations between disease occurrence and selected variables derived from the Normalized Difference Vegetation Index (NDVI), number of buildings, building area, and street density using binary logistic regression models. Finally, we predicted the spatial odds of CVL using a Generalized Additive Model (GAM) that incorporated the significant covariates. Our results revealed significant clustering of cases up to a range of 1.7 km. Mean NDVI, street density, and sparse vegetation were statistically significant, increasing the odds of CVL by 431%, 109%, and 100%, respectively, per unit change. The predictive performance of the GAM, evaluated through cross-validation, indicated that the model incorporating mean NDVI achieved the best fit, with an area under the receiver operating characteristic (ROC) curve of 0.74 (CI 0.72–0.76). Our findings demonstrate that CVL is widespread across the city, predominantly in urban fringe areas, with nearly 45% of the city classified as having increased odds of CVL (>1). In contrast, the downtown area exhibited lower odds of disease. Furthermore, we identified distinct parasite genotypes across the city, primarily in areas with higher disease odds. Altogether, our results highlight how biological and environmental data can be integrated into mapping to enhance the understanding of the spatial dynamics of disease transmission in urban areas.