Modelagem espacial dos dados de Fasciola Hepática Bovina no Sul do Espírito Santo

Abstract

The fascioliasis is a liver disease caused by the parasite Fasciola, trematoda that parasite liver and biliary tract of cattle and sheep, but also diagnosed in goats, horses, buffalo, human and wild animals such as capybara and eagle rays bathed . Currently, fascioliasis has become a serious economic problem for producers from southern state of Espirito Santo and possibly for the entire land mass of the state, in view of the possibility of contamination of cattle herds, sheep, goats and buffaloes for shipping of infected animals and the intermediate host presence in several river basins in the region. The study area was represented by 23 municipalities in the southern state of Espirito Santo. The unit of analysis were property whose economic goal bovine livestock, totaling 115 properties, five in each municipality. The epidemiological survey in these properties was made during the period between 2009 and 2011. The objectives of this study are: to determine the geographical distribution of the fluke in the southern state of Espirito Santo checking for clusters, analyze the possible factors that may be associated with pathology (risk factors) and to propose and discuss some techniques of mathematical modeling based on heuristics to detect spatial clusters in the affected properties with liver fluke. It was found that the average prevalence of the fluke properties studied was 19.52% [13.41%, 27.35%] and the variation coefficient was estimated at 8.24%. It was verified by the kernel statistics of occurrence of the highest intensity in the central fluke is the area under study. By calculating the Moran's correlation coefficient between the municipalities of the study area, the value of 0.443 (p-value <0.001) was observed indicating a significant spatial autocorrelation between the areas studied, ensuring the formation of geographical clusters. After analysis of the models and the literature review, it can be mean that the GAM multivariate logistic model is the most parsimonious model for the study of fluke and its risk factors. This way you can see that variables other hosts and previous cases are characterized as epidemiological risk factors for fluke. For detection of geographic clusters were used two approachs, k-means with elbow and genetic algorithm with silhouette function, both were significant but to this data the second proved more accurate than the first, returning the value of 5 (five) clusters and showing the cluster with the highest concentration is located in the center of the study area. This result validates that was found in the kernel statistics. Thus we conclude that the properties belonging to the cluster 1 (one) require priority attention.