Vector-borne diseases are a world concern, representing 17% of all infectious dis- eases and more than 1 million deaths annually. In particular, dengue and malaria, trans- mitted by mosquitoes, are the most alarming vector-borne diseases because the former is the mosquito-borne viral disease that had the highest incidence growth in the last 50 years (30-fold) and the latter has the highest mortality incidence with an estimate of 627 thousand deaths in 2012. Predicting the incidence of these diseases is an important step in improving control programmes in order to prevent outbreaks with an efficient distribution of logistics and human resources to the affected zones within a reasonable time; however, the risk factors that determined the incidence are not fully understood. In order to deter- mine the main risk factors affecting malaria and dengue incidence in the Brazilian Amazon between 2006-2013, Bayesian hierarchical latent Gaussian models were used through the integrated nested Laplace approximation (INLA) inference approach. The area of study covers 310 municipalities of 6 Federative Units and the considered factors include climatic and socio-economic variables in space, time and space-time domains. It has been showed that the Poisson distribution is not adequate for the observed data suggesting the use of the Negative Binomial distribution. Then, the Besag-York-Molliè (BYM) model in the spatial and spatio-temporal scale outperformed the Negative Binomial generalized linear model thanks to the inclusion of unstructured and structured random effects. The main findings confirmed that the temperature, precipitation and the Oceanic Niño Index were highly associated with dengue risk and that the urbanization, measured through the pop- ulation density, is a main risk factor for both dengue and malaria. Finally, the discussion and disadvantages of the BYM model and the INLA inference approach along with possi- ble competitor models are presented.
Keywords:
Bayesian hierarquical modelling, dengue, i n l a , latent gaussian models, malaria, spatial models, spatio temporal models..
DOI
PDF