B0382
Title: Spatiotemporal modelling for multiple mosquito-borne diseases: A flexible Bayesian clustering approach
Authors: Jessica Pavani - Pontificia Universidad Catolica de Chile (Chile) [presenting]
Fernando Quintana - Pontificia Universidad Catolica de Chile (Chile)
Abstract: Disease mapping has become increasingly important in public health analysis. In this context, the data are typically collected for specific regions over time and modelled using parametric spatiotemporal techniques. As an alternative contribution to the literature on multivariate disease, a flexible model is developed to identify and cluster areas where multiple diseases behave similarly. To do so, a spatiotemporal model is established where temporal dependence is defined for areal clusters induced by product partition models (PPM). Unlike similar methods, PPM produces more flexible clusters, even allowing them to be non-contiguous. To model the temporal component, a structure that considers lagged values of observed data is defined, including a seasonal effect. The model also considers a multivariate directed acyclic graph autoregressive structure to accommodate spatial and inter-disease dependence, which allows the interpretation of a spatial correlation parameter. As an illustration, the proposed modelling is first tested using simulation studies, then it is applied to a real dataset. For this application, the number of cases of two tropical diseases is considered, dengue and chikungunya, transmitted by the same mosquito, for all 145 microregions in Southeast Brazil from 2018 to 2022.