Estimativas bayesianas da taxa de letalidade do infarto agudo do miocárdio

• Ano de defesa: 2013
• Autor(a) principal: Fernanda Rodrigues Vargas
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Epidemiologia espacial; MCMC; INLA; Estatística espacial; Mapeamento de doenças; Estatistica
• Link de acesso: http://hdl.handle.net/1843/BUOS-98KHRK

The analysis of the geographical distribution of the incidence of a disease and its relationship with risk factors are sources of relevant information in epidemiological studies and public health. They suggest hypotheses that lead to investigation and to monitoring of the possible causes of the disease. In disease mapping, Bayesian methods are widely used, especially by the possibility of adopting a hierarchical structure for modeling data. The purpose of this dissertation is to estimate the mortality for acute myocardial infarction (AMI) in the Brazilian microregions. Since infarcts are few in small microregions, lethality ends up being poorly estimated in these cases. Thus, we study together another phenomenon associated with mortality, hospitalization for AMI. This phenomenon is more common and this allows the Bayesian model to borrow strength to better estimate the lethality. We analyze jointly the data of hospitalization and mortality due to AMI from Brazilian microregions. We used the shared component model proposed by Knorr-Held e Best (2001) specied by a new formulation here proposed, assessed and validated through a simulation study. We used the INLA (Integrated Nested Laplace Approximations ) method to estimate the model. The smoothed estimates for mortality decreased the effect of random uctuations not associated with risk, making visible the microregions with estimates above the national average. Furthermore, we found that the relationship between mortality and hospitalization rate is nonlinear, having a variability dependency between lethality and the logarithm of the admission rate.