Arquétipos latentes: os módulos dos padrões espaciais complexos do câncer

Abstract

Studies on the epidemiological behavior of several diseases have been the focus of many researchers who analyze incidence and/or mortality maps in search of geographic patterns capable of explaining the behavior of the disease in question. When analyzing regions with a higher rate, it is possible to draw conclusions that indicate, for example, that a certain disease is associated with high air pollution due to the high concentration of cases in industrial areas. The problem with this procedure is that diseases like cancer have different types, which makes their study more complicated because it is necessary to analyze several maps individually. Based on this, the objective of this work is to propose a method capable of reducing the amount of maps to be analyzed without the detriment of the analysis performed. The basic idea is to use the decomposition of the singular value to find latent maps that are able to explain the geographical patterns of different diseases. Before applying the decomposition theorem, Bayesian smoothing methods will be used and the rates will be obtained through the SMRi form to remove age-bias. In this work, we will analyze the cancer mortality data from Brazil and USA and show how this reduction of maps can be made, understood and explored. As a result, an efficiency of 90% is obtained in the reduction of the United States and of 63.33% in the Brazilian case.