Estudo sobre a mortalidade adulta, para Brasil entre 1980 e 2000 e unidades da Federação em 2000: uma aplicação dos métodos de distribuição de mortes

Abstract

The study of trends and levels of adult and elderly mortality in developing countries is hampered by a lack of complete vital registration data. Even when coverage is adequate, information regarding age is often inaccurate. Census data, necessary to estimate denominators in standard mortality rates, are also of uncertain quality (United Nations, 1983; Preston, Elo, Stewart, 1999). In recent decades, different methods known as Death Distribution Methods, based on formal demographic equations, were developed to study quality of data and adult mortality.The purpose of this dissertation is to evaluate the quality of population and deaths data, and obtain estimates of mortality for the adult population, in Brazil, between 1980 and 2000, and for states in 2000. In common, these methods estimate death coverage, in relation to the age distribution of the population, adjusting the level and maintaining the structure of mortality constant (Brass, 1975, Preston et al., 1980, Bennett and Horiuchi, 1981; Hill, 1987; Hill and Choi, 2004; Hill and Queiroz, 2004). The use of different techniques is important because literature does not indicate any gold standard to study adult mortality data problems. Furthermore, by applying and evaluating different methods we expect obtain more robust estimates of mortality, since using a set of estimates, we are able to produce an expected range of estimates. The application of alternative methods allows one to understand more about the level and trend of mortality in Brazil, and is also an interesting methodological exercise.We obtained different results based on the different methods and their assumptions. In methods that relaxed the assumption of stable population, the results indicated that the coverage of death registration was more than 100% in the states of Southeast, South and Center-West, indicating the existence of errors in data sources that generated overestimate results. Thus, measures of mortality was calculated using the coverage estimated by the methods of Growth Balance Equation (Brass, 1975) and Preston et al. (1980), which have the assumption of stable population. Given the difficulty in defining what is the most appropriate method and which one best fits the data, estimates for these two methods was interpreted as upper and lower bounds of results.