Content area
Abstract
This thesis presents an analysis of the relationship among community attributes, individual and household characteristics, and child survival in Brazil. The study is based on a multilevel proportional hazards model for data that are correlated at two hierarchical levels. The model provides corrected parameter estimates and standard errors with typical data sets from developing countries, such as the 1986 DHS survey of Brazil used in this thesis. By design, these data sets are clustered by family and by community.
Our analysis begins with an investigation of how the social and environmental context in which a child is raised affects his or her survival chances, and we study how individual and household characteristics serve to modify the effects of community characteristics. Subsequently, we examine the demographic and reproductive pathways through which maternal education and household income operate to affect child mortality risks. We also provide an analysis and interpretation of community frailty-effects in models that include no community-level covariates. The ultimate goal of the thesis is to understand the role played by individual, household and community characteristics in shaping child mortality differentials in the context of a developing country.
Our results from the study of community characteristics indicate that these are important covariates of child mortality risks. Analysis of interaction effects provided us with a deeper understanding of the effect of community characteristics on child survival chances by illuminating the likely pathways through which these variables operate. This information is useful for predicting who is most likely to benefit from public policies to improve community infrastructure, education and health care. This study has also contributed to our understanding of the factors responsible for the large regional differential in child mortality between the Northeast region and other parts of Brazil. The analysis of community frailty-effects in Northeast Brazil shows the importance of community clustering in studies of the proximate determinants of child mortality that do not include community covariates. In addition, it provides a more accurate estimate of the magnitude and significance of unmeasured family effects.
