Analysis of risk factors for both, all-cause and cause-specific mortality in developing countries is fundamental for the design of interventions and for monitoring their performance. An important resource for such analyses is the network of over 30 demographic surveillance systems (DSS) across the developing world. In the absence of useful national registries of vital statistics these sentinel systems monitor large populations within defined geographical areas, collecting longitudinal data on births, deaths, and migrations. People within DSS sites are associated with households which are geo-coded. Verbal autopsies are used to determine cause-specific mortality. Health service access, socio-economic and other household characteristics and risk factors are monitored via household surveys. DSS mortality data have spatial and temporal characteristics that make their analysis challenging. These data are collected at very large number of households (typically 12000-25000) repeatedly over time, and thus they are correlated in space and time. Common exposures (i.e., social, economic, climatic, and behavioural factors), whether measured or not, may influence mortality similarly in households of the same geographical area, introducing spatial correlation in mortality outcomes. Risk factor analysis and mapping, without taking into account this correlation, result usually in under-estimation of the precision of the parameter estimates and of the predictions. Geostatistical models which are appropriate for this type of spatial data are highly parameterised. Their full estimation was only possible over the last decade by formulating them within the Bayesian modelling framework and employing Markov chain Monte Carlo (MCMC) simulation for model fit. The complexity, however in fitting these models increases with the number of locations and for the very large number of households present within the DSS sites, model fit is not practical.The main objectives of the proposed project will be (i) to develop Bayesian statistical methodology for the analysis of very large, longitudinal, Gaussian and non-Gaussian geostatistical data, which will allow applications to large datasets with spatial and temporal correlations, and (ii) to validate and implement this methodology on data from DSS sites in Africa to answer specific problems that cannot adequately be addressed using other datasets or existing statistical methods. Specifically, we will (a) identify risk factors (demographic, socio-economic, nutritional and health service-related) and spatio-temporal patterns of infant, child and adult mortality within the Agincourt DSS area in South Africa; (b) map space-time patterns of malaria transmission measures within a malaria endemic DSS site in Rufiji, Tanzania and assess the relation between malaria transmission with all-cause and cause-specific infant and child mortality; and (c) analyse the relationships of mortality, at the above sites, with socio-economic status measured by household asset indices, in order to explore the feasibility of developing mortality-based asset indices for comparisons of health equity over time and between DSS sites.The proposed project will make the following contributions: (i) It will develop novel methodologies in the area of spatio-temporal modelling of very large geostatistical data that will enable rigorous statistical analyses of DSS sites data; (ii) It will provide a better understanding of factors related to infant and child mortality and of specific health system access issues with regards to specific mortality impacts; (iii) It will provide within-site estimates of spatio-temporal patterns of all-cause and cause-specific mortality. This is very important not only for planning appropriate interventions within a DSS site but for understanding cross-site differences taking into account within site variation in mortality and mortality related factors; (iv) It will improve our understanding on the relationships between mortality rates and malaria endemicity in Africa, which will again be of value in health planning since the estimates of how age specific mortality rates depend on the endemicity of malaria will make it possible to predict the consequences of reduction in malaria transmission of malaria control initiatives and (v) It will define socio-economic asset indices more comparable across sites that will help to better understand socio-economic differentials.