Spatial Patterns of Childhood Mortality and Morbidity In Sub-Saharan Africa: A Bayesian Geo-Additive Multinomial Models Approach

Abstract

Background: In epidemiological studies, several diseases share common risk factors. The jointly modelling of the risks of multiple diseases can provide the epidemiologists and health practitioners the etiological patterns of the incidence (mortality rates) across ecological areas. Methods: This paper investigates the differences in small scale geographical variations and the risk factors on child’s health outcomes (infant mortality) and co-infections (diarrhea, fever, cough and low birth weight) in the African sub-regions. The cross sectional data was obtained from Demographic and Health Surveys from Nigeria and Tanzania. We model spatial heterogeneity within the sample population using a flexible structured geo-additive regression model. The inference was based on the Bayesian MCMC simulation technique. Results: The results indicated that the proportion of low birth weight mortality deaths was found to 43.4% for Nigeria and 31.4% in Tanzania. The overall prevalence were: diarrhea (9.4%), fever (35.5%) and cough (13.8%) in Nigeria; and diarrhea (10.9%), fever (16.3%) and cough (14.9%) for Tanzania. We estimated correlation between the diseases to evaluate the `share risks' in the co-morbidity across the geographic areas. The multivariate analysis revealed that the risk factors such like non-antenatal attendance, multiple birth, short birth intervals, low maternal education, and poor sanitation were associated with infant mortality and childhood morbidity. In addition to the statistical relevance, we produce predictive spatial maps that detect high risk regions, “hot spots" which can assist developing partners and government to channel scarce health resources in an effective manner. Conclusion: The findings can guide in evidence-based allocation of scarce health resources in the sub-region with the aim of improving the chance of child survival.