Bayesian spatial analysis of childhood diseases in Zimbabwe.

Rodney Godfrey Tsiko

doi:10.1186/s12889-015-2182-7

Abstract

BackgroundMany sub-Saharan countries are confronted with persistently high levels of childhood morbidity and mortality because of the impact of a range of demographic, biological and social factors or situational events that directly precipitate ill health. In particular, under-five morbidity and mortality have increased in recent decades due to childhood diarrhoea, cough and fever. Understanding the geographic distribution of such diseases and their relationships to potential risk factors can be invaluable for cost effective intervention.MethodsBayesian semi-parametric regression models were used to quantify the spatial risk of childhood diarrhoea, fever and cough, as well as associations between childhood diseases and a range of factors, after accounting for spatial correlation between neighbouring areas. Such semi-parametric regression models allow joint analysis of non-linear effects of continuous covariates, spatially structured variation, unstructured heterogeneity, and other fixed effects on childhood diseases. Modelling and inference made use of the fully Bayesian approach via Markov Chain Monte Carlo (MCMC) simulation techniques. The analysis was based on data derived from the 1999, 2005/6 and 2010/11 Zimbabwe Demographic and Health Surveys (ZDHS).Results and conclusionsThe results suggest that until recently, sex of child had little or no significant association with childhood diseases. However, a higher proportion of male than female children within a given province had a significant association with childhood cough, fever and diarrhoea. Compared to their counterparts in rural areas, children raised in an urban setting had less exposure to cough, fever and diarrhoea across all the survey years with the exception of diarrhoea in 2010. In addition, the link between sanitation, parental education, antenatal care, vaccination and childhood diseases was found to be both intuitive and counterintuitive. Results also showed marked geographical differences in the prevalence of childhood diarrhoea, fever and cough. Across all the survey years Manicaland province reported the highest cases of childhood diseases. There is also clear evidence of significant high prevalence of childhood diseases in Mashonaland than in Matabeleland provinces.

Highlights

Many sub-Saharan countries are confronted with persistently high levels of childhood morbidity and mortality because of the impact of a range of demographic, biological and social factors or situational events that directly precipitate ill health
Deviance Information Criterion (DIC) values for M1 were very high when compared to other models, signifying that for this study, structured spatial effects alone do not give a true reflection of the variation in the prevalence of childhood diseases and that other factors should be taken into account
M3 is an improvement on M2, this improvement is considered indistinguishable for differences in DIC of < 3, for example, the 1999 diarrhoea, fever and cough datasets (ΔDIC values of 1.52, 0.65 and 1.57 respectively); 2005–6 fever and cough datasets (ΔDIC values of 1.86 and 1.55, respectively)

Summary

Introduction

Many sub-Saharan countries are confronted with persistently high levels of childhood morbidity and mortality because of the impact of a range of demographic, biological and social factors or situational events that directly precipitate ill health. Population surveys of health and fertility provide valuable insight into the prevalence of diseases in thirdworld countries. They can provide nationally and regionally representative estimates on a range of epidemiological variables, such as time of exposure, age, gender and occupation of the person exposed. The Demographic and Health Surveys (DHS) are a wellestablished source of reliable population data with a substantial focus on childhood diseases, which are a great deterrent militating against the demographics of underdeveloped nations. The under-five mortality rate rose from 77 deaths per 1000 live births in 1994 to 84 deaths

Methods

Results

Discussion

Conclusion