|
Spatial analysis of the relationship between early childhood mortality and malaria endemicity in Malawi
|
| 12 May 2008 | Lawrence N. Kazembe, Christopher C. Appleton, Immo Kleinschmidt
|
Malaria is common in Malawi and children <5 years of age reportedly suffer a mean of 9.7 malaria episodes a year (Wirima, 1996; NSO, 1997). Not only is this parasite infection the cause of over 40% of deaths in children under two, but also one of the most frequent causes of morbidity and mortality in all Malawian children <5 years of age. The annual cost of malaria is estimated at US$35 per household, i.e. 7.2% of the average household income (Ettling et al., 1994).
Spatial differences have been reported in Africa in mortality amongst children under-five years of age (Balk et al., 2004; Gemperli et al., 2004; Kandala and Ghilagaber, 2006). Risk factors contributing to the geographical variation include bio-demographic and socio-economic factors, the prevalence of infectious diseases and the variability in the quality of child health care (Madise and Diamond, 1995; Root, 1999).
This paper is concerned with investigating the link between early childhood mortality and malaria. A spatial Cox regression model was applied (Hennerfeind et al., 2006) to analyze child survival assessing the influence of both individual-specific factors, malaria endemicity and group-specific environmental factors, approximated by geographical location. Geographical location was considered at the sub-district level using this unit for spatial analysis. This choice was driven by the need to highlight small area variation which should not only be more meaningful for health policy decision makers but also improve our understanding of the epidemiology of early childhood mortality in the country. This information is critical for designing interventions to achieve the greatest impact.
Three models were fitted for infant mortality. Model 3 did not include malaria and was the best fit (DIC = 7204.48). The risk varied with demographic and socio-economic factors. The risk was higher for boys than for girls and rural children were more at risk than urban children. The level of risk was inversely proportional to the level of education of the mothers, i.e. children of mothers with high-school education were less at risk than children born to mothers with only primary school education, or to mothers without any education. In addition, first-born children were at increased risk and birth interval, if less than two years also placed children at increased risk.
Figure A shows the plot of residual spatial effects based on model 3 representing unobserved, or unknown, environmental or geographical factors. A clear spatial variation is displayed in the map with increased hazard along the lakeshore regions and in the southern region and decreased hazard in the central- western areas and in the north of the country. The sub-district-specific geographical variation in infant and child mortality also showed consistently higher risk in rural than in urban settings.
On the one hand, the present analysis shows that malaria endemicity, in addition to the effect of socio-economic and bio-demographic factors, is associated with early childhood mortality. On the other hand, this effect was shown to be different when infants and children were compared. Indeed, the risk of infant mortality is similar at all levels of malaria endemicity, despite the bivariate data showing that the risk of infant mortality marginally increased at the highest level of malaria endemicity. For child mortality, however, the malaria risk was found to impose a greater risk at the intermediate or higher levels (≥50% prevalence) than at lower levels of endemicity.
This study fills a significant gap in the knowledge of geographical variations of early childhood mortality in Malawi. The maps identify areas of increased hazard and these patterns have important implications for health policy aimed at reducing all causes of child mortality by 50% by 2010.
To read the full article, go to: www.geospatialhealth.unina.it/articles/v2i1/gh_v2i1_05_kazembe.pdf |
|
|
|
|
View archived features |
|
|
|
