Addressing diarrhea prevalence in the West African Middle Belt: social and geographic dimensions in a case study for Benin

Abstract

Background: In West Africa, the Northern Sahelian zone and the coastal areas are densely populated but the Middle Belt in between is in general sparsely settled. Predictions of climate change foresee more frequent drought in the north and more frequent flooding in the coastal areas, while conditions in the Middle Belt will remain moderate. Consequently, the Middle Belt might become a major area for immigration but there may be constraining factors as well, particularly with respect to water availability. As a case study, the paper looks into the capacity of the Middle Belt zone of Benin, known as the Oueme River Basin (ORB), to reduce diarrhea prevalence. In Benin it links to the Millennium Development Goals on child mortality and environmental sustainability that are currently farthest from realization. However, diarrhea prevalence is only in part due to lack of availability of drinking water from a safe source. Social factors such as hygienic practices and poor sanitation are also at play. Furthermore, we consider these factors to possess the properties of a local public good that suffers from under provision and requires collective action, as individual actions to prevent illness are bound to fail as long as others free ride.

Methods: Combining data from the Demographic Health Survey with various spatial data sets for Benin, we apply mixed effect logit regression to arrive at a spatially explicit assessment of geographical and social determinants of diarrhea prevalence. Starting from an analysis of these factors separately at national level, we identify relevant proxies at household level, estimate a function with geo-referenced independent variables and apply it to evaluate the costs and impacts of improving access to good water in the basin.

Results: First, the study confirms the well established stylized fact on the causes of diarrhea that a household with access to clean water and with good hygienic practices will, irrespective of other conditions, not suffer diarrhea very often. Second, our endogeneity tests show that joint estimation performs better than an instrumental variable regression. Third, our model is stable with respect to its functional form, as competing specifications could not achieve better performance in overall likelihood or significance of parameters. Fourth, it finds that the richer and better educated segments of the population suffer much less from the disease and apparently can secure safe water for their households, irrespective of where they live. Fifth, regarding geographical causes, it indicates that diarrhea prevalence varies with groundwater availability and quality across Benin. Finally, our assessment of costs and benefits reveals that improving physical access to safe water is not expensive but can only marginally improve the overall health situation of the basin, unless the necessary complementary measures are taken in the social sphere.

Conclusion: The ORB provides adequate water resources to accommodate future settlers but it lacks appropriate infrastructure to deliver safe water to households. Moreover, hygienic practices are often deficient. Therefore, a multifaceted approach is needed that acknowledges the public good aspects of health situation and consequently combines collective action with investments into water sources with improved management of public wells and further educational efforts to change hygienic practices.

Figure 1

Population density in West Africa. The insert map shows infrastructure (black lines), the Oueme River (grey line) and settlements (dots) in the Beninese ORB.

Figure 2

Diarrhea prevalence at (a) municipality, (b) diarrhea prevalence at cluster level, and (c) spatial difference in diarrhea prevalence across the country: deviation of diarrhea prevalence at cluster level from the national average.

Figure 3

(a) Accessibility to good water source, (b) aquifer suitability stress index, (c) ground water quality stress index. Figure 3 (a) is created by averaging binary variable, GW, of accessibility to good water source at household level to the cluster level. In figures 3 (b) and (c) higher values indicate poorer aquifer suitability under higher population densities and poorer ground water quality under higher population densities respectively.

Figure 4

(a) Indicator of household characteristics at cluster level, and (b) variability in diarrhea prevalence by socio-economic conditions at household level. In Fig 4 (a), the indicator of household characteristics is closer to 1 for poorer household conditions. The x-axis in Fig 4 (b) shows 1-Ind1, classified into 5 equiprobable classes with increasing magnitude of 1-Ind1, wherein higher class values imply poorer household conditions.

Figure 5

(a) Scatterplot of observed versus predicted prevalence averages (at cluster level) for Benin, and (b) the predicted cluster-wide averages in the ORB. The solid line in (a) is a 1:1 line, benchmark for a perfect fit.

Figure 6

(a) Population map of the ORB by settlements, (b) modelled prevalence map by settlements in the ORB, (c) average modeled prevalence rates for the lowest tertile, and (d) the highest tertile with poor access to good water source. In figures 6 (c) and (d) prevalence rates are averaged over households of nearest cluster that have GW = 0.

Figure 7

(a) Change in average prevalence over the basin with policy intervention of increasing accessibility to the lowest household class (red) and total cost of these interventions (blue). (b) Spatial distribution in the ORB for the cost of increasing accessibility of the lowest household class by reducing the fraction of households with poor access by half.