Geostatistical analysis and mapping: social and environmental determinants of under-five child mortality, evidence from the 2014 Ghana demographic and health survey

Abstract

Background: Under-five mortality (U5M) rates are among the health indicators of utmost importance globally. It is the goal 3 target 2.1 of the Sustainable Development Goals that is expected to be reduced to at least 25 per 1000 livebirths by 2030. Despite a considerable reduction in U5M observed globally, several countries especially those in sub-Saharan Africa (SSA) like Ghana are struggling to meet this target. Evidence-based targeting and utilization of the available limited public health resources are critical for effective design of intervention strategies that will enhance under-five child survival. We aimed to estimate and map U5M risk, with the ultimate goal of identifying communities at high risk where interventions and further research can be targeted.

Methods: The 2014 Ghana Demographic and Health Survey data was used in this study. Geostatistical analyses were conducted on 5884 children residing in 423 geographical clusters. The outcome variable is child survival status (alive or dead). We employed a geostatistical generalised linear mixed model to investigate both measured and unmeasured child specific and spatial risk factors for child survival. We then visualise child mortality by mapping the predictive probability of survival.

Results: Of the total sampled under 5 children, 289 (4.91%) experienced the outcome of interest. Children born as multiple births were at increased risk of mortality with an adjusted odds ratio (aOR) (aOR: 8.2532, 95% CI: [5.2608-12.9477]) compared to singletons. Maternal age increased risk of mortality (aOR: 1.0325, 95% CI: [1.0128-1.0527]). Child's age (aOR: 0.2277, 95% CI: [0.1870-0.2771]) and number of children under 5 within each household (aOR: 0.3166, 95% CI: [0.2614-0.3835]) were shown to have a protective effect. Additionally, mothers with secondary education level (aOR: 0.6258, 95% CI: [0.4298-0.9114]) decreased the risk of U5M. The predicted U5M risk in 2014 was at 5.98%. Substantial residual spatial variations were observed in U5M.

Conclusion: The analysis found that multiple births is highly associated with increased U5M in Ghana. The high-resolution maps show areas and communities where interventions and further research for U5M can be prioritised to have health impact.

Keywords: Child deaths; Developing countries; Geostatistical analysis; Ghana; Mapping under-five mortality; Risk factors; Sub-Saharan Africa; Under-five mortality.

Fig. 1

Child mortality in Ghana. The locations for clusters where DHS surveys were conducted in Ghana in 2014. Note: This map was produced by the authors using the free open software R

Fig. 2

Child mortality in Ghana. Results from variogram diagnostic check for compatibility of data with the fitted geostatistical model. The solid line is the empirical variogram of the data. The dashed lines are the 95% confidence bands under the fitted spatial model. Note: This map was produced by the authors using the free open software R

Fig. 3

Empirical child mortality in Ghana. Child mortality prevalence from the sampled cluster locations in Ghana in 2014. The size (from small to large) and colour (from blue to yellow) of the dot represents child mortality prevalence, ranging from low to high, respectively. Note: This map was produced by the authors using the free open software R

Fig. 4

Child mortality in Ghana. Predicted child mortality risk map in Ghana showing relatively low mortality on average with pockets of high mortality in parts of Northern, Ashanti and Eastern regions of the country. Note: This map was produced by the authors using the free open software R

Fig. 5

Child mortality in Ghana. Lower and upper 95% confidence maps of predicted child mortality in Ghana. Note: This map was produced by the authors using the free open software R

Fig. 6

Child mortality in Ghana. Uncertainty measure for predicted child mortality. Note: This map was produced by the authors using the free open software R