The Geography of Child Stunting in Kenya: What Can We Learn From the Sub-National County Variations of the Kenya Household Surveys?

Abstract

Background: Reduction in child malnutrition is one of the key components of the 2030 Sustainable Development Goals. National-level monitoring efforts are very useful for closing inequalities in the burden of stunting, but they usually mask huge sub-national variations, making targeting of the most affected population groups in specific locations difficult. In this study, we examine the geographical variation of stunting at the sub-national county-level in Kenya, to highlight the importance of estimates at sub-national level for programming and for monitoring of the 2025 internationally agreed targets for stunting and the 2030 SDG target 2.2. \n \nMethods: We used a nationally representative cross-sectional household sample of 18,600 children with anthropometric measurements for height-for-age (stunting) who were under five years in the 2014 Kenya Demographic and Health Survey.\xa0 We analysed geographical variations of stunting across the 47 counties of Kenya and used a Bayesian geo-additive mixed model while accounting for observable and unobservable risk factors. This approach isolates geographical disparities, net of known covariates associated with stunting, and may point to unobservable factors that need attention. \n \nFindings: In 2014, prior to the launch of the SDGs and agreement of the 2025 Global Targets for Stunting, the national stunting rate in Kenya was 26 percent, meaning that if the 2025 targets of reducing stunting by 40 percent are to be achieved, the country should be aiming at stunting rates among under-five children of 10.4 percent or lower. Furthermore, the national average masks sub-national county level variations ranging from 10.7 percent in Kirinyaga to 37.7 percent in Kitui. Factors significantly associated with higher risk of stunting were found for boys [posterior odds ratio and 95% credible region: 1.40 (1.24, 1.58)]; children from the Pokomo ethnic group: 2.54 (1.34, 4.98); children from households with 10 or more members: 1.35 (0.99, 1.75); child’s age ( higher up to 10 months of age then levelling off at 20 months); and lower risk of stunting among children from mothers with secondary education; children from the highest wealth quintile: 0.38 (0.27, 0.53) and children with >24-month birth interval: 0.71 (0.62, 0.83). After adjusting for known covariates, there were striking residual geographical variations of stunting at county-level with the highest risk found in Madera county [odds ratio and 95% credible region: 2.00 (1.20, 3.61)], a county bordering Ethiopia and Somalia, and the lowest risk was found in the counties of Kirinyaga [0.90(0.60, 1.34)], Nairobi [1.10(0.79, 1.60)] and Nyeri [0.98(0.67, 1.43)].\xa0 \n \nInterpretation: Overall, the observed sub-national stunting estimates point to great variations in child stunting in Kenya, which show that the task of achieving the 2025 international targets for stunting and the achievement of the SDG2 is made harder by geographical disparities that would need to be addressed. The geographic patterns in undernutrition rates show the potential role of intervening at sub-national and county levels. Uneven distribution of resources, and poor access to health services and nutrition programs may explain some of these county-level differentials. \n \nFunding Information: This study was funded by the Children’s Investment Fund Foundation (CIFF). \n \nDeclaration of Interests: We declare no competing interests.