Sampling and Analysing Spatial Determinants of Undernutrition in Northern Rwanda – A Multidisciplinary Approach
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background: In Rwanda, various intervention initiatives have led to progress in reducing maternal and child undernutrition. However, the rate of stunting, the chronic form of malnutrition, remains high with about 33% of children under five years stunted countrywide, especially in the Northern Province where the stunting rate is 40.5%. This study aimed to sample and analyse spatially-explicit undernutrition determinants in Northern Rwanda through a multidisciplinary approach. Methods: A multidisciplinary team of experts in public health, geographical information science, epidemiology, animal science and veterinary medicine conducted a population-based cross-sectional study in five districts of the Northern Province of Rwanda. Targeting households with mothers of children aged 1-36 months, we used a structured household-level questionnaire to collect information on socio-demographic and economic factors, child health, childcare practices, presence of violence against children and mothers, maternal health, livestock production and animal health. In addition to child and maternal anthropometric measurements, we collected rectal swabs and blood samples from children to identify potential gastrointestinal pathogens and estimate haemoglobin levels. Data on environmental and physical characteristics of the study area were obtained from existing national datasets. To capture GPS coordinates of each visited household, we customized a web-based geodata management platform into a mobile application for in-field data collection (emGeo). To analyse the determinants underlying stunting, we used the following key techniques: feature selection using random forest to identify 26 most pertinent predictors, univariable and multivariable logistic regression analyses, and geographically weighted logistic regression (GWLR) to account for spatial variation. Results : The multivariable logistic regression accounted for 30% (R 2 = 0.30) of the variation in stunting, with higher odds with increase in child’s age and sex, mother’s decision-making autonomy in major household purchases, having a friend to assist the mother when ill, access to electricity, having a home garden, and handwashing practices. The geographically weighted logistics regression improved explanatory power to 35%, highlighting spatial heterogeneity in the predictors' effects across the study area. Conclusion : A multidisciplinary approach is essential to tackle multifaceted challenges like stunting. Combining logistic regression and geospatial statistics enabled us to identify areas with high stunting risk factors and guide policymakers toward geographically targeted interventions, particularly in limited settings, including Rwanda.