Risk factors for stunting among children under five years: A cross-sectional population-based study in Rwanda using the 2015 Demographic and Health Survey

Background: Child growth stunting remains a challenge in sub-Saharan Africa, where 34% of children under 5 years are stunted, and causing detrimental impact at individual and societal levels. Identifying risk factors to stunting is key to developing proper interventions. This study aimed at identifying risk factors of stunting in Rwanda. Methods: We used data from the Rwanda Demographic and Health Survey (DHS) 2014-2015. Association between children’s characteristics and stunting was assessed using logistic regression analysis. Results: A total of 3594 under 5 years were included; where 51% of them were boys. The prevalence of stunting was 38% (95% CI: 35.92-39.52) for all children. In adjusted analysis, the following factors were significant: boys (OR 1.51; 95% CI 1.25-1.82), children ages 6-23 months (OR 4.91; 95% CI 3.16-7.62) and children ages 24-59 months (OR 6.34; 95% CI 4.07-9.89) compared to ages 0-6 months, low birth weight (OR 2.12; 95% CI 1.39-3.23), low maternal height (OR 3.27; 95% CI 1.89-5.64), primary education for mothers (OR 1.71; 95% CI 1.25-2.34), illiterate mothers (OR 2.00; 95% CI 1.37-2.92), history of not taking deworming medicine during pregnancy (OR 1.29; 95%CI 1.09-1.53), poorest households (OR 1.45; 95% CI 1.12-1.86; and OR 1.82; 95%CI 1.45-2.29 respectively). Conclusion: Family-level factors are major drivers of children’s growth stunting in Rwanda. Interventions to improve the nutrition of pregnant and lactating women so as to prevent low birth weight babies, reduce poverty, promote girls’ education and intervene early in cases of malnutrition are needed. We used the 2014–2015 Rwanda DHS open access dataset [19]. The DHS included a randomly selected national total of 12,793 households from five provinces. Of the 12,793 households, a sub-sample of 6350 (50%) households was randomly selected and data on child anthropometric measurements and development indicators were collected (N = 3594 children) [15]. A full protocol explaining the data collection process and sampling methods of DHS can be reviewed elsewhere [19]. We have included a total of 17 variables related to three categories in the WHO stunting framework [17, 18]: i) Individual-level factors (sex, age group, parity, child’s weight at birth, history of diarrhea in two weeks prior to the survey); ii) Maternal (height, highest educational level, intake of parasite controlling drugs for mothers during pregnancy, number of days of daily intake of iron tablets or syrup by mothers during pregnancy, breastfeeding within the first hour after birth and household (household wealth index, household size, access to improved water at household, availability of improved toilet facility) factors and iii) Community level factors (household location data including province, sector and village; altitude (highland vs. lowland) and location (urban vs. rural). Stunting was measured by DHS, using the WHO Child Growth Standards and collected data on every child’s length/height, age and sex to calculate the number of standard deviations (Z-score) that his/her length/height is below or above the median of the 2006 WHO growth reference population [20]. These measures were recorded with two implied decimal places, thus we divided all values of DHS standard deviations by a hundred. Stunting was defined as a z-score lower than − 2. Mother’s height was considered low if it was < 145 cm [21]. Household wealth index was calculated using principal components analysis (PCA) of data on a household’s ownership of assets like television, radio, mobile telephone, computer and car, and housing characteristics (access to electricity, access to source of drinking water, type of floor material, type of toilet facility, number of bedrooms, and type of cooking fuel) [15]. Wealth index scores were adjusted for the type of residence (rural or urban). Then, DHS assigned each person in the population his/her household’s score. The study population was ranked according to their social economic scores and divided into five equal groups (20% of the population), known as wealth quintiles [22]. For this study, we re-grouped our sample of children into three groups: High-social economic status (SES) (highest two wealth quintiles), middle SES (middle wealth quintile) and low-SES (lowest two wealth quintiles) [15]. Child weight at birth was defined as low if it was less than 2.5 kg [23]. Location of the household was defined as highland if the household location is at an altitude of > 1642 m (median altitude of households), and lowland if the household location is at an altitude of < 1642 m. To determine risk factors for stunting, we first conducted a full logistic regression model with all 17 variables by calculating odds ratio (OR), 95% confidence interval (CI) and p-value. Then, we conducted a final logistic regression model by controlling for sex, province, and altitude. All variables with p-values 0.10 or less in the full model were considered in our final logistic regression model, and removed using backward stepwise selection stopping when all the final variables were significant at the α = 0.05 significance level. Stata/SE 13.1 was used for data analysis [24]. We used the svy commands to account for the complex survey sampling and used sampling weights to account for unequal probability sampling in different strata.