Determinants of chronic malnutrition among under-five children in Ethiopia using simultaneous quantile regression

Abstract

Child malnutrition remains a challenge in Ethiopia despite progress in development goals. Stunting in children under five leads to disease, impaired development, and increased mortality. Earlier studies have used linear and logistic regressions to identify the drivers of stunting. These models overlook variations across outcome distributions. This study employed simultaneous quantile regression to identify the association between chronic malnutrition across different height-for-age z-score (HAZ) quantiles in children under the age of five. Data were drawn from the 2016 Ethiopian Demographic and Health Survey, including 8,592 women aged 15 – 49 years and their under-five children. After cleaning missing variables, HAZ score served as the dependent variable. Simultaneous quantile regression modeled covariates across multiple quantiles. The findings indicated that 34.75% of children were stunted. Significant variables associated with HAZ score included place of residence, shared toilet facility, respondent employed, vaccination, succeeding birth interval in months, frequency of checking antenatal care, literacy, type of toilet facility, anemia level, wealth index of household, twin status, place of delivery, highest level of education of mother and father and age of child. The association of these factors varied across quantiles, with slope differences between the 10th and 90th quantiles. The quantile regression plots for the selected quantiles 10th to 90th revealed significant differences in the association of the covariates across the HAZ quantiles under consideration. Quantile regression revealed that various factors work differently across the HAZ distribution. Findings demonstrate the benefit of quantile regression in revealing differential impacts and guiding targeted policy. Addressing stunting requires coordinated efforts to enhance child nutrition and achieve the Sustainable Development Goals by 2030.