Exploring determinants of under-5 stunting in Malawi using a generalised linear mixed model

Abstract

Background. Although several studies have examined determinants of stunting, most fail to account for unobserved heterogeneity in clustered survey data. 

Objective. To investigate the determinants of stunting among Malawian children under-5 years of age. 

Methods. The study utilised cross-sectional data on 5 686 children aged 0 - 59 months extracted from the 2015 - 2016 Malawi Demographic and Health Survey dataset. Data were analysed using a generalised linear mixed model applied to clustered data. 

Results. Children who were female (adjusted odds ratio (aOR) 0.83; 95% confidence interval (CI) 0.73 - 0.94), from wealthier households (aOR 0.68; CI 0.58 - 0.81), and whose mothers were overweight/obese (aOR 0.78; CI 0.65 - 0.94), aged 25 - 34 years (aOR 0.76; CI 0.62 - 0.93) or had at least a secondary level education (aOR 0.73; CI 0.56 - 0.94) were less likely to be stunted. The likelihood of stunting was higher among children who were anaemic (aOR 1.38; CI 1.20 - 1.59) and in whom respondents reported small birth size (aOR 1.61; CI 1.34 - 1.93). Age ≥12 months, being a twin or triplet and living in the Central region of Malawi also increased a child’s risk of being stunted. About 3.0% of the variance in likelihood of being stunted occurred across communities (clusters). 

Conclusions. Hidden community variations of child stunting in clustered-survey data need to be accounted for. Stunting strategies should be context specific.