This open-access article is distributed under Creative Commons licence CC-BY-NC 4.0.Stunting is a risk factor for poor health and psychosocial development arising from in utero and/or early childhood malnutrition. [1] As the most prevalent form of child malnutrition, stunting remains a public health problem. [2] Although efforts have been put in place by various stakeholders to ameliorate these negative impacts through several nutrition interventions, [1,3,4] reduction in the trend of malnutrition in under-5-year-olds (hereafter 'under-5 malnutrition') in Malawi and many sub-Saharan African countries is still suboptimal. Worryingly, the 2015 -2016 Malawi Demographic Health Survey (MDHS) reported a prevalence of 37.1% stunting in under-5s. [5] Previous studies have used varying statistical approaches to examine factors associated with stunting in Malawi, [6][7][8][9] in other sub-Saharan African countries [10][11][12][13] and elsewhere. [14][15][16] These studies attest to individual setting or country peculiarities and the multifaceted nature of the associated risk factors. In Malawi, the likes of quantile regression, [6] linear random effect model, [17] generalised estimating equation, [9] logistic regression, [7,18] and multilevel logistic regression model [8] have been employed to identify determinants of stunting. Many of these studies, however, do not account for unobserved heterogeneity in clusteredsurvey data.We sought to identify potential risk factors of stunting among under-5s in Malawi using a generalised linear mixed model (GLMM) approach. The GLMM was employed to appropriately adjust for the peculiar nature of hierarchical data, including missing observation, using the most recent MDHS clustered dataset. This was necessary to attain objective and unbiased inferences on predictors of child stunting.
Methods
Study design, setting and populationThe present analysis used the 2015-16 MDHS data, a cross-sectional design aimed at providing population and maternal-and-childhealth indicators. [5] Ethical approval for the parent study was obtained from the National Health Sciences Research Committee, Malawi. The Demographic and Health Surveys Program approved the use of the dataset for the present analysis.Malawi has a population of 18.1 million, of whom 2.9 million were under 5 years of age in 2016. Only 16% of the population resided in urban areas. [4,5] The multistage cluster sampling technique was used for the survey, based on the sampling frame containing enumeration areas (EAs), adopted from the 2008 Malawi Population and Housing Census. At the first stage, 850 EAs referred to as clusters (communities) were selected as the primary sampling units. A total of 26 361 of the sampled households within the selected EAs were interviewed at the second stage. The detailed sampling procedure has been reported previously. [5] Of 6 033 under-5s eligible for anthropometric measurement within households, 5 686 (94%) had complete and valid height measurements. [5] The term 'cluster' is used interchangeably with 'commun...