Background: Bangladesh has been struggling to reduce the prevalence of childhood undernutrition, which impedes physical and mental capability and accelerates morbidity and mortality.
Objectives: The objective of the paper is to examine the changes over time in the association between potential covariates and nutritional status of Bangladeshi children.
Methods: The study combined and analyzed data from six waves of Demographic and Health Surveys between 1997 and 2014. Multivariable binary logistic regression models have been fitted to data from individual waves. Overall association has been investigated using forest plots, and meta-regression has been utilized to assess the pace of change in the association over time.
Results: Parental education and place of residence showed a consistent association with nutritional status of children. Children from parents with no little education were more likely to be undernourished than those from parents with secondary or higher level of education (odds ratio [OR] in 1997 = 3.34, 95% confidence interval [CI] = 2.65-4.22, OR in 2004 = 1.93, 95% CI = 1.58-2.37). On the other hand, gaps in the association of wealth and childhood nutrition have been widening consistently so that in 2014 children from households from the lowest 40% wealth category were 2.66 times (OR = 2.66, 95% CI = 2.13-3.33) as likely as to be undernourished than those from upper 20%.
Conclusions: The findings have policy implications in terms of developing programs directed to mothers with a relatively poor socioeconomic background. A specific example would be providing nutritional education in relation to importance of childhood nutrition or cheaper nutritious food.
Keywords: Forest plots; logistic regression; meta-regression; stunting.