Changes in the determinants and spatial distribution of under-five stunting in Bangladesh: Evidence from Bangladesh Demographic Health Surveys (BDHS) 1996-97, 2014 and 2017/18.

backgroundBangladesh has experienced tremendous change in child nutrition over the past few decades, but there are large differences between different regions in progress made. The question is whether continuation of current policies will bring the progress needed to reach national and international targets on child nutrition security. DATA AND

methodsUsing national data BDHS 1996/97, 2014, and 2017, this study attempts to map such reductions across Bangladesh and to explore the distribution of covariate effects (joint effects) that are associated with childhood stunting over these two periods, overall and by region. The main contribution of this paper is to link observed stunting scores to a household profile. This implies that different variables are evaluated jointly with stunting to assess the likelihood of being associated with stunting.

resultsOverall, the covariates: 'Parental levels of education', 'children older than one year old', 'children live in rural area', 'children born at home' formed the country winning profile in 1996/97, whereas parental levels of education disappear in the winning profile for children stunted in 2014. This implies that over the years, Bangladesh has been successful in addressing parental education for long-term reductions in child undernutrition. In addition, the diversity of profiles of households with stunted children increases over time, pointing at successful targeting of policies to increase food security among children over the period. However, in areas where improvements have been insignificant, also the profiles remain stable, indicating a failure of policies to reach the target populations. The analysis for 2017 confirms this picture: the diversity of profiles remains high, with little change in the dominant profiles.

conclusionFurther decline in stunting is possible through region specific multipronged interventions, targeting children older than one year among vulnerable groups, in addition with strengthening family planning programs as larger families also have a higher risk to have stunted children. In general, the profiles in 2014 and 2017/18 are much more diverse than in 1996, which can be explained by the relative success of specific targeted policies in some divisions, while being much less successful in other regions. In sum, our results suggest that the challenge lies in the implementation of policies, rather than in the generic approach and assumed theory of change.