Abstract
BACKGROUND: One of the reported causes of high malnutrition rates in Burundi and Rwanda is children's inadequate dietary habits. The diet of children may be affected by individual characteristics and by the characteristics of the households and the communities in which they live. We used the minimum dietary diversity of children (MDD-C) indicator as a proxy of diet quality aiming at: 1) assess how much of the observed variation in MDD-C was attributed to community clustering, and 2) to identify the MDD-C associated factors. METHODS: Data was obtained from the 2010 Demographic and Health Surveys of Burundi and Rwanda, from which only children 6 to 23 months from rural areas were analysed. The MDD-C was calculated according to the 2007 WHO/UNICEF guidelines. We computed the intra-class coefficient to assess the percentage of variation attributed to the clustering effect of living in the same community. And then we applied two-level logit regressions to investigate the association between MDD-C and potential risk factors following the hierarchical survey structure of DHS. RESULTS: The MDD-C was 23% in rural Rwanda and 16% in rural Burundi, and a 29% of its variation in Rwanda and 17% in Burundi was attributable to community clustering. Increasing age and living standards were associated with higher MDD-C in both countries, and only in Burundi also increasing level of education of the mother's partner. In Rwanda alone, the increasing ages of the head of the household and of the mother at first birth were also positively associated with it. Despite the identification of an important proportion of the MDD-C variation due to clustering, we couldn't identify any community variable significantly associated with it. CONCLUSIONS: We recommend further research using hierarchical models, and to integrate dietary diversity in holistic interventions which take into account both the household's and the community's characteristics the children live in.