Socio-Demographic Determinants of Anaemia in Children in Uganda: A Multilevel Analysis

Abstract

AbstractBackground: Anaemia is a worldwide public health problem. Recently it affected two billion people (WHO. (2008). Worldwide prevalence of anaemia 1993–2005, WHO Global database. Geneva: World Health Organisation) which accounts for about 25 % of the world population. It appears to be not only a major cause of pre- and post-partum morbidities and mortalities in developing countries but also it affects the physical and cognition development of children and its impact in increasing children’s risk of getting other infections is of major concern. Although the immediate biological causes of anaemia are well documented, socioeconomic factors associated with anaemia and the fact that anaemia differs markedly between individuals, within households and communities have rarely been explored.Aims: To investigate background characteristics of anaemia (socioeconomic and demographic related factors) in a selected sample of children in Uganda in 2006 after accounting for some proximate determinants and use multilevel logistic regression models to account and quantify variability due to individual (child), household and community (cluster) levels.Methods: this is a retrospective cross-sectional study of 2,110 children aged between 5 and 54 months from Uganda. The 2006 Uganda Demographic Health survey (UDHS 2007) was utilised for this analysis–more detail on data and sampling design and methodology can be found in (www.measuredhs.com). Potential risk factors of anaemia which were previously reported were studied and contextual variables were also included. Bivariate and multivariate (multilevel logistic regression) analysis were carried out. In the bivariate analysis, Chi-square test was used to test the significance of the selected potential risk factors and the Wald test was used in the multivariate analysis to test the joint significance of categorical factors. For both tests a p-value of <0.05 was used as a cut off point, odd ratios and associated 95 % confidence interval were computed for multilevel logistic model. Estimates and odds ratios for simple logistic regressed model are presented for comparison purposes.Results: Of the 2,110 children whose haemoglobin level was tested, 73 % are anaemic. Anaemia is multifactorial, however this study suggests that maternal and paternal education (respective p < 0.001 and 0.002), maternal occupation (p <0.001), wealth index (p < 0.001), religion (p < 0.001), mother’s smoking habit (p = 0.04), the place of residence (p < 0.001), the place of delivery (p = 0.05), region (p < 0.001), whether the child was given meat (p < 0.001) and whether the child ate green leafy vegetables (p < 0.001) are factors associated with anaemia in children in Uganda. It also suggests that children’s age is an important risk factor and was significantly related to anaemia in children and that younger children, male, children of mothers who are not working at all or who are not skilled and children of anaemic mothers are associated with an increased risk of anaemia compared with those children aged between 2 and 5 years old, female, children of mothers who work in services sectors or are skilled and children whose mother are not anaemic. In addition, the results indicate that in situations where children are nested within households and households within communities the hierarchical nature of data needs to be accounted for. Otherwise, factors that are not significant at all could appear to be significant. More variability in children’s risk of having anaemia is at community level followed by individual level. No association was found between anaemia and maternal education, birth order, preceding birth interval, birth weight, maternal age and breastfeeding, region, the place of residence and household wealth status.Conclusions: Anaemia is highly prevalent among children in Uganda and may be due to diverse factors. However, the results from the present study suggests that children age, gender, maternal occupation and whether the mother is anaemic are significantly related to children’s risk of having anaemia. The results indicate that in situations where subjects are nested within units, the hierarchical nature of the data should be accounted for otherwise the standard errors of variables included in the model would be biased downward, resulting in over statement of the significance of variables that in reality are not significant. It also shows the importance of assumption regarding planning and developing anaemia intervention at community level followed by individual level.KeywordsWealth QuintileGreen Leafy VegetableSimple Logistic RegressionAnaemia InterventionMultilevel Logistic Regression ModelThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.