Topical Area: Maternal, Perinatal and Pediatric Nutrition
Objectives : Anemia is a public health problem in Ethiopia. There is strong evidence to show that both individual and community level factors play an important role in the prevalence of anemia and should be considered while designing effective programs to reduce anemia. The study aims to explore the determinants of anemia at both individual and community levels and to analyze any differential effect of a predictor at the two levels.
Multilevel random intercept model was estimated using Ethiopia DHS 2016 with group mean centered variables at the individual level and grand mean level centering for cluster-level variables. The analysis was conducted in Stata/SE 15.1 using “melogit”. The final sample included 14,489 women of reproductive age group (WRA) belonging to 25 strata with an average cluster size of 580.
The variation between strata accounted for almost 11 percentage of the variability in the outcome. Woman’s age and number of children at the individual level as well as at the cluster level, and religion and education at the individual level are significant predictors. In addition, significant cross-level interactions include mean number of children in cluster versus woman’s age and household wealth; religion versus cluster means of household wealth and prevalence of high decision-making power and child marriage; and number of children versus mean age in cluster.
Results suggest that women who are younger than the average cluster age and reside in relatively younger clusters have a higher probability of being anemic as compared to women who are older than the average cluster age. On the other hand, the probability of being anemic is the lowest across ages for younger women if they reside in clusters with highest mean age. Similarly, the association of poverty at the individual level reduces with increasing average wealth at cluster level while the opposite is true for number of children. Cluster average of number of children has a differential association with the probability of being anemic across religions. Cluster prevalence of high decision-making power among women and child marriage have a negative association with the probability of being anemic though the association varies by religion.
Conclusions : Results showed the importance of variation between strata and clusters; and the importance of these when interacting with some individual level characteristics, that are not modifiable. These results have important implications for anemia program and policy
Funding Sources : FHI 360, Bill & Melinda Gates Foundation