Background Childhood anaemia is highly prevalent in Nigeria. According to the 2015 Nigeria Malaria Indicator Survey (NMIS) report, more than 68% of children aged 6-59 months were found to be anaemic. This estimate is far above the World Health Organization’s 40% cut off point which classifies anaemia a severe public health challenge in the country. Identifying environmental, health, socioeconomic and demographic influential factors and mapping the prevalence of anaemia can help guide geographically targeted intervention programmes to reduce the risk of anaemia associated morbidity among vulnerable children in Nigeria.
Methods Geographically linked national level datasets obtained from the 2015 Nigeria Demographic and Health Survey (NDHS) and Malaria Indicator Survey (NMIS) programmes were used for this study. For the analysis, a binary structured additive regression (BSAR) model was explored. This model incorporated a Markov Random Field (MRF) prior, with posterior parameters estimated via Bayesian framework.
Results After accounting for the spatial heterogeneity, we found a strong negative association between the odds of anaemia and the child’s demographic variables in terms of increasing age and being female. Increased odds of anaemia were also associated with child’s malaria and fever status, living in a rural environment, lower household wealth quintile, having younger and illiterate mothers. We also found that a decreased distance to vegetated areas was associated with increased childhood anaemia risk, while the odds of anaemia decreased as cluster altitude increased at 95% credible interval. The maps of the posterior means of the spatial random effects revealed evidence of spatial variation in the odds of childhood anaemia, while accounting for the model covariates (Fig. 4). Greater risk of anaemia was observed for children who resided in Adamawa, Ebonyi, Edo, Cross river, FCT, Jigawa, Kaduna, Kano and Kebbi states in Nigeria.
Conclusions In this study, a binary structured additive regression model was utilized, allowing for a flexible semiparametric predictor that accounted for the effects of different types of covariates, while simultaneously incorporating spatial variables directly. Our results revealed significant spatial variability of childhood anaemia, suggesting that spatially targeted interventions could result in efficiency gains for anaemia control in Nigeria.