STATISTICAL MODELLING OF HAEMOGLOBIN LEVEL IN UNDER FIVE YEARS OLD CHILDREN IN NIGERIA

Abstract

Haemoglobin concentration is a clinical indicator that examines or measures the presence or otherwise of an anaemia in an individual subject particularly due to iron deficiency. Normal Haemoglobin distributions vary with age, sex, life style, race/ethnicity, socio-economic status, regional difference, drug related issues and clinical characteristics including guidelines, and protocols. Issues arising from the recent spikes in the number of new-borns who were anaemic have attracted attention globally and these are of great concern particularly in Sub-Sahara Africa in which Nigeria has huge share of statistics. Hence, this study examined haemoglobin levels in under 5 children in Nigeria and the set of factors driven it from various statistical approaches. These models were: linear regression model (LM), Linear mixed model (LMM) and multilevel model (MM). This study used dataset of children included in the Nigeria Malaria Indicator Survey, 2015. The results of LM, identified two significant predictors of under 5 children haemoglobin level, also only three predictors were significant under LMM. The random effect of household number under LMM setting had higher variability than the state as random effect. In MM with state and household number as random effects, area of residence of the child, head of household wealth index, and the age of the child were all significant. The estimates of MM produced smaller standard errors compared to LMM. This implies that multilevel model is more competitive than other models considered in this study. Therefore, it could be applied to predict the haemoglobin level of under 5 children in Nigeria.