A multilevel multinomial logistic regression model for identifying risk factors of anemia in children aged 6-59 months in northeastern states of India

Abstract: In this article, we use multilevel multinomial logistic regression model to identify the risk factors of anemia in children of northeastern States of India. The data consisted of 10,136 children of age group 6-59 months. We considered the level of anemia as the outcome variable with four ordinal categories (severe, moderate, mild, and non-anemic) based on hemoglobin concentration in blood as per WHO guidelines. A two-level random intercept model was considered with state of residence as the level-2 variable. The intra-class correlation (ICC) between states is 0.0577 indicating approximately 6% of the total variation in the response variable accounted for by the state of residence. Several multilevel models have been compared, and a final model was decided based on deviance test. We observed that predicted probability of being at or below severely anemic level to be 0.1247, at moderately anemic level: 0.3578, at mildly anemic level: 0.0698, and being non-anemic to be 0.4477. We found that age at marriage (OR=1.13, 95% CI: 1.05, 1.21) and the number of children even born (OR=1.09, 95% CI: 1.03, 1.15) have significant effect on being at or below lower hemoglobin level (severely anemic). Furthermore, age of child (OR=0.92, 95% CI: 0.86-1.00) was a significant predictor, indicating that odds of severe anemia decreases if the child is 48 months or older.