Multilevel modeling of geographically distributed vitamin A deficiency

Vitamin A deficiency is a common health problem in developing countries like India. Present study involves data on children aged between 6-36 months from northern part of India collected geographically, to find prevalence and important factors for risk of nightblindness. Both traditional logistic models and multilevel logistic models were applied to achieve our aim. All Individual level variables vitamin A diet intake, age, vitamin A capsule intake and awareness about vitamin A were found significant for risk of night-blindness (p < 0.05) in individual level analysis. The effect of risk factors for night-blindness was smaller in multilevel modeling as compared to individual level model. The reason is that the previous model takes into account the within-block as well as among-block variations. Multilevel analysis, did not find, individual level variables vitamin A diet intake, awareness of vitamin A and vitamin A capsule intake significant for the outcome variable (p > 0.10). There was about 139% change in odd-ratio for vitamin A capsule taken once. Block level variable, average age of subjects in blocks comes out as significant factor (p = 0.01) for night-blindness. Thus, this paper demonstrates the usefulness of multilevel modeling in the analysis for epidemiology of disease risk, which is structured in a hierarchiary, with particular reference to geographical analyses of small area data.