Predictors of stunting in states of India with highest and lowest prevalence of under-five stunted children

Abstract

Background: According to 2017 data, globally, 155 million under five children or 23% of all under-five children are stunted. Over the last decade (2006 to 2016), in India the prevalence of stunted children has decreased from 48% to 38%, but this prevalence is higher when compared with other lower and middle income countries such as Sri Lanka (15%), Vietnam (25%). In India, prevalence of stunting varies by the states, it ranges from about 20% in Kerala to about 48% in Bihar. Aim: The study aimed to evaluate predictors of stunting among children under five years of age living in the states of Kerala and Bihar, to assess the potential differences in the sets of predictors between these states. Methods: Secondary analysis of the National Family Health Survey 2015-16 (NFHS-4) data was conducted. Descriptive data analysis was conducted using the chi-square and t test. For identifying the predictors, those variables with different distribution between the stunted and non-stunted groups at the level of significance p<0.25 in the chi-square or t test were entered into the logistic regression analysis, first in univariate, then in multivariable. For categorical variables an effect size of 5% was used. For the continuous variables the clinical significance was checked, as the continuous variables are unit sensitive. For the association between stunting and food diversity score, the confounders of association were identified. Multivariable regression model was used to identify the association between stunting status and food diversity score after adjusting for the identified confounders. Results: For Kerala after controlling for all the other significant variables the predictors of stunting were; each month increase in age of the child (O.R. = 0.99), low wealth index category (OR = 2.01) or middle wealth index category (OR = 1.45), Children from Muslim families (O.R. = 1.87), each month increase in the duration of breastfeeding (O.R. = 1.01). The stunting status was not statistically significantly associated with the food diversity score after adjusting for the identified confounders. For Bihar after controlling for all the other significant variables the predictors of stunting were; each one-month increase in the age of the child (O.R. = 1.03), low birth weight (O.R. = 1.81), each unit rise in the birth order (O.R. = 1.06), birth interval of less than 24 months (O.R. = 1.24). Mothers who were 20 years old or younger at the time of the delivery (O.R. = 1.17). Each one centimetre rise in the height of the mother (O.R. = 0.94). Mother having no education (O.R. = 1.33), belonging to scheduled caste, scheduled tribe or other backward classes caste (O.R. = 1.24), low (O.R. = 1.82) and middle wealth index (O.R. = 1.38), each month increase in duration of breastfeeding (O.R. = 1.04). The stunting status was statistically significantly associated with the food diversity score (O.R. = 0.97) after adjusting for the identified confounders. Conclusion: Study hypothesis was correct we found different sets of predictors of stunting for Kerala and Bihar. Identified modifiable variables were food diversity score, birth weight, birth interval. State specific interventions should be designed to reduce stunting. Interventions should target vulnerable population groups with low socio-economic status, Hindu and Muslim religion, lower education, and backward class. Educational interventions should cover; importance of diversity in the children’s diet, importance of antenatal care and family planning to promote birth intervals of more than 24 months.