Statistical analysis

We use counts and percentages to describe the distribution of anaemia and BMI among 15–19 years old by residence, wealth and education, both for the India 2005 and the India 2015 DHS.

Chi-squared or Fisher exact and trend P-value tests were utilised to evaluate differences across the levels of the covariates. To evaluate the risk of anaemia by adolescent nullipara women v. older nullipara women, we computed unadjusted and adjusted OR by logistic regression. For multivariable analysis, we adjusted for residence, wealth and education. Within-DHS correlation was controlled for by running a non-linear mixed logistic random effect model⁽¹⁶⁾.

We used Stata v13.1 SE (StataCorp. LP.) for statistical analysis.