Maternal-related deaths and impoverishment among adolescent girls in India and Niger: findings from a modelling study

Background High levels of maternal mortality and large associated inequalities exist in low-income and middle-income countries. Adolescent pregnancies remain common, and pregnant adolescent women face elevated risks of maternal mortality and poverty. We examined the distribution across socioeconomic groups of maternal deaths and impoverishment among adolescent girls (15–19 years old) in Niger, which has the highest total fertility rate globally, and India, which has the largest number of maternal deaths. Methods In Niger and India, among adolescent girls, we estimated the distribution per income quintile of: the number of maternal deaths; and the impoverishment, measured by calculating the number of cases of catastrophic health expenditure incurred, caused by complicated pregnancies. We also examined the potential impact on maternal deaths and poverty of increasing adolescent girls' level of education by 1 year. We used epidemiological and cost inputs sourced from surveys and the literature. Results The number of maternal deaths would be larger among the poorer adolescents than among the richer adolescents in Niger and India. Impoverishment would largely incur among the richer adolescents in Niger and among the poorer adolescents in India. Increasing educational attainment of adolescent girls might avert both a large number of maternal deaths and a significant number of cases of catastrophic health expenditure in the 2 countries. Conclusions Adolescent pregnancies can lead to large equity gaps and substantial impoverishment in low-income and middle-income countries. Increasing female education can reduce such inequalities and provide financial risk protection and poverty alleviation to adolescent girls.

India and 553 in Niger, respectively; Table 1  Subsequently, we estimate the number of adolescent maternal deaths in income quintile and among age group in country as: (1)
Subsequently, using elements from (1) above, we can estimate the amount of OOP costs incurred by adolescents in income quintile and age group as: (2)

Adolescent maternal-related impoverishment
In country , adolescent medical impoverishment was quantified by the estimated number of cases of catastrophic health expenditure incurred, which depended on assumed income and OOP costs (see section 1.2). A case of catastrophic expenditure was counted when OOP costs were found to be higher than 10% of income . Specifically, we estimated the number of adolescents, per income quintile, for whom the size of OOP costs (sum of direct medical costs and transportation costs) would exceed 10% of their income.
For country , we derived a distribution of income drawn from a simulated gamma distribution whose shape and scale parameters were based on gross domestic product per capita ($1596 for India and $427 for Niger, respectively; Table 1 in the main text) and Gini coefficient (0.34 for India and 0.32 for Niger; Table 1 in the main text) [1][2][3].
Subsequently, for each complicated delivery with incurred OOP costs, we assigned an annual income , based on that derived income distribution. The annual income was also used to define the income quintile into which each individual belonged ( Table 1 in the   main text).
Finally, per income quintile, we calculated the number of cases of catastrophic health expenditure occurring. This was done by combining the estimate of annual income with OOP incurred costs estimated (section 1.2); in other words, we counted the number of adolescent women whose OOP costs were larger than 0.10 * .

Linear relationship between increases in female education levels and adolescent pregnancy rate
We examined the relationship between mean years of education among women aged 15-44 (denoted ) [4] and adolescent (15-19 year-olds) pregnancy rate (denoted ) (percentage of women aged 15-19 who have had children or are currently pregnant), in a given low-and middle-income country [3], controlling for gross domestic product per capita ( ) and additional variables (Table S2), using the following type of linear model: The complete results of the linear models tried are given in Table S2. For our analysis we retained model (4) where ! = −0.18. This meant that an increase by 1 year of the mean number of years of education among women aged 15-44 would lead to a relative decrease of the adolescent pregnancy rate of 18% in a given country.

Estimation of adolescent maternal-related deaths, out-of-pocket costs, and impoverishment averted
We

Sensitivity analysis
We conducted a Monte Carlo probabilistic sensitivity analysis to estimate aggregate uncertainty from key inputs. Parameters were given values using probability distributions (details are given in Table S3).
We also pursued univariate sensitivity analyses where: (1)