Growth failure among children of adolescent mothers at ages 0–5 and 6–12 years in India

Adolescent motherhood has been linked with poor health outcomes at birth for children, including high neonatal mortality, low birthweight, and small‐for‐gestational‐age rates. However, longer‐term growth outcomes in the children of adolescent mothers in low‐resource settings remain inadequately studied. We used longitudinal data from the India Human Development Surveys, 2004–2005 and 2011–2012 (n = 12,182) and employed regression and propensity score matching analysis to compare the following growth indicators of children born to adolescent mothers (ages 19 years or below) with those born to older mothers. Growth indicators included height and weight during ages 0–5 years and 6–12 years and change in height and weight between the two periods. In regression‐based estimates, children born to adolescent mothers were 0.01 m shorter and weighed 0.2 kg less than children of older mothers at ages 0–5 years. At ages 6–12 years, those born to adolescent mothers were 0.02 m shorter and weighed 0.97 kg less. The height difference between the two groups increased by 0.01 m and the weight difference grew by 0.77 kg over time. Height and weight difference between the two groups worsened among boys over time, while for girls, only the weight gap worsened. The results were similar when using propensity score matching methods. Public policies for reducing child marriage, combined with targeted health, nutrition, and well‐being programs for adolescent mothers, are essential for both preventing adolescent childbearing and reducing its impact on growth failure among children in India.


INTRODUCTION
An estimated 12 million girls aged 15-19 years give birth every year in low-and middle-income countries (LMICs). 1 India remains one of (27 per 1000 girls) areas of India, and in the states of Bihar, Assam, Andhra Pradesh, Tripura, and West Bengal, the rate was over 60 per 1000 girls. 4 Adolescent childbearing carries substantial health risks for the mother and the child. Maternal conditions, including pregnancy and birth complications, are the second most common cause of death, after tuberculosis, among adolescent girls 15-19 years of age worldwide. 5 A recent meta-analysis of studies from 18 LMICs estimated that neonatal mortality, low birthweight, and small-for-gestationalage rates were 12-25% higher among children of mothers aged [15][16][17][18][19] years as compared with those of 20-to 29-year-old mothers. 6 Another World Health Organization study used 2010-2011 data from 29 countries and found similarly higher rates of low birthweight, preterm birth, and severe neonatal conditions among children of adolescent mothers. 6 Adverse outcomes during the first 5 years of life, such as stunting and anemia, have also been documented among children of teenage mothers in many countries. [7][8][9][10][11] The current literature linking maternal age and child health outcomes in India and other LMICs focuses heavily on outcomes during birth and the neonatal phase. [7][8][9][10][11] The "fetal origins" hypothesis of early childhood development suggests that low birthweight and other adverse conditions around birth can have potential long-term negative effects on children's health, cognition, schooling, and labor market outcomes. [12][13][14] A recent systematic review of 11 studies from mainly high-income countries found that catch-up growth-defined as reaching normal height later in life-occurred in 87% of smallfor-gestational-age children. 15 While catch-up growth in the overall population of stunted LMIC children has been also documented to a limited extent, 16,17 whether catch-up growth can occur among the children of adolescent mothers remains inadequately known.
Only a few studies have examined the outcomes of children born to adolescent mothers beyond the first 5 years of life in LMICs. Fall et al. used the Consortium for Health-Orientated Research in Transitioning Societies (COHORTS) data from five LMICs and found that besides low birthweight and preterm birth risks, adolescent motherhood was associated with 38% higher likelihood of failing to complete secondary school among children. 18 Benny et al. used the Young Lives longitudinal data in four LMICs, including India, and found that children born to stunted adolescent mothers had a 15 percentage point higher chance of being stunted in infancy-but not at older ages-as compared to those born to nonstunted older mothers. 19 Another longitudinal study from South Africa found that children of teenage mothers were more likely to be underweight and stunted as compared with those born to older mothers. 20 In India, where adolescent motherhood is correlated with child marriages, a 1 year delay in marriage-controlling for background socioeconomic indicators-has been linked with 2.2% lower likelihood of home births, 5.1% higher duration of breastfeeding, 4.6% higher likelihood of receiving age-appropriate childhood vaccines, and 0.08 higher weight-for-age z-score among under-5 children. 7 The authors also linked delayed marriages of women with small improvements in schooling enrollment, attainment, and cognitive test scores ranging from 2% to 3% among their 8-to 11-year-old children. 7 Another important limitation of the existing literature is that many clinical studies do not fully account for the socioeconomic conditions of adolescent girls. Adolescent motherhood in LMICs is strongly correlated with lower levels of maternal schooling, standard of living, female empowerment, and access to family planning and prenatal care. 18,21,22 These factors may also be associated with lower human capital investment in children, confounding the link between adolescent motherhood and child development. Finally, the literature is largely silent on the health implications for girls versus boys born to teenage mothers.
In India, there is a strong preference for sons over daughters, leading to sex-selective abortion of female fetuses and reduced health and education resource allocation for young girls vis-à-vis boys. [23][24][25][26] Because adolescent motherhood and neglect of the female child can both be driven by lower female autonomy, it is important to understand the long-term health effects separately by sex.
To address these major gaps, we examined the long-term trajectories of growth indicators from birth to 12 years of age in Indian children born to adolescent mothers in comparison with those born to older mothers. We used nationally representative longitudinal data from a large household survey and employed regression and propensity score matching (PSM) methods to account for a wide array of maternal health and background socioeconomic conditions.

India Human Development Survey data
We used longitudinal data from two rounds of the India Human Devel- We examined the height (meter) and weight (kg) of the same set of children across IHDS (0-5 years old) and IHDS-2 (6-12 years old).
We excluded outlier values of height and weight by truncating the distributions at 1% and 99% as is commonly done in the literature. Our final sample included approximately 12,000 children whose height and weight data were available in both rounds of the survey. Data were subject to minor variations in the analysis due to missing values of covariates. Nine and nine-tenths percent of the sample of children were born to adolescent mothers (the age of the mother was 19 years of age or below at the time of birth). The age of the mother at the time of birth was calculated based on mother-reported information on her own age and the age of the child at the time of the survey.

Regression analysis
We evaluated the relationship between adolescent childbearing and child growth over time using the following outcome variables: child height and weight during ages 0-5 years and 6-12 years, and change in height and weight between the two periods. We regressed these outcomes on a binary indicator of whether the child was born to an adolescent mother (ages 19 years or below), controlling for a large set of demographic and socioeconomic indicators, which included the age of the child in years, age squared, sex (whether female), and household characteristics, such as location (whether rural), whether the household belonged to any government-recognized socioeconomically disadvantaged group (known as scheduled caste, scheduled tribe, or other backward classes), religion (Muslim or Christian), and sex of the household head (whether female). To account for the intergenerational transfer of resources, we included the age and years of schooling completed by the household head among the covariates of the regression.
Finally, we accounted for standard of living by dividing the monthly per capita consumption expenditure of the household into five quintiles and including indicators of the top four quintiles in the model. All covariates were drawn from the IHDS data, and all regression models included state fixed effects. Standard errors were clustered at the primary sampling unit (village or urban ward) level.

PSM analysis
As discussed earlier, the relationship between adolescent motherhood and child growth may be determined by several intertwined community, family, and child background characteristics. Children born to adolescent versus older mothers may be systematically different (e.g., the former may belong to poorer and rural households), which may affect the long-term nutritional status. To mitigate systematic baseline differences between the two groups, we employed a widely used quasiexperimental methodology of PSM. 27-31 PSM compares the outcomes of the affected group (children of adolescent mothers) with those in the nonaffected group (children of older mothers) who had the same or very similar probability of being included in the affected group based on their observable characteristics. If the unobservable characteristics between the two groups were assumed to be distributed at random, the group difference in child growth indicators could be attributed solely to adolescent motherhood.
Using IHDS data, we first conducted a probit regression of the binary indicator of the mother's age (whether the child was born to an adolescent mother aged 19 years and below) on the set of background socioeconomic variables and state fixed effects described in the previous section. The predicted probability from this model (known as the propensity score) was then used to match each child born to an adolescent mother with similar children born to an older mother.
We only considered observations across the two groups with overlapping propensity scores (known as "common support"). We employed a Kernel (Epanechnikov) matching algorithm which used the entire set of nonaffected observations and assigned decreasing weights based on propensity score distance from the affected child. Matching was done separately for each outcome variable and survey round to reduce missing values of covariates and the effect of attrition. We report the difference in outcome indicators between the matched affected and nonaffected children, which is also known as the average treatment effect on the treated.

Validity of PSM and additional matching algorithms
We tested the validity of PSM (i.e., whether the methodology substantially reduced the initial differences in the characteristics between affected and nonaffected children) in three different ways. First, we computed the mean standardized percentage bias (i.e., the average percentage difference in the values of the explanatory variables in the propensity score estimation equation between the two groups before matching). We then calculated the mean bias between matched affected and nonaffected children and checked if the bias was substantially reduced from before to after matching.
We also conducted a likelihood ratio test of the joint significance of the propensity score estimation covariates separately on the original data and the matched sample. If matching quality was high (i.e., systematic group differences were reduced substantially), the covariates would not be able to explain the affected status (born to an adolescent mother) in the sample of matched children anymore. As a result, the likelihood ratio test would fail to reject the null hypothesis after matching.
Another test of matching quality involved estimating the pseudo-  Table 1 presents the summary statistics of the study sample. Figure 1 presents Kernel density plots of the distribution of outcome variables in IHDS and IHDS-2, along with the change over time, separately for affected and nonaffected children. Children born to adolescent mothers (ages 19 years or below) were 0.02 m shorter at both ages as compared with those born to older mothers (ages 20 years and above).

Characteristics of the study sample
They also weighed 0.24 kg less at ages 0-5 years and 0.84 kg less at ages 6-12 years. All differences were statistically significant. There were some significant differences in the background characteristics of TA B L E 1 Summary statistics of study sample.

Regression results
Regression results for both sexes are presented in Table 2

PSM estimates
Results from PSM analysis are presented in Table 3 The estimates were similar when we used radius and three nearest neighbor matching algorithms. They are presented in Table 3 and not discussed here separately. Findings from matching quality tests are presented in Tables S3 and S4. There were statistically significant differences (percent bias) in almost all covariates-except age, schooling of household head, and some social group or religion indicatorsbetween the affected and nonaffected groups before matching. There was no significant percent bias after matching in any of the covariates.
The mean percentage bias reduced from 7.5% to 1.2% in all models.
The pseudo-R 2 and likelihood ratio tests of PSM validity were also satisfied. The pseudo-R 2 reduced substantially from 0.073 before matching to 0.003-0.004 after matching. In all models, the likelihood ratio test failed to reject the null hypothesis of no joint significance after matching. Together, these results indicate that PSM produced consistent and valid estimates of the link between adolescent motherhood and child growth.

DISCUSSION
We used nationally representative longitudinal data and found that well as the health of the next generation. For example, higher maternal height is known to be positively associated with higher birthweight and child growth. [7][8][9][10][11]18 A subset of the childhood development literature is devoted to catch-up growth, focusing mainly on an accelerated increase in height or weight during the first 2 years of life. A recent systematic review examined the effect of catch-up growth for low birthweight infants in two high-income countries and five LMICs. 35 Accelerated weight increase was linked to higher body mass and glucose metabolism and lower mortality and hospitalization rates in the short-term (mean age of 13 months). Catch-up growth was also linked to longer-term (mean age of 11 years) higher body mass, body mass index, and cholesterol levels.
Our study contributes to the understanding of long-term trajectories of growth which until now remain inadequately investigated in  36 As a result, the negative effect of adolescent motherhood on child growth in our study may be underestimated.
Selective attrition of samples between IHDS and IHDS-2 may also distort our results. For example, children born to adolescent mothers in IHDS may have higher infant or child mortality rates than those born to older mothers. This selection effect could imply that only children with higher growth rates are captured in IHDS-2, while the rest are lost to attrition because of death. 37 As a result, growth failure and catchup failure among children of adolescent mothers in IHDS-2 may be underestimated.
Finally, adolescent motherhood may be associated with or a mediating factor for lower maternal height and education. Early menarche and pregnancy have been linked with growth cessation in adolescents, and in LMICs, they may also increase the risk of dropping out of school, both of which could be negatively associated with offspring growth. 31,[38][39][40][41][42][43] Therefore, our model should ideally account for only the factors which were predetermined at the time of motherhood, such as socioeconomic and demographic characteristics of the woman's natal family.
Data from women's natal family were not collected by the IHDS surveys. However, considering that marriages in India primarily occur within the same religion, caste, and economic strata, [44][45][46][47][48] in-law family characteristics should serve as adequate substitutes for natal family characteristics. Furthermore, we tested if our findings were sensitive to factors that were not predetermined at the time of motherhood.
We repeated our analysis after including the mother's height and completed years of schooling (from IHDS data) among covariates of the model. The results showed negligible to no difference from the original model and, therefore, are not presented separately.
Despite these limitations, our study makes a significant contribution to the understanding of long-term trajectories of growth among children born to adolescent mothers in India, showing a widening gap over time with respect to children of older mothers. Our findings were robust to an array of sensitivity analyses involving regression and matching methods. We accounted for a wide range of potential confounding factors at the child, mother, household, and community levels.
Adolescent childbearing in India has reduced substantially from 67 births per 1000 women aged 15-19 years in 2000 to 10 births in 2020. 49 While improvements in socioeconomic status, female education, intrafamily bargaining power, and economic empowerment may have contributed to this rapid decline, 50 key systemic barriers to preventing adolescent pregnancy, such as child marriages, are yet to be fully addressed. Almost a quarter of young Indian women reported getting married before the age of 18 years in 2020. 6 The ongoing COVID-19 pandemic has exacerbated this challenge by putting an additional 5-10 million girls worldwide at risk of child marriage. 51,52 In addition to policies for reducing the incidence of child marriage and adolescent pregnancies, a robust set of programs are also required for addressing undernutrition among the children of adolescent mothers. Integrated Child Development Services-which is India's national program of community-based supplemental nutrition for pregnant women, new mothers, adolescent girls, and children under the age of 6 years-could be used for targeted nutrition delivery and accelerated catch-up growth. Early childbearing is also a major driver of school dropout among adolescent girls in India and other LMICs. 43,53 Programs and policies that encourage and support re-entry into the schooling system and retention for adolescent mothers must also be considered. Future research should evaluate the effectiveness of these and other policies for addressing adolescent childbearing in India.

AUTHOR CONTRIBUTIONS
A.N. designed the study, conducted the analysis, and wrote the first version of the manuscript. A.N., F.Z., K.A., N.H., and T.D.N. interpreted the findings and critically evaluated and edited the manuscript. All authors approved the final draft for publication.

COMPETING INTERESTS
The authors declare no competing interests.

DATA AVAILABILITY STATEMENT
All data are publicly available and can be accessed through the India Human Development Survey website https://ihds.umd.edu/.