We found an increased risk of infant mortality among children born to adolescent mothers compared to those who were born to mothers aged 20-29 years, after socioeconomic factors and parity were controlled for. This excess was due to postneonatal deaths. This association was attenuated when mediators were introduced into the model, suggesting that much of the excessive risk among infants from adolescent mothers is explained by behavioral and health care characteristics. Maternal age was not associated with perinatal mortality or either of its components - fetal or early neonatal deaths.
When comparing our results to the literature, it is important to consider which confounding variables were adjusted for in each study, as well as the characteristics of the populations where the studies were carried out. In terms of confounding factors, one should ideally include valid and multiple indicators of socioeconomic position (SEP), rather than rely on single indicators such as income, education, occupation or assets. Lack of thorough adjustment for SEP may lead to residual confounding given the strong association between poverty and adolescent childbearing. In addition, it is important to present analyses adjusted for mediating factors - such as antenatal care, weight gain during pregnancy, type of delivery, birth weight or breastfeeding - separately from those adjusted for confounding factors, as these models have different causal interpretations.
Two review articles are available in the literature. In 2001, Cunnington's systematic review identified four studies from high-income countries which assessed the effect of maternal age on neonatal mortality, of which two reported associations . In 2009, a new review identified three papers on the same topic, of which one was included in the 2001 review . All three papers reported increased odds of neonatal deaths among adolescent mothers when compared with mothers aged 20-29 years (from 1.1 to 2.7) . All studies were carried out in high-income countries. Three out of four studies showing associations used educational level as a proxy for SEP and two of them also controlled for adequacy of prenatal care and for tobacco consumption [9, 13]. No reviews of the effect of maternal age on infant mortality (rather than neonatal deaths) were located.
Through a systematic search of PubMed since the 1960's, we identified 22 studies reporting on maternal age and risk of fetal or infant mortality. In general, studies which failed to adjust for SEP tended to report that adolescent childbearing increased the risk of fetal [8, 26, 28], perinatal , neonatal [10, 12, 18, 21, 29], postneonatal [21, 23, 29, 30], and infant mortality [12, 16, 37], whereas these outcomes were not associated to maternal age in analysis with adjustment for SEP [15, 17, 27, 38]. Authors who adjusted for several socioeconomic variables reported ORs from 1.1 to 1.5 [15, 17, 27] and those who used just educational level as indicator of SEP or just ethnicity and marital status reported ORs from 1.6 to 3.0 [12, 18, 21, 26, 28–30]. Sixteen of the 22 studies were from high-income countries [8, 10, 12, 15, 16, 18, 19, 21–23, 26, 28–30, 37, 38]. Comparing results of studies from high-income and those from low and middle-income countries, it appears that associations with maternal age tend to be stronger in the former for fetal mortality and neonatal mortality [8, 10, 12, 18, 21, 26, 28, 29]. Differences between studies settings may be due to baseline risk of mortality, social characteristics of adolescents, type of health care system and social support available in each setting, among other factors. Methodological differences may also account for discrepancies in findings, including the fact that several studies from high-income countries relied on secondary databases with very large sample sizes but possibly lower data quality.
In addition, some studies [14, 19] may have failed to detect associations between adolescent motherhood and offspring mortality because of adjustment for mediators such as birth weight, gestational age or medical and behavioral risk factors during pregnancy, which in fact may be a consequence of adolescent pregnancy rather than true confounding factors. Analyses adjusted for mediating factors are important but their interpretation is different from analyses adjusted for true confounders.
An interesting finding in our analyses was the increase in the effect of adolescent motherhood on infant mortality as child age increases from 1 to 12 months, which supports the social and environmental explanations for this relationship. If the excess of mortality among children of adolescent mothers is due to maternal physiological immaturity, then the effect of maternal age should be more pronounced for periods of the children closer to the time of birth, or equally pronounced across all child ages. Similar findings were described in other studies [17, 21, 30]. The fact that the association with post-natal mortality completely disappeared after adjustment for factors such as weight gain during pregnancy, antenatal care and breastfeeding is particularly important because it suggests potential areas for interventions to reduce mortality among offspring of pregnant adolescents.
Major strengths of this study are the population-based sample from birth cohort studies, the very high (over 99%) response rates at baseline, the detailed assessment of maternal characteristics, and active surveillance for fetal and infant deaths. In spite of the large sample sizes, deaths are rare events and some of our analyses - particularly for mothers aged 12-15 years - had low power, and this group had to be pooled with older adolescents, whose risk may be considerably lower. Further studies would be needed to replicate this analysis in a larger sample and in a similar setting.