Using three indicators to understand the parity-specific contribution of adolescent childbearing to all births

Indicator definitions and data requirements

We define the indicator of the contribution of live births to adolescents (up to age 20 years) to live births to women in all age groups, and, separately, the contribution of first-order vs second-order and higher-order births, as shown in table 1.

Country selection and data sources

For LMICs, we used the raw data from Demographic and Health Surveys (DHS) and included countries which conducted at least one survey in each of the three time periods—1990–1999, 2000–2009 and 2010–2015; 30 countries met these criteria (table 2). If more than three surveys were available, we included the earliest in the 1990–1999 period and the most recent 2010–2015, with a survey in the 2000–2009 period that was most equally situated between the first and most recent. This sample included countries covering a range of geographical regions, income levels, data types (aggregate and raw) that had easily accessible data to describe trends over time (table 2). For comparative purposes, we included two high-income countries in Europe with aggregate data available from vital registration: Finland (data available for all three indicators) and Slovakia (data for indicator one only).

DHS data analysis

DHS are cross-sectional nationally representative household surveys and use model questionnaires which are adapted to each country’s circumstances. The DHS use a multilevel, cluster sampling survey design; survey weights are needed in analysis to adjust for this and for non-response. We used the children ever born section of the women’s questionnaire. Respondents aged 15–49 years were included in the analysis; in 16 of the 90 surveys, only ever-married women were sampled (online supplementary material 1). We calculated the numbers of live births for the numerators and denominators for the 10-year period before each survey, because of likely under-reporting of early adolescent births by girls and young women who were in the 15–19 years age group at the time of the survey.34

Multiple births (twins, triplets, etc) were considered one delivery “event” and were thus classified as the same birth order. In the calculation of numerators and denominators, births resulting in multiples contributed only one live birth to avoid multiple live births with the same birth order number. Hence, a woman who had twins followed by a singleton, while being of parity 3, contributed one live birth of first birth-order and one live birth of second/higher birth-order to the analysis. While this treatment slightly underestimates the absolute numbers of live births, it does not bias the indicators assuming that women’s risk of having live births resulting in twins, and higher order multiples is not significantly different between adolescents and older women.

The three indicators were calculated overall and by age thresholds and exact ages, for each country and each survey. Data were disaggregated by cumulative thresholds and exact age groups. Absolute and relative changes in indicators for the three cumulative thresholds are described over time. All DHS analysis was conducted in Stata/SE14, with adjustment for survey design (weights, clusters and stratification). The DHS receive government permission, use informed consent and assure respondents of confidentiality.

Most recent survey data

figure 1Figure 1 shows, for the most recent survey data, the contribution of adolescent births to all live births (indicator 1), to first-order live births (indicator 2) and to second-order and higher-order live births (indicator 3), by country, for each annual age increment. The median value for the percentage of all live births to adolescents <20 years of age in the 30 LMICs was 17.5% (table 3). Only three countries (Jordan, Indonesia and Rwanda) had levels below 10%. At the other end of the spectrum, Bangladesh had the highest level at 33.1%, and an additional nine countries had levels above 20%. The percentage of all live births that were among the youngest group of adolescents (<16 years) ranged between 0.1% (Jordan) and 6.3% (Bangladesh), with a median value of 1.5%. In all LMICs, there was a steep rise in the percentage of all live births occurring to adolescents with each annual age threshold segment. At 6.2%, the contribution of adolescent births (<20 years) to all births in Slovakia was similar to the two LMIC countries with the lowest percentages; Rwanda (6.5%) and Jordan (6.0%). However, in Finland, the percentage was much lower at 1.7%. The estimates of the percentages of all, first-order, and second-order and higher order births occurring to adolescents by exact ages in the included countries, and surveys are shown in online supplementary material 2.

Request permission

Looking at the contribution of adolescents to first-order births for the most recent survey, table 4 shows that more than half of all first-order live births are to women <20 years of age in 14 of the 30 LMICs assessed, 13 of them in sub-Saharan Africa. In the 30 LMICs, the median value of the percentage of all first-order live births occurring to adolescents <16 years was 6.1% and to adolescents <20 years was 48.6%. The country with the highest contribution of adolescent first-order births to all first-order births, in both these cumulative age thresholds, was Chad (21.4% and 72.8%, respectively). The lowest percentage of first-order births occurring to adolescents of all age thresholds among the 30 LMICs was in Jordan (0.5% and 19.0%, respectively), and Rwanda was a low outlier in sub-Saharan Africa with 0.7% and 21.5%, respectively. In comparison, in Finland in 2015, the percentage of first-order births occurring to adolescents <16 years was only 0.1% and to adolescents< 20 years was only 3.8%. Parity disaggregated data were not available in Slovakia.

Table 5 shows the contribution of adolescent second-order and higher-order births to all second-order and higher-order births. In the 30 LMICs, the median value for the most recent survey for adolescents <20 years was 6.3% (lowest in Rwanda at 0.8% and highest in Bangladesh and Chad at 12.5% and 12.1%, respectively). The estimates for the youngest adolescents (<16 years) ranged from ≤0.01% in 16 countries to 0.8% in Bangladesh, Chad and Mali, with a median value of 0.1% across the 30 included LMICs. There were no second-order or higher-order live births reported to girls aged <16 in Finland in 2015, and the contribution of adolescents <20 years to second-order and higher-order births in Finland was negligible at 0.3% in 2015.

Time trends

The change in the three proposed indicators over time is shown for all 32 countries in tables 3–5. Comparing these trends for adolescents <20 years of age between the earliest and most recent survey, we saw seven patterns emerge across the 30 LMICs (table 6).

Pattern 1, which was most common (17 countries), was defined by a reduction in the percentage of adolescent births among all births, first births and subsequent births to women <20 years. In some countries, such as Senegal, Uganda and Niger, these reductions were very marked. For these countries with a very marked reduction, it appears that not just the adolescent and percentage of women having a first birth before the age of 20 are declining, but that adolescent fertility is falling at a greater rate than for other age groups.

In Senegal, for example, there has been a marked fall in the proportion of adolescent births in all three indicators. This is reflected by a fall in the proportion of women who had experienced a first birth before 20 years. In 1992, 55.7% of women aged 20–49 reported that they had given birth before the age of 20, compared with only 38.5% in 2014, a 31% decline (table 7). The median age at first birth in the age group 25–29 years increased by over 2 years in the same time period, from 19.3 to 21.4. The ASFR for 15–19 year olds fell sharply (by 29% between 1992 and 2014), and reduced at a greater rate than in other age groups (the general fertility rate fell by only 17%). A further factor that influenced the decrease in the proportion of births to adolescents was that the proportion of all women of reproductive age 15–49 years who were aged 15–19 years fell from 25% in 1990 to 21% in 2015.35 Similar large reductions in the three indicators were found in Egypt, Jordan, Indonesia, Pakistan and Rwanda. Finland also belonged in this pattern of consistent progressive decline.

There was one country each in pattern 2 (Benin) and 3 (Chad). Pattern 2 was defined as a country where the overall percentage of adolescent births among all live births has decreased, the percentage of adolescent first births among all first births also decreased, but the percentage of adolescent higher-order births among all higher-order births increased. Pattern 3 described a situation where the overall percentage of adolescent births among all births and contribution of adolescent higher-order births to all higher-order births decreased, but where the percentage of adolescent first births to all first births increased. Chad is also the only country in this sample of countries that saw a rise, although modest, in the percentage of first births occurring to adolescents. This is confirmed by the decline in the median age at first birth between the first and last surveys, from 18.2 in 1996–1997 to 17.9 in 2014–2015.

Colombia and the Philippines are the only countries that saw a rise in the contribution of adolescent births among all three indicators (pattern 4). This reflects a rise in the percentage of women reporting a first birth before the age of 20 years and the 15–19 years ASFR, but the effect is magnified by the fact that fertility in older age groups has declined. In the case of the Philippines, the rise is greatest for first births but is high for both first and subsequent births in Colombia.

In four countries (Nepal, Guinea, Mali and the Dominican Republic), the overall percentage of births among adolescents <20 years (indicator 1) among all live births increased, but there were discordant trends among first-order and higher-order births. Nepal, Guinea and Mali showed a decrease in the contribution of adolescent to first-order births, but an increase in the adolescent contribution to second-order and higher-order births (pattern 5). The opposite was seen in the Dominican Republic which saw an increase in the contribution of adolescent first-order births, but a decrease in the adolescent contribution to second-order and higher-order births (pattern 6). Patterns of adolescent fertility may be partly responsible for these patterns, but wider fertility processes that affected the denominator, such as falling general fertility which would affect the proportion of all births that are first or higher-order, may well also have contributed. In the case of the Dominican Republic, the result is largely due to adolescent fertility rates increasing slightly while the general fertility rate declined by a third over this time period (table 7).

An apparent anomaly is the last pattern 7 (Namibia, Haiti, Peru, Bangladesh, and Malawi), where the overall proportion of births to adolescents <20 years among all live births has increased, but both of the component indicators have fallen. In all these countries except Peru, the ASFR 15–19 years fell, but by a smaller extent than the general fertility rate. In Peru, the 15–19 years ASFR increased slightly, while general fertility declined substantially. This pattern is consistent with the overall increase in the proportion of all births that are to adolescent mothers. By contrast, the fact that the proportion of both first births and higher-order births fell could have been driven by changes over time in the percentage of all births that were first births, as is the case in the context of declining overall fertility. Hence, the proportion of all births that occur to adolescents will be driven higher because a higher proportion of adolescent births are first births.

Summary

In this paper, we propose an additional set of three indicators which capture the contribution of adolescent births to all births, and separately for first-order and second/higher-order births. These indicators help to place adolescent fertility in a broader demographic context of overall fertility and childbearing, and can thus serve as a basis for improving the focus on and provision of reproductive/maternal health services to adolescent girls and their newborns. We found that in the 30 included LMICs, a substantial proportion of all live births occurred to adolescents. The extent of this phenomenon varied across countries, with a median of 18% for adolescents <20 years. This was mainly due to the high contribution of adolescent first-order births to all first-order births, showing a median of 49% across the 30 LMICs. However, even among second-order and higher-order births, adolescent childbearing was appreciable (median across countries of 6%).

Over the period under examination, the proportion of adolescent births among all live births declined in the majority of the countries. However, as the number and complexity of patterns of change over time showed, the interpretation of changes in these indicators over time is complex because they are influenced by demographic factors other than variation in the adolescent fertility rate and percentage of women having their first birth before the age of 20 years. Due to being expressed as a proportion of all births, the proposed indicators are influenced by the levels of and trends in fertility rates not just among adolescents but also among older age groups, as well as by changes in the proportion of all women of reproductive age who are below the age of 20 years.

Limitations

The main strength of this set of proposed indicators is that they can be constructed from available secondary data. However, like indicators of adolescent and general fertility levels based on population-representative surveys, they have limitations. The first issue is the focus on live births—mainly based on data availability in LMICs—leads to the exclusion of pregnancies that resulted in miscarriage, induced abortion or stillbirth. This is a problem as adolescent women are at higher risk of negative perinatal outcomes, so a lower proportion of their pregnancies result in live births. Second, population-based surveys such as the DHS, only sample women aged 15–49 years old at time of the survey, therefore missing live births to girls who are younger than 15 years at the time of the survey. This would have likely resulted in a slight underestimation of the contribution of adolescent live births to all live births.

Third, we used the period 0–9 years before each survey for indicator estimation due to age-related biases in reporting live births, particularly among young adolescents.25 However, this also means that the estimates cover a longer period (not providing the most up-to-date picture even when the most recent survey is used), and that these 10-year survey recall periods overlap in some countries included in this analysis. Last, we recognise that as the proposed indicators are affected by a number of demographic processes, it might be difficult to interpret trends over time. This relates particularly to the birth order disaggregated component indicators (indicators 2 and 3), which are therefore probably best only used as cross-sectional indicators for most recent data available. We reiterate that these indicators are not suitable for tracking and monitoring adolescent fertility per se as they are influenced by complex population changes over time and are to be used in conjuction with other commonly used indicators of adolescent fertility.

Interpretation and case for including proposed indicators

Notwithstanding these limitations, an understanding of the proportion of all births that are to adolescents through the set of three indicators proposed in this paper is important for two main reasons. First, these proposed indicators provide valuable demographic information. For example, when considered in conjunction with overall adolescent fertility, they give an indication of adolescent fertility in relation to fertility rates in older women. As we show, a number of countries within Latin America and parts of Asia have lowered overall fertility without making similar gains in adolescent fertility. Such enriched understanding of demographic processes should be used for planning sexual and reproductive healthcare provision, including contraception, safe abortion care and maternal healthcare. In this regard, the proposed set of indicators can be used to calculate the total number of births expected in a given year, disaggregated by age, parity and region.

Second, by providing a more health system perspective on adolescent reproduction than currently available indicators of fertility, these indicators can support the planning and delivery of appropriate and good quality reproductive and maternal care to adolescents and their infants. For example, it can be impactful to note that three of every four first-order births in Bangladesh or Chad are to adolescents <20 years of age. Because the majority of adolescent births are first order, our second indicator may be seen as more important to policy and practice. However, second-order and higher-order births among adolescents should not be disregarded as these indicate, to a certain extent, a failure to prevent repeat pregnancies among this vulnerable population. Recent evidence also suggests that the increased risk of neonatal mortality for second-order and higher-order births is actually greater than for first-order births when compared with similar parities in women in their 20s.27 In most of the countries, we included in our analysis, the contribution of adolescent (<20 years) second-order and higher-order births to all second-order and higher-order births is relatively small (<10%), but the use of this indicator has highlighted several countries where more than 1 in 10 second-order and higher-order births is to adolescents (Bangladesh, Chad, Mali, Niger). This is not insignificant from the point of view of providing pregnancy, intrapartum and postpartum care, and it gives valuable information to policy makers on the failure to prevent these pregnancies.

Health service provision to pregnant, delivering and postpartum adolescents can be improved by recognising their biological and social needs and vulnerabilities. While recent studies show that the increased risk of maternal mortality associated with adolescent motherhood is not as large or as consistent across countries as previously thought, women aged 15–19 years still have higher MMRs than women in their 20s in most countries.14 Further evidence suggests that pregnant adolescents are at greater risk of developing certain conditions such as pre-eclampsia or eclampsia and fistula resulting from obstructed labour compared with older women.36 37 While adolescents are more likely to have malaria infection during pregnancy than older women,38 a study in Uganda found that pregnant adolescents are less able to recognise the risks posed by this condition.39

Adolescents also face specific barriers when seeking maternal healthcare, and it is particularly important for countries with a high proportion of adolescent pregnancies to ensure that the services provided are adolescent friendly. In select countries, there is evidence that adolescents make less use of antenatal care and skilled care at delivery.40–42 In addition, a study in West Africa found that adolescents were less likely to commence antenatal care within the first trimester, to attend four or more antenatal visits, and reported receiving fewer components of care.43 While in some cases, poorer access to appropriate levels of maternal healthcare may reflect socioeconomic factors, studies also suggest that many adolescents face considerable stigma from healthcare providers, which presents a barrier to seeking care.40 Postpartum contraception and other effective interventions to reduce rapid repeat pregnancies (very common in adolescents) are also an important factor.44