Significance

Despite downward trends over time, the United States has the highest rate of teen pregnancy among western countries. Given that a repeat pregnancy to teen mothers compounds the already great challenges faced by an adolescent and her family, it is important to identify the predictors that might reduce the odds of a repeat pregnancy, and to use this knowledge to inform intervention efforts. This study is among the first to examine distal factors—school, family, peer, and religious ties—and their association with rapid repeat pregnancies (RRP) among adolescent mothers using a nationally representative sample of adolescents.

Introduction

Despite declining rates for almost two decades, and a 2014 record low of 24.2 births per 1000 adolescents aged 15-19 (Hamilton et al. 2015), the United States (U.S.) continues to have the highest adolescent birth rate among most other developed countries (United Nations Statistics Division 2015). Although some scholars have proposed that not all adolescent births are harmful (Furstenberg 2007; SmithBattle 2009), most research suggests that secondary births remain a public health concern. A rapid repeat pregnancy (RRP) among adolescent mothers (aged 19 and younger), defined as a pregnancy occurring within 24 months of a previous birth, compounds the challenges faced by an adolescent mother and her family. Historically, RRP rates among adolescents in the U.S. have been high; 30 % of adolescent mothers become pregnant in their first postpartum year and another 25 to 50 % become pregnant in their second year (Mosher et al. 2012).

Adolescent childbearing has negative consequences for mother and child; both generally complete less education, have lower socioeconomic status, and have poorer health compared to individuals without an adolescent birth (Raneri and Wiemann 2007; Rosengard 2009). These challenges increase with a second birth. There has been considerable research examining the determinants and consequences of early motherhood (Klein 2005; Klerman 2004). However, research on RRP is limited, especially using nationally representative samples (Klein 2005; Klerman 2004). Of the research that exists, many studies focus on sociodemographics, contraception, pregnancy intentions, and other proximal individual level characteristics (Boardman et al. 2006; Crittenden et al. 2009; Gillmore et al. 1997; Kelly et al. 2005; Kershaw et al. 2003; Klerman 2004; Nelson 1990; Pfitzner 2003; Stevens-Simon et al. 2005). Although clearly important, these studies ignore distal, sociocontextual factors that also influence behavior.

Of the few studies that have examined adolescents’ social context, factors most consistently associated with RRP include attachment to school, family and/or peers, and church. Adolescent mothers who drop out of school before (Kalmuss and Namerow 1994; Manlove et al. 2000; Van Horne et al. 2009) or directly after (Barnet et al. 2004; Kalmuss and Namerow 1994; Raneri and Wiemann 2007; Stevens-Simon et al. 1986; Stevens-Simon 1996) first birth are more likely to have an RRP. Raneri and Wiemann (2007) found one of the strongest community-level predictors of RRP was “not being in school at 3 months postpartum (Raneri and Wiemann 2007).” Given child rearing responsibilities, some adolescent mothers may find it difficult to maintain student status after the initial birth. Further, for some young women, family is a higher priority than school (Bull and Hogue 1998), as low educational and occupational aspirations are associated with RRP (Barnet et al. 2010; Klerman 2004; Manlove et al. 2000; Raneri and Wiemann 2007; Stevens-Simon et al. 1998; Van Horne et al. 2009).

Previous research also suggests that family and peer attachment is associated with likelihood of RRP among adolescents, especially poor maternal relationships (Bull and Hogue 1998) and the lack of family (Raneri and Wiemann 2007; Stevens-Simon, 1996) or peer support (Rigsby 1998). Adolescent mothers who receive inadequate social support from their families or peer networks could see childbearing as a pathway to a better sense of belonging and empowerment. A lack of family and peer connectedness may encourage an adolescent mother to seek sexual intimacy or to bear additional children to fulfill her desire to be loved (Raneri and Wiemann 2007; Rigsby 1998; Stevens-Simon et al. 2001). Adolescents who are able to seek and receive emotional and financial support from their parents during a first pregnancy are less likely to become pregnant again (Klerman 2004; Manlove et al. 2000).

Evidence regarding the potential role of attachment to religion in RRP is limited, and findings are inconsistent (Klerman 2004; Manlove et al. 2000). Using the 1988 National Education Longitudinal Study (NELS), Manlove et al. (2000) found that religious involvement (attending church and school religious clubs more than once a month) is unrelated to “closely spaced births” (Manlove et al. 2000). Contrastingly, Van Horne et al. (2009) found that low religiosity (never attending weekly services) is associated with never or only sometimes using condoms among adolescent mothers (Van Horne et al. 2009). Failure to find other associations between religiosity and RRP in previous studies may be an artifact of limited measures of religiosity. Focusing on attending church or participating in school religious clubs ignores subjective and private religious experiences, such as how often one prays, that may be important. Studies using more comprehensive measures have found that greater religiosity is associated with a reduced likelihood of first adolescent pregnancy (Jeynes 2003); it seems plausible that similar patterns are associated with RRP.

Despite increased attention to the determinants of repeat pregnancy among adolescent mothers, the current literature is limited in that many studies do not examine diverse sociocontextual predictors of adolescent RRP. Further, of the studies noted above, only one (Manlove et al. 2000) utilized a nationally representative sample of U.S. adolescents; most studies rely on convenience samples. Providing evidence about contributors to RRP, based on a nationally representative and diverse sample, would be useful for the development of comprehensive sexual health promotion efforts for adolescent mothers in the U.S.

Few extant studies on RRP applied theory. We use social development theory (Catalano and Hawkins 1996) to select distal potential determinants of adolescent RRP. Social development theory suggests that social bonds, consisting of attachment to others and commitment to conventional values and behaviors, promotes healthy development and protects against risk behaviors, including sexual risk (Catalano et al. 2004). Conversely, social rewards from associating with antisocial peers can promote risk behaviors (Catalano et al. 2004). Further, key sources of social ties—family, peers, school, and community—influence adolescent behavior by modeling socially desired behaviors, and providing opportunities for youth to develop social skills that build resiliency (Catalano et al. 2004). Given complex biopsychosocial changes during adolescence and challenging parenting responsibilities, positive social attachments may help adolescent mothers manage competing responsibilities and promote healthy behavior.

To address past study limitations, we examined whether attachments to conventional institutions of family, peers, school, and religion/church—factors highlighted in social development theory—are associated with RRP among adolescent mothers in a nationally representative sample of U.S. adolescents. We hypothesized that adolescent mothers with strong social bonds (attachment) to family, peers, school, and religion are less likely to experience an RRP.

Methods

Data

We used data from Waves I and IV of the National Longitudinal Study of Adolescent to Adult Health (Add Health), an ongoing prospective study of a U.S. nationally representative probability sample of 20,745 adolescents in grades 7–12 during the 1994–1995 school year [see Harris (2013) for study design details]. To date, one in-school and four in-home interviews have been completed; data for present analyses came from the Wave I/baseline in-home adolescent and parent (usually the resident mother) interviews, and the Wave IV respondent in-home interview (2008; respondents aged 24–32). Add Health procedures were approved by the University of North Carolina, Chapel Hill IRB. Present analyses were deemed exempt and were conducted in accord with prevailing ethical principles.

Inclusion criteria for the current study were participation in Waves I and IV (n = 15,701), valid sampling weight (n = 14,800), female biological sex (n = 8352), reporting at least one birth before age 20 (n = 1259), and non-missing data on analytic variables (n = 1158).

Measures

Because of their sensitive nature, all pregnancy and sexual behavior questions were self-administered using computer-assisted self-interviewing (CASI).

Rapid Repeat Pregnancy (RRP)

A complete pregnancy and childbearing history for each respondent was collected during the Wave IV in-home interview. RRP was defined as a pregnancy that occurred within 24 months of the first adolescent birth.

Unless otherwise indicated, all attachment variables described below are self-reported in the adolescent Wave I/baseline interview. Variables were selected based on use in prior studies of the distal/sociocontexual factors associated with adolescent pregnancy (e.g., Raneri and Wiemann 2007) and relevance to social development theory (Catalano and Hawkins 1996).

Attachment to Family and Peers

Parent-adolescent relationship quality was measured by summing answers to four questions about perceptions of closeness, communication, relationship satisfaction and warmth with each resident parent. In households with two parents, we selected the higher of the two scores; scores ranged from 4 to 20, and higher values reflect better relationship quality (Cronbach’s alpha = 0.85).

Adolescents were asked about their perceptions of their mothers’ attitudes toward their engaging in sexual activity and their education attainment. Three items measuring adolescent perception of maternal disapproval of sex were, “How would your mother feel about your having sex at this time in your life?”; “How would your mother feel about your having sex with someone who was special to you and whom you knew well-like a steady boyfriend?”; and “If you had sexual intercourse, it would upset your mother.” Responses ranged from 1 to 5, from “strongly approve/agree” to “strongly disapprove/disagree.” Item scores were summed; higher values indicate greater perceived maternal disapproval of sex (range: 3-15; Cronbach’s alpha = 0.95).

Two questions measuring adolescent perception of maternal attitudes toward education were, “On a scale of 1 to 5, where 1 is low and 5 is high, how disappointed would she [resident mother] be if you did not graduate from high school [graduate from college]?” Item scores were summed; higher values reflect greater perceived maternal disappointment (range: 2–10; Cronbach’s alpha = 0.71). Anticipated social consequences of sex were evaluated by four items: “If you had sexual intercourse, your friends would respect you more;” “If you had sexual intercourse, your partner would lose respect for you;” “If you had sexual intercourse, it would make you more attractive to men/women;” and “If you had sexual intercourse, you would feel less lonely.” Item responses ranged from 1 to 5, from “strongly disagree” to “strongly agree.” Scores were summed such that higher values indicate fewer negative social consequences (and more anticipated rewards) of having sex (range: 3-15; Cronbach’s alpha = 0.68).

Attachment to School

School connectedness (Cronbach’s alpha = 0.78) was a summary score measuring adolescents’ feelings of being close to people at school, feeling a part of school, feeling happy to be at school, and feeling safe at school. Each item was measured on a five-point Likert scale and summed to create an overall measure of school connectedness (range = 4–20; higher values indicate greater attachment).

For two indicators of adolescent educational aspirations, the desire to attend college and the perceived likelihood of attending college, scores on each item ranged from 1 to 5, “strongly disagree” to “strongly agree.” Lastly, we included dichotomous indicators for ever repeated a grade, received an out-of-school suspension, and expulsion from school.

Attachment to Religion/Church

Attachments to religion reflect private and public indicators, including how important religion is to the adolescent (scores range from 1 to 5, “not important at all” to “very important”), frequency of prayer (“at least once a day [referent],” “once a week,” “once a month,” “less than once a month” and “never”), frequency of religious services attendance, and participation in church-related youth activities in the past 12 months. Response options for the latter two were “once a week or more (referent),” “once a month or more,” “less than once a month” and “never.”

Control Variables

(Self-reported in the Wave I adolescent in-home interview, unless otherwise indicated): Race/ethnicity, categorized as non-Hispanic white (referent), Hispanic (any race), non-Hispanic black, and non-Hispanic Asian/Pacific Islander, and other. Age at Wave IV interview is a continuous variable in years. Family of origin structure, categorized as living with two biological parents (referent), other two parent household, single mother, single father, and all other structures. Parental education attainment, used as a proxy for socioeconomic status, is the highest level of education obtained by either of the respondent’s parents or caregivers (less than high school; high school graduate/general education diploma; some college or post-high school business, trade, or vocational school; or college graduate or more [reference category]) and was reported by the resident mother in the parent wave I in-home interview (supplemented by adolescent report if parent information missing).

Neighborhood poverty, the proportion of families in the respondents’ Census block group with dependents younger than 18 years and income below the federal poverty level (FPL) in 1989, was categorized as low (<11.6 % of families below FPL; referent), medium (between 11.6 and 23.9 % below FPL), and high (>23.9 % below FPL) (Billy et al. 1998).

Neighborhood connectedness, a protective factor for adolescent pregnancy and childbearing (Desikan 2011; Small and Luster 1994), was indexed by two questions: “People in this neighborhood look out for one another” and “Do you usually feel safe in your neighborhood?” (affirmative responses coded 1, negative coded 0). Responses were summed. Scores ranged from 0 to 2; higher values indicate greater connectedness.

We also controlled for age at first pregnancy, adoption status of first child, and marriage status at time of adolescent birth. Further, we controlled for birth control use before first pregnancy, based on respondent report about use in the month before first pregnancy (yes/no). Wanted first birth was based on the question, “Thinking back to the time just before this pregnancy with {initials}, did you want to have a child then?” (yes/no).

Additionally, we controlled for self-reported childhood sexual abuse (“Did a parent or other adult caregiver touch you in a sexual way, force you to touch him or her in a sexual way, or force you to have sexual relations?”), dichotomized as never/≥1 times before age 18). Self-esteem was evaluated by a composite of seven items from Rosenberg’s self-esteem scale (range = 7–35; Cronbach’s alpha = 0.95) (Rosenberg 1965).

Statistical Analyses

Bivariate analyses (Pearson Chi square test for categorical variables and 2-sample t-tests for continuous variables) were used to identify variables potentially associated with RRP. Those significant at p < 0.10 were retained and entered into one multivariate logistic regression model (Barlett et al. 2001) to test whether indicators of attachment to family, peers, school, and church are associated with odds of RRP among adolescent mothers, controlling for all covariates. We used Stata’s (version 14.0) margins command to calculate the predicted probability of an adolescent reporting an RRP for each significant variable in the multivariate model, by averaging probabilities across cases (College Station, TX, USA). All results were adjusted to account for Add Health sampling weights and study design.

Results

Descriptive Analyses

Table 1 presents weighted sample characteristics, by RRP status. Approximately one-third of the adolescent mothers (n = 367) reported an RRP. Of adolescents with an RRP, approximately 53 % were non-Hispanic white, 27.5 % were non-Hispanic black, 14 % were Hispanic, and 2 % were non-Hispanic Asian/pacific islander. On average, adolescent mothers with an RRP reported slightly lower parent-adolescent relationship quality, lower ratings of perceived maternal disapproval of sex, and fewer social consequences (or more rewards) from sex compared to adolescent mothers who did not have an RRP. Additionally, adolescents with an RRP prayed and attended religious services and youth activities more often than adolescents without an RRP.

Table 1 Weighted sample characteristics of adolescent mothers in the United States, by rapid repeat pregnancy (n = 1158)
Full size table

Multivariate Analyses

Table 2 presents results of multivariate logistic regression models. Two measures of family and peer connectedness, and two measures of private and public dimensions of religiosity, were associated with RRP. Adolescent mothers who had stronger parent-adolescent relationships had reduced odds of RRP (adjusted odds ratio [AOR]: 0.83; 95 % confidence interval [CI]: 0.71–0.99), whereas adolescents who anticipated fewer negative social consequences of having sex had increased odds of RRP (AOR: 1.18; 95 % CI: 1.02–1.35). Adolescents who never prayed had increased odds of RRP compared to those who prayed every day (AOR: 1.47; 95 % CI: 1.10–1.96). Adolescents who never participated in church-related youth activities had increased odds of RRP compared to respondents who participated in youth activities once a week (AOR: 1.04; 95 % CI: 1.01–1.07). No school attachment variable was associated with odds of RRP among adolescent mothers in either bivariate or multivariate models.

Table 2 Multivariate logistic regression analysis of the association between attachment to conventional institutions and the likelihood of rapid repeat pregnancy among adolescent mothers, variables selected from significant bivariate associations (n = 1158)
Full size table

Figure 1 shows the predicted probabilities of an RRP for each significant attachment variable in the multivariate logistic regression model, with other variables held at their means. Adolescent mothers who never prayed had a 50 % chance of reporting an RRP compared to 21 % for mothers who prayed once a day or more. Likewise, adolescent mothers who anticipated fewer negative consequences and high social rewards from sex (defined as one standard deviation [SD] [2.68] above the mean [9.89]) had a 37 % chance of having an RRP, compared to 26 % for adolescent mothers anticipating low social rewards (one SD below the mean). Adolescent mothers who reported weak parent-adolescent relationships (one SD [3.05] below mean [17.9]) had a 35 % chance of reporting an RRP, compared to 27 % for adolescent mothers with strong (1 SD above the mean) parental bonds.

Fig. 1
figure 1

Predicted probability of reporting a rapid repeat pregnancy (RRP) among adolescent mothers in the United States (n = 1158). a Predicted probabilities represent illustrative values of the average probability of having an RRP across cases, with the following control variables held at their means: race/ethnicity, family of origin structure, parental education attainment, age at Wave IV, neighborhood poverty, neighborhood connectedness, age at first pregnancy, adoption status of adolescent birth, marriage status at adolescent birth, adoption status of first child, marriage status at adolescent birth, birth control use before first pregnancy, wanted first birth, childhood sexual abuse history, and self-esteem. Only predictors that were statistically significant in the multivariate model are included in the figure. b Strong parent-adolescent relationship quality is 1 SD (3.05) above the mean for the sample (17.9). Parent-adolescent relationship quality measures adolescent perceptions of closeness, communication, relationship satisfaction, and warmth with each resident parent as reported by the adolescent at the Wave I in-home interview, and is a summed scale ranging from 4-20. c Weak parent-adolescent relationship quality is 1 standard deviation (SD) (3.05) below the mean value for the sample (17.9). d Anticipated more negative social consequences of sex is 1 SD (2.68) above the mean value for the sample (9.89). Adolescent anticipated social consequences of sex measures adolescent perceptions of the negative social consequences (and anticipated rewards) from having sex among peers at reported at the Wave I in-home interview, and is a summed scale ranging from 2-10. e Anticipated fewer rewards negative social consequences of sex is 1 SD (2.68) below the mean for the sample (9.89). f Frequency of participation in youth activities within church, as reported by the adolescent at the Wave I in-home interview. g Frequency of prayer, as reported by the adolescent at the Wave I in-home interview

Full size image

Discussion

Guided by social development theory, we examined associations between attachment to conventional institutions (i.e., family, peers, school, and religion/church) and the likelihood of RRP among adolescent mothers. As hypothesized, we found that attachment to conventional institutions is associated with the likelihood of an RRP. Specifically, adolescent mothers who never prayed, never participated in church-related youth activities, or adolescents who anticipated more peer-related social rewards from having sex were more likely to have an RRP. By contrast, adolescent mothers who had strong parent-adolescent bonds were less likely to report an RRP. We did not find significant associations between school attachment and RRP.

Present findings support the importance of private aspects of religion; frequency of prayer yielded the largest probability difference in the likelihood of RRP, a relationship not evident in previous studies. Private religiosity may increase self-esteem, a sense of self-efficacy, or openness to social support (Nonnemaker et al. 2003). Private religiosity might also buffer the effects of life stress and provide a coping mechanism for new adolescent mothers (Wills et al. 2003). Adolescent mothers who never pray may lack these buffers, putting them at elevated risk of RRP. Attachment to public aspects of religion also plays an important role. Involvement in youth activities within a religious setting may provide protection against RRP by motivating adolescents to remain abstinent or to use contraception. Our findings suggest that fostering spirituality or associating with pro-social peers in youth groups may promote positive behaviors, although we could not test that idea directly.

Adolescents who reported better relationships with their parents were also less likely to have an RRP, suggesting the importance of communication between adolescent mothers and their parents, especially in conversations about sexual activity and birth control. Literature addressing early childbearing seldom mentions adolescents in the context of their families, and programs directed at pregnant and childbearing adolescents often do not consider family environments (Barnet et al. 2010). RRP prevention efforts should involve the parents (or guardians) of the adolescent mother, if possible, to strengthen support, communicate effectively, and increase involvement in the adolescent mother’s life (Markham et al. 2010). Prevention efforts could help parents become better prepared to discuss contraception, childrearing, and goal-setting. Further, our findings suggest that adolescent perceptions of peers’ reactions to sexual activity might have an influence on RRP risk, therefore prevention programs that create networks of pro-social peers with positive attitudes about contraception could support adolescent mothers (Markham et al. 2010).

Although no true consensus exists among previous studies largely based on convenience samples, the lack of significant associations between RRP and any school attachment indicator was unexpected. More research using nationally representative samples is needed. In addition, different indicators of school attachment might reveal associations not evident here (Barnet et al. 2010; Klerman 2004; Manlove et al. 2000; Raneri and Wiemann 2007; Stevens-Simon et al. 1998; Van Horne et al. 2009). Future research could examine whether other indicators (e.g., school club participation, career aspirations) affect the likelihood of RRP.

Limitations

Present study strengths include a large, diverse national sample and theoretically driven hypotheses. However, findings should be considered with limitations in mind given the risk of type 1 error. First, we relied on adult retrospective fertility reports, which are subject to recall error and bias. Second, pregnancy and childbearing during adolescence are sensitive topics that may be under-reported. Although Add Health’s use of CASI diminishes these concerns, the influence of social connections cannot be fully understood without qualitative inquiry. Future qualitative studies could explore how connections to family and church protect against RRP. Third, with the exception of marital status at the time of the first adolescent birth, we have not taken into account the respondent’s relationship with the father(s). The quality of this relationship could influence likelihood of RRP. Fourth, 20 % of our analysis sample had their first and/or second pregnancies before Wave I, thus measures of adolescent attachment were collected after the pregnancies. However, we conducted a sensitivity analysis eliminating these respondents, and findings were similar. Finally, not all potential correlates of adolescent RRP were examined, there was no adjustment for multiple comparisons and because of sample size limitations, interactions among variables were not tested. Future research should examine potential moderating and mediating processes.

Conclusion

This study is among the first theoretically grounded investigations to examine adolescent mothers’ family, peer, school, and religious ties and their associations with RRP using nationally representative data. Intervention efforts should be geared toward building on individual needs and environmental strengths. Present findings suggest that adolescent mothers might become pregnant again because they lack a firm connection to supportive social institutions like family and religion. Increasing engagement with these institutions might contribute to postponing second births.