Processes Underlying Mexican-origin Adolescent Mothers’ BMI

Abstract

OBJECTIVE

To examine potential mediating and moderating factors in the longitudinal association between contextual stressors (economic hardship, ethnic discrimination) and subsequent engagement in risky behaviors and Body Mass Index (BMI) of Mexican-origin adolescent mothers.

METHODS

Participants were Mexican-origin adolescent mothers (N = 204) who were recruited from community agencies and high schools in a Southwestern metropolitan area. Contextual stressors and risky behaviors were assessed 3 and 4 years postpartum. Adolescent mothers’ BMI was assessed 5-years postpartum. Path analyses assessed moderated mediation with risky behaviors as a mediator of associations between contextual stressors and BMI and family and friend support as moderators of the mediated pathways.

RESULTS

At low levels of family support, economic hardship at 3-years postpartum positively predicted engagement in risky behaviors at 4-years postpartum, which in turn positively predicted BMI at 5-years postpartum. At high levels of family support, all relations were not significant. At low levels of friend support, ethnic discrimination at 3-years postpartum positively predicted engagement in risky behaviors at 4-years postpartum, which in turn positively predicted BMI at 5-years postpartum. At high levels of friend support, there were no significant associations.

CONCLUSIONS

Among adolescent mothers who receive low levels of family and friend support, engagement in risky behaviors may function as a mechanism through which contextual stressors are linked to adolescent mothers’ BMI. Findings have implications for prevention efforts aimed at attenuating unhealthy weight status among Mexican-origin adolescent mothers by reducing engagement in risky behaviors and bolstering family and friend support.

Obesity disproportionately affects Mexican-origin adolescents (Ogden & Carroll, 2010) and adolescent mothers postpartum (Lemay, Elfenbein, Cashman, & Felice, 2008), compared to non-minority and non-parenting adolescents in the United States. Obesity is considered a major health risk and has been linked with the onset of disease (e.g., heart disease, diabetes, hypertension), disability, and morbidity (Maggi, Busetto, Noale, Limongi, & Crepaldi, 2015). Despite health risks associated with obesity for adolescent mothers postpartum (Lemay et al., 2008), limited work has simultaneously examined processes and protective factors that are associated with adolescent mothers’ body mass index (i.e., BMI). A comprehensive understanding of factors that influence and protect against unhealthy BMI among understudied at-risk populations, such as Mexican-origin adolescent mothers, is especially warranted.

The mechanisms underlying an unhealthy BMI may be understood through bioecological theory (Bronfenbrenner & Morris, 2006), which underscores the importance of individuals’ social environmental contexts in development. In considering the development of an unhealthy BMI among Mexican-origin adolescent mothers over time, two contextual stressors (i.e., ethnic discrimination and economic hardship) may be particularly salient. Because Latino youth and Mexican-origin adolescent mothers are disproportionately at risk for experiencing ethnic discrimination and poverty (Lemay et al., 2008; Pérez, Fortuna, & Alegria, 2008), a focus on these key stressors is critical. Indeed, higher perceived discrimination was associated with abdominal obesity among an adult multiethnic sample (Hunte & Williams, 2009). Furthermore, analyses with data from the National Longitudinal Survey of Adolescent Health revealed that greater economic hardship was associated with higher incidence of obesity for adult women (Averett & Smith, 2014). The mechanisms through which these contextual stressors shape BMI, however, are not well understood.

Prior work has indicated that contextual stressors may be associated with BMI through behavioral risks. Specifically, higher perceived discrimination and economic hardship among adolescent and ethnic minority youth have been associated with greater engagement in risky behaviors (e.g., substance use, delinquency, unprotected sex; Hurd, Varner, Caldwell, & Zimmerman, 2014; Martin, Tuch, & Roman, 2003; Pascoe & Richman, 2009). Furthermore, findings from cross-sectional and longitudinal studies reveal that risky behavior engagement may heighten individuals’ risk for unhealthy BMI (Huang, Lanza, & Anglin, 2013; Pasch, Nelson, Lytle, Moe, & Perry, 2008; Pasch, Velazquez, Cance, Moe, & Lytle, 2012; Ratcliff, Jenkins, Reiter-Purtill, Noll, & Zeller, 2011). Importantly, some studies examining sequential and bidirectional associations between BMI and risky behaviors have found that risky behavior engagement was predictive of increased BMI, however BMI did not predict subsequent risky behavior engagement (Pasch et al., 2008; Pasch et al., 2012).

Scholars have proposed that negotiating stressful experiences may influence individual capacities for making safe behavioral choices by decreasing resources for self-control (Martin et al., 2003; Pascoe & Richman, 2009). Lacking self-control to avoid behavioral risks, in response to stressful experiences, may also signify poor coping skills that pose a risk for developing unhealthy traits or practices (e.g., poor eating habits, sedentary behaviors; Pasch et al., 2008), which may increase risks for unhealthy gains in BMI (Pasch et al., 2012). Moreover, researchers have noted that youth who engage in high levels of risky behavior may become isolated from contexts (e.g., pro-social peer interactions, recreational activities) that promote normative healthy lifestyles or physical health needs. Collectively, findings support notions that links between stressors and BMI may be mediated through behavioral risks, such that adolescents may engage in risky behaviors following experiences with contextual stressors, given limited resources for self-control or poor coping skills, which in turn increases the risk of unhealthy BMI due to unhealthy lifestyles.

Although much work on BMI has centered on behavioral risks involving substance use (i.e., cigarette smoking, alcohol and marijuana use; Huang et al., 2013), other works have highlighted the significance of considering a broad spectrum of risky behaviors (e.g., fighting, shoplifting, vandalism; Pasch et al., 2008; Pasch et al., 2012; Ratcliff et al., 2011). Consistent with Problem Behavior Theory (Jessor & Jessor, 1977), which suggests that problem behaviors tend to be systematically interrelated, scholarship with Latino populations suggests that indeed patterns of multiple risky behaviors may be explained by one single underlying factor (Mobley & Chun, 2013; Wheeler et al., 2017). Accordingly, it is possible that perceptions of contextual stressors may influence adolescent mothers’ behavior systems more generally, and it is this underlying maladaptation that may trigger unhealthy lifestyles that lead to higher BMI. Thus, an understanding of how mechanisms involving a broad range of risky behaviors are linked to stressors and risks for unhealthy BMI is warranted.

Alongside considering factors that place adolescent mothers at risk for increased BMI, it is critical to consider how protective factors may simultaneously protect them from maladaptive processes associated with contextual stressors and risky behaviors. Theoretical models of Latina adolescent parenthood highlight the role of support from family members and close friends as critical for Latina adolescent mothers’ adjustment (Contreras, Narang, Ikhlas, & Teichman, 2002). Indeed, one study found that support (e.g., emotional support, providing resources and problem-solving ideas) from family and peers was associated with positive health-related behaviors (e.g., healthy eating) among low-income ethnic minority adolescents (St. George & Wilson, 2012). Furthermore, previous work with ethnic minority adolescents found that supportive friendships served an important protective role, such that higher levels of support protected against the negative influence of stressful events on risky behavior engagement (Brady, Dolcini, Harper, & Pollack, 2009). Similarly, support and emotional closeness from the family have been found to serve a protective function against the negative effects of discrimination on risky behaviors among ethnic minority adolescents (Delgado, Updegraff, Roosa, & Umaña-Taylor, 2011). Thus, high levels of support from family and friends may protect against the negative impact of perceived discrimination, economic hardship, and engagement in risky behaviors on adolescent mothers’ BMI.

The current longitudinal study examined the role of ethnic discrimination and economic hardship at 3 years postpartum on adolescent mothers’ BMI assessed at 5 years postpartum via adolescent mothers’ engagement in risky behaviors at 4 years postpartum. An understanding of stressors and behaviors related to adolescent mothers’ BMI during this developmental period is critical given findings linking adolescent mothers’ unhealthy postpartum BMI with increased risk of obesity among their toddler offspring (Lemay et al., 2008). We hypothesized moderated mediation such that risky behaviors would function as a mediator of the relation between contextual stressors and BMI and the association between contextual stressors, risky behaviors, and BMI would be moderated by family and friend support, such that associations between contextual stressors and BMI would be weaker under conditions of high family and friend support. Age was included as a control in analyses, given the variability by age that has been noted in prior work with respect to experiences with discrimination (Hurd et al., 2014) and economic hardship (Mirowsky & Ross, 2001). Additionally, because prior work has noted variability by nativity status in adolescents’ engagement in risky behaviors, rates of poverty, and experiences of discrimination (Pew Hispanic Center, 2009), analyses controlled for adolescents’ nativity. Finally, given expected variability in BMI due to subsequent pregnancy, adolescents’ pregnancy status across time was included as a control in the analyses.

Method

Participants and Procedure

Data were from a longitudinal study of 204 Mexican-origin adolescent mothers (author citation). Pregnant adolescents were recruited from high schools, health centers, and community agencies in a large Southwestern metropolitan area. To participate in the study, participants had to be 15 to 18 years old, in their third-trimester of pregnancy, of Mexican descent, not legally married, and to have a mother figure who was also willing to participate. Of 305 adolescent mothers recruited for participation, 45 were ineligible for participation due to not being currently pregnant (n = 23), not being between 15 and 18 years of age (n = 12), not having a mother figure willing to participate (n = 4), not being of Mexican origin (n = 3), and reporting married status (n = 3). From those eligible to participate (n = 260), 39 declined, 14 were unreachable for scheduling or interviews, and 3 were unable to participate due to extenuating circumstances. Adolescents completed face-to-face interviews in their home or in a lab, in which all questions were read aloud by an interviewer. The current study used three waves of data gathered at 3 years postpartum (i.e., Wave 4; W4); 4 years postpartum (i.e., W5); and 5 years postpartum (i.e., W6). Of the 204 adolescent mothers who participated at W1, 195 participated at W2, 173 participated at W3, 171 participated at W4, 173 participated at W5, and 172 participated at W6, reflecting a retention rate of 87% across the six-year period. Attrition analyses revealed no significant differences on demographic factors or key study variables between participants with complete versus missing data (results available upon request). Participants were interviewed in their preferred language (i.e., English, Spanish), and received $40 at W4, $50 at W5, and $60 at W6. Across W4 to W6, 66% to 69% of participants completed interviews in English. All procedures were approved by the Human Subjects Review Board of [blinded]. Participant characteristics are displayed in Table 1.

Table 1.

Demographic Characteristics of Study Participants

	%	M	SD	Range
Nativity Demographics (N = 204)
W1 U.S. born	65%	-	-	-
W1 Mexico born	35%	-	-	-
W1 Length of time in U.S. a	-	7.8	4.7	0–18 years
W4 Age	-	19.94	.99	18.08–22.17
W4 Education Demographics (n = 167) b
Enrolled in School	21%	-	-	-
GED/High School graduate c	41%	-	-	-
Dropout	38%	-	-	-
W4 Relationship Status (n = 170) b
Single	28%	-	-	-
Married	11%	-	-	-
Co-habiting/not married	38%	-	-	-
Dating/not co-habiting	23%	-	-	-
W4 Employment Status (n = 170) b
Not working for pay	53%	-	-	-
< 40 Hours per week	30%	-	-	-
40 Hours or more per week	17%	-	-	-
W4 Household Information b
Household income (n = 147) d	-	$24,714	$19, 545	$1,053– $105,000
Number of people in home (n = 170)	-	5	2.4	1–13

Note. W1 = Wave 1, W4 = Wave 4, W5 = Wave 5.

^aLength of time in U.S. for foreign born participants;

^bBased on valid cases at each wave;

^cEarned GED or High School graduate but no longer enrolled in school;

^dMean household income was reported by young mothers’ mother figures.

Measures

Adolescent mothers’ nativity and pregnancy status

Adolescent mothers’ country of birth was self-reported at W1. It was coded as 0 = Mexico-born, 1 = U.S.-born. In addition, adolescent mothers reported whether or not they were pregnant at each wave (W1 to W6). We created a single composite variable indicating whether adolescent mothers reported being pregnant in any of the years following their first pregnancy, or across multiple waves of data (i.e., W2, W3, W4, W5, W6). Pregnancy status was coded as 0 = Not pregnant after W1 or 1 = Pregnant at least once between W2 to W6.

Economic hardship

The Economic Hardship Measure (Barrera, Caples, & Tein, 2001) was used to assess mothers’ perceptions of economic hardship at W4. The 17-item measure (e.g., “In the next three months, how often do you expect that you/your family will have to do without the basic things that your family needs?”) includes four subscales that assess Financial Strain, Inability to Make Ends Meet, Not Enough Money for Necessities, and Economic Adjustments or Cutbacks. Following procedures outlined by Barrera and colleagues (2001), subscales were weighted and summed to create a composite of economic hardship; higher scores indicated greater economic hardship. The unweighted 17 items had an alpha coefficient of .85.

Ethnic discrimination

A revised version of the Perceived Discrimination Scale, originally adapted from the Schedule of Racist Events (Landrine & Klonoff, 1996; Whitbeck, Hoyt, McMorris, Chen, & Stubben, 2001) was used to assess Latina adolescent mothers’ perceived ethnic discrimination at W4 (Umaña-Taylor & Updegraff, 2007). Responses for the 10-item measure (e.g., “How often have others said something bad or insulting to you because you are Hispanic/Latina?”) ranged from Almost never (1) to Very often (4). Higher scores indicated higher perceived ethnic discrimination. Alpha coefficient was .84.

Family and friend support

The family and friend support subscales from the Multidimensional Scale of Perceived Social Support (Canty-Mitchell & Zimet, 2000) were used to assess family and friend support at W4 and W5. Responses for the 4-item family subscale (e.g., “My family really tries to help me”) and 4-item friend subscale (e.g., “I can talk about my problems with my friends”) ranged from (1) Strongly disagree to (7) Strongly agree. Alpha coefficients were .92 at W4 and W5 for the family support subscale, and .93 at W4 and .94 at W5 for the friend support subscale.

Engagement in risky behaviors

An adapted version of Eccles and Barber’s (1990) measure of risky behaviors was used to assess adolescent mothers’ engagement in risky behaviors (including substance use, violence, criminal behavior, and risky sexual practices) at W4 and W5. Additional items were from the Penn State Family Relationships Study to measure risky behavior from early through late adolescence (Crouter, Head, McHale, & Tucker, 2004). Adolescents reported on how frequently they engaged in each risky behavior in the past year. Responses for the 24-item measure (e.g., “done something you knew was dangerous just for the thrill of it?”, “gotten drunk or high”, “vomited, taken laxatives, or taken diet pills to lose weight?”, “started a fight with someone”) ranged from (1) Never to (5) More than 10 times. Alpha coefficients were .90 at W4 and .89 at W5.

Body Mass Index

BMI at W6 was calculated for each participant using a method from the Centers for Disease Control and Prevention (National Institutes of Health, 1998), which involved dividing adolescent mothers’ reports of weight by their reports of height squared and multiplying by a conversion factor of 703. The resulting categories for BMI include underweight (below 18.5), normal (range of 18.5–24.9), overweight (range of 25.0–29.9), and obese (range of 30.0 and above). For descriptive purposes, in the current study, 2.3% of adolescent mothers were underweight, 41.9% were normal, 30.2% were overweight, and 25.6% were obese. Due to small counts in the underweight BMI group and consistent with scholars’ recommendations cautioning against the use of categories of BMI, given restrictions in power (Hamilton & Filardo, 2006; Taylor & Yu, 2002), BMI was treated as a continuous variable.

Analytic Approach

Hypotheses were tested via two path models in Mplus version 7.11 (Muthén & Muthén, 2013), with a separate model conducted for each moderator (i.e., friend support or family support). Model fit was considered to be good if the comparative fit index (CFI) was greater than or equal to .95, the root-mean-square-error of approximation (RMSEA) was less than or equal to .05, and the root-mean-square residual (SRMR) was less than or equal to .05 (Hu & Bentler, 1999). Missing data were handled using full information maximum likelihood (Arbuckle, 1996). As recommended by MacKinnon (2008), moderated mediation was tested by modeling the moderator and mediator variables simultaneously, followed by tests of significance for any mediation pathways. Regarding interactions, continuous measures were mean-centered prior to the creation of interaction terms (Aiken & West, 1991). Simple slopes analyses were used to decompose significant interaction terms (Preacher, Curran, & Bauer, 2006), and significant interactions were probed at one standard deviation above (i.e., high) and below (i.e., low) the mean of each moderator. To test mediation for significant pathways, we used the bias-corrected bootstrap method recommended by Taylor, MacKinnon, and Tein (2008), in which confidence intervals are created for the significant mediational pathways. In this approach, mediation is significant if the confidence intervals do not contain zero (Taylor, MacKinnon, & Tein, 2008).

Results

Prior to testing the hypothesized models, correlations, means, and standard deviations were computed for all study variables (Table 2). The first model examined family support as a moderator of the association between contextual stressors and BMI via risky behaviors, which demonstrated good fit: χ² (df = 16) = 13.80, p > .05; CFI = 1.00; RMSEA = .00 (90% C.I.: .00 – .06); SRMR = .03 (see Figure 1). Results indicated that W4 family support was negatively associated with W5 risky behaviors. Additionally, the interaction between W4 economic hardship and W4 family support predicting W5 risky behaviors was significant. Furthermore, the interaction between W5 risky behaviors and W5 family support predicting W6 BMI was significant. Simple slopes analyses indicated that at low (i.e., −1 SD) levels of family support, there was a significant positive association between W4 economic hardship and W5 risky behaviors (b = .21, p = .03, 95% CI: .03–.39; see Figure 3a) and between W5 risky behaviors and W6 BMI (b = .24, p = .01, 95% CI: .05–.43; see Figure 3b); however, at high (i.e., +1 SD) levels of family support, the association was not significant between W4 economic hardship and W5 risky behaviors (b = −.11, p = .22, 95% CI: −.28-.06) nor between W5 risky behaviors and W6 BMI (b = −.04, p = .72, 95% CI: −.28-.19). Given that a mediation pathway existed (i.e., W4 economic hardship to W5 risky behaviors to W6 BMI) at low (i.e., −1 SD) levels of family support (i.e., a moderated mediation pathway), we tested the significance of this moderated mediation pathway (see Figure 4a). Results indicated that mediation was significant (unstandardized indirect effect = .14, 95% CI = .01, .33).

Table 2.

Means, Standard Deviations, and Correlations among Study Variables for Sample (N = 204)

	1	2	3	4	5	6	7	8	9	10	11	12
1. W1 age	--
2. W1 nativitya	.06	--
3. W4 economic hardship	−.05	−.18**	--
4. W4 ethnic discrimination	−.01	.13	.24***	--
5. W4 family support	.07	−.06	−.19**	−.19**	--
6. W4 friend support	.03	−.03	−.14*	−.16*	.37***	--
7. W4 risky behaviors	−.04	.16*	.18**	.27***	−.19**	−.11	--
8. W5 family support	.05	−.09	−.19**	−.22**	.64***	.25***	−.19**	--
9. W5 friend support	.18**	.04	−.20**	−.14*	.28***	.63***	−.08	.40***	--
10. W5 risky behaviors	.01	.13	.20**	.28***	−.30***	−.06	.60***	−.36***	−.05	--
11. W6 body mass index	−.10	.06	.10	.13	−.03	.00	.11	−.04	−.02	.13	--
12. W2 to W6 pregnancy statusb	−.08	−.04	.08	.09	−.09	−.15*	−.10	−.04	−.17*	−.06	−.01	--
Mean	16.81	.64	.01	1.32	5.90	5.50	1.24	5.92	5.53	1.24	26.71	.55
Standard Deviation	.99	.48	2.37	.42	1.17	1.34	.30	1.21	1.27	.28	5.26	.50

Note.

^aNativity coded as: 0 = Mexico-born, 1 = U.S.-born.

^bPregnancy status was coded as 0 = Not pregnant after W1 or 1 = Pregnant at least once between W2 to W6. W = Wave. Means and standard deviations reported here were calculated prior to centering. W4 economic hardship is calculated as a weighted summed score.

^*p < .05

^**p < .01.

^***p < .001.x

Figure 1.

Examination of ethnic discrimination and economic hardship as predictors of adolescent mothers’ body mass index via engagement in risky behaviors with family support as a moderator.

Note. W = Wave. Standardized coefficients are displayed. Variables in grey boxes represent control variables. Grey lines indicate estimates involving controls. Black lines represent hypothesized paths. Dashed lines indicate non-significant paths. Nativity coded as: 0 = Mexico-born, 1 = U.S.-born. Pregnancy status coded as: 0 = Not pregnant after W1, 1 = Pregnant at W2, W3, W4, W5 or W6. *p ≤ .05. **p ≤ .01. ***p ≤ .001.

Figure 3.

Moderation effects of (a) family support on the association between economic hardship and risky behaviors, (b) friend support on the association between ethnic discrimination and risky behaviors, (c) family support on the association between risky behaviors and body mass index, and (d) friend support on the association between risky behaviors and body mass index.

Note. Low = −1 SD above the mean and High = +1 SD above the mean. *Denotes significant slope at p < .05; **denotes significant slope at p < .01.

Figure 4.

Moderated mediation process of ethnic discrimination and economic hardship predicting adolescent mothers’ body mass index via engagement in risky behaviors at (a) low (−1 SD) family support, and (b) low (−1 SD) friend support. Standardized coefficients are displayed.

Note. W = Wave. *p ≤ .05. **p ≤ .01.

The second model examined friend support as a moderator of the association between contextual stressors and BMI via risky behaviors, which demonstrated good fit: χ² (df = 16) = 16.11, p > .05; CFI = 1.00; RMSEA = .01 (90% C.I.: .00 – .07); SRMR = .03 (see Figure 2). Results indicated that W4 discrimination was positively associated with W5 risky behaviors. In addition, the interaction between W4 discrimination and W4 friend support predicting W5 risky behaviors was significant. Furthermore, the interaction between W5 risky behaviors and W5 friend support predicting W6 BMI was significant. Simple slopes analyses indicated that at low (i.e., −1 SD) levels of friend support, there was a significant positive association between W4 ethnic discrimination and W5 risky behaviors (b = .30, p = .01, 95% CI: .08–.51; see Figure 3c) and between W5 risky behaviors and W6 BMI (b = .33, p = .01, 95% CI: .12–.53; see Figure 3d); however, at high (i.e., +1 SD) levels of friend support, the association was not significant between W4 ethnic discrimination and W5 risky behaviors (b = −.02, p = .78, 95% CI: −.19-.14) nor between W5 risky behaviors and W6 BMI (b = −.07, p = .49, 95% CI: −.28-.13). Given that a mediation pathway existed (i.e., W4 ethnic discrimination to W5 risky behaviors to W6 BMI) at low (i.e., −1 SD) levels of friend support (see Figure 4b), we tested the significance of this moderated mediation pathway. Results indicated that mediation was significant (unstandardized indirect effect = 1.15, 95% CI = .15, 3.27).

Figure 2.

Examination of ethnic discrimination and economic hardship as predictors of adolescent mothers’ body mass index via engagement in risky behaviors with friend support as a moderator.

Note. W = Wave. Standardized coefficients are displayed. Variables in grey boxes represent control variables. Grey lines indicate estimates involving controls. Black lines represent hypothesized paths. Dashed lines indicate non-significant paths. Nativity coded as: 0 = Mexico-born, 1 = U.S.-born. Pregnancy status coded as: 0 = Not pregnant after W1, 1 = Pregnant at W2, W3, W4, W5 or W6. *p ≤ .05. **p ≤ .01. ***p ≤ .001.

Discussion

This study contributes to the limited longitudinal work examining processes and protective factors simultaneously associated with adolescent mothers’ BMI in the postpartum period. Descriptively, findings revealed that 55.8% of Mexican-origin adolescent mothers in this sample had unhealthy (i.e., overweight or obese) BMI, a rate that was 30 percentage points greater than prevalence rates for 18 to 24-year old females reported from national data (i.e., 25.8%; Flegal et al., 2012). To address processes that account for unhealthy BMI, we explored potential mediating and moderating factors involved in the associations between contextual stressors, risky behaviors, and BMI. Although the mediation path from contextual factors to BMI was not supported, findings supported moderated mediation, such that mediation paths existed at low levels of the moderator (i.e., family and friend support).

As expected, findings indicated that the relation between contextual stressors at 3 years postpartum and BMI at 5 years postpartum via engagement in risky behaviors at 4 years postpartum varied as a function of support. Importantly, the link between risky behaviors and BMI emerged with the inclusion of interaction terms in study models. Although risky behaviors, friend support, and family support, did not emerge as direct predictors of BMI, significant associations emerged in the context of low levels of family and friend support. Findings indicate that the associations between risky behavior and BMI would be masked without taking into account family and friend support. In particular, at low levels of family support, higher economic hardship was associated with adolescent mothers’ unhealthy BMI over time via higher engagement in risky behaviors. In addition, at low levels of friend support, higher ethnic discrimination was related to unhealthy BMI over time via higher engagement in risky behaviors. Thus, in the context of low levels of support from family and friends, there appears to be a maladaptive pathway in which Mexican-origin adolescent mothers who experience greater levels of stress from ethnic discrimination and economic hardship engage in more risky behaviors (e.g., delinquency, substance abuse), which heighten risks for unhealthy BMI (Ratcliff et al., 2011). It is possible that as youth experience stress, they engage in risky behavior as a coping mechanism, due to limited resources for self-control (Martin et al., 2003; Pascoe & Richman, 2009). Given that we were unable to test this notion directly, future work is warranted that tests whether limited self-control resources or poor coping skills also function as mediators of the association between stress and BMI.

Further, another possibility for this link may be that youth who engage in risky behaviors may be more inclined to develop unhealthy practices (e.g., poor eating habits, sedentary physical habits), which ultimately increase BMI (Pasch et al., 2008; Pasch et al., 2012). Because unhealthy practices were not directly assessed in the current study, future work should consider how specific health-related practices may contribute to gains in BMI, secondary to risky behavior engagement. Given that ethnic discrimination and poverty are pervasive among Latino adolescents (Pérez, et al., 2008) and Mexican-origin adolescent mothers in the U.S. (Lemay et al., 2008), findings highlight the importance of developing alternative coping strategies for Mexican-origin adolescent mothers with low levels of support to reduce risky behavior engagement and subsequent unhealthy BMI. Findings underscore the need to directly target engagement in risky behaviors among at-risk adolescent mothers as one possible avenue to protect against unhealthy postpartum BMI.

Interestingly, contextual stressors were differentially associated with risky behavior in each moderator model (i.e., friend and family support). In the context of low friend support, high levels of ethnic discrimination were associated with greater engagement in risky behavior. In the context of low family support, economic hardship emerged as a significant predictor of risky behavior. Regarding the direct association between ethnic discrimination and risky behavior, it may be that the link becomes more robust for Mexican-origin adolescent mothers in the context of low friend support. Findings coincide with prior work with ethnically diverse college students in which greater friend support, but not family support (i.e., cohesion), buffered the negative effects of racial-ethnic discrimination on somatization (Juang, Ittel, Hoferichter, & Gallarin, 2016). Scholars have suggested that youth may perceive peers as better providers of support given that they are more likely to share in comparable experiences of discrimination. In the context of the current findings, it is possible that the support provided by family members, who are more likely to communicate or share in economic hardships with adolescents, as well as provide tangible support (e.g., financial resources, material goods), may be more effective than the support provided by friends in buffering against the effects of economic hardship. Indeed, research on the implications of Mexican-origin adolescents’ perceptions of family economic hardship for adolescent adjustment (i.e., depression, risky behavior) provides support for the important role of family relationships (Delgado, Killoren, & Updegraff, 2013). Thus, in the context of low support from either friends or family members, the risks of specific contextual stressors may become more salient given supporters’ proximity to shared experiences. These findings highlight the importance of identifying and understanding the role of contextual stressors in the context of different sources of support. Importantly, because family support and friend support were not tested in the same model, given sample size limitations, it will be important for future work with larger samples to examine these associations together.

It is worthy to note that the maladaptive mediational pathways from contextual stressors to BMI via risky behaviors were not found in the context of high levels of family and friend support. When adolescent mothers experience high levels of family and friend support, these supportive networks may mitigate the effects of ethnic discrimination and economic hardship on engagement in risky behaviors by providing effective resources (e.g., encouragement, tangible support) to adolescent mothers. The presence of supportive networks in the early years of parenting may be particularly meaningful for Mexican-origin adolescent mothers, given work highlighting the significance of familism values (i.e., family commitment and obligation; Knight et al., 2010) and dependence on family and friends for support among Latina adolescent mothers, in general (Contreras et al., 2002). Thus, it is possible that some Mexican-origin adolescent mothers may be more resilient to the risks associated with salient contextual stressors in the context of supportive family and friend environments.

Although the current study advances the understanding of processes underlying adolescent mothers’ unhealthy BMI, findings must be considered alongside important limitations. First, the current study assessed BMI with adolescent mothers’ self-reports of weight and height. To eliminate reporter bias, it will be important to directly assess participants’ current weight, along with their lowest and highest weight in the past few months. Future work should also replicate findings with other measures, such as trained staff assessments of waist circumference or body fat percentage with professional grade equipment. Second, adolescent mothers in the current study were all of Mexican origin; because ethnic differences have been identified in individuals’ BMI (Flegal et al., 2012), it will be important for future studies to test whether the strength of the associations underlying BMI generalize to adolescent mothers from different ethnic backgrounds (e.g., African American) as well as other Latino subgroups.

Next, BMI was only measured at 5 years postpartum (i.e., W6) and not at prior waves (i.e., W1–W5); thus, the current findings are limited to understanding predictors of BMI at one point in time. Future studies should assess BMI at multiple time points to examine the role of contextual stressors and risky behavior on changes in BMI over time. By understanding the rate of change in BMI, it will be possible to identify specific periods of vulnerability or detect trajectory patterns for unhealthy weight gain over time, in relation to contextual stressors. Additionally, given established links between BMI and unhealthy behavioral patterns (e.g., eating disorders; Haines & Neumark-Sztainer, 2006) and caloric intake (Flegal et al., 2012), it will be important for future studies examining the mediating role of risky behaviors to control for these factors. Lastly, future work should examine whether study associations vary by nativity status. Indeed, prior work has demonstrated that the effects of contextual stressors may differentially occur among Latino youth based on their country of origin (Pew Hispanic Center, 2009), as foreign-born Latino youth tend to report lower risky behavior engagement compared to their U.S-born counterparts (Killoren & Deutsch, 2014). Examining multi-group differences was not possible in the current study, given the relatively small sample size.

Despite these noted limitations, findings from the current longitudinal study provide important insights regarding predictors of Mexican-origin adolescent mothers’ BMI and underscore the critical role of support from family and friends as salient protective factors for these young mothers’ well-being and health status. Study findings contribute to the limited work simultaneously examining protective and risk factors among a marginalized at-risk population and highlight two potential areas for intervention. First, Mexican-origin adolescent mothers who experience low levels of support from family and friends may benefit the most from learning about positive ways to cope with salient contextual stressors, particularly related to economic hardship and ethnic-racial discrimination. Exposing adolescent mothers to positive coping strategies may reduce risky behavior engagement as a way of coping with stressors. Second, given that maladaptive pathways from contextual stressors to risky behaviors to unhealthy BMI did not exist in the context of high levels of family and friend support, intervention efforts with Mexican-origin adolescent mothers should (a) include supportive family members and friends, and (b) include strategies for improving and maintaining supportive relationships with family members and friends over time.