Does context and adversity shape sexual behavior in youth? Findings from two representative samples of Puerto Rican youth

Abstract

CONTEXT:

Sexual risk behaviors often co-occur. Understanding the heterogeneity in patterns of sexual behavior among youth, and how context of majority and minoritized status may be related to these behaviors, can inform targeted STIs/HIV interventions.

METHODS:

Data are from the Boricua Youth Study, a longitudinal study of two probability samples of Puerto Rican youth recruited in the South Bronx (SBx) and the metropolitan area in Puerto Rico (PR). We identified patterns of sexual behaviors among young adults (ages 15–24) with sexual experience (N=1,203) using latent class analysis. Analyses examined context differences and the prospective relationship between adverse childhood experiences (ACEs) (childhood maltreatment/violence, family/parental dysfunction) and patterns of sexual behaviors (age at first sex, number of sex partners, sex with a high-risk partner, condom use, sex while intoxicated, oral sex, anal sex).

RESULTS:

We identified five classes of sexual behaviors: (1) currently inactive (16.51%); (2) single partner, low activity (13.49%); (3) single partner, inconsistent condom use (32.19%); (4) single partner, sex without a condom (27.65%); and (5) multi-risk (10.16%). Young adults from the SBx (minoritized context), those who identified as male, and those with higher child maltreatment/violence ACEs were more likely to be in the multi-risk class relative to the single partner, inconsistent condom use class. Those from the SBx were also more likely to be in the single partner, sex without condom class, relative to the single partner, inconsistent condom use class. Differences in young adults’ patterns of sexual behaviors between the two contexts, one representing the minoritized context (SBx) contrasted to the majority context (PR), were not explained by ACEs.

CONCLUSION:

Findings highlight the heterogeneity in the patterns of sexual behaviors among Puerto Rican young adults as well as how such patterns vary based on sociocultural contexts. Exposure to child maltreatment/violence ACEs was related to the riskier patterns, however they did not explain why riskier patterns of sexual behaviors were found in the SBx compared to PR. Results underscore the need for tailored interventions and more in-depth examination of differences across contexts.

Adolescents and young adults ages 15–24 account for half of all new sexually transmitted infections (STIs) in the U.S. (Centers for Disease Control and Prevention, 2019b). Systemic racism and social inequities often result in unequal distribution of disease in the population (Alson et al., 2021; Bailey et al., 2021). Hispanic young adults ages 20–24 have 1.3 times the reported rate of chlamydia, 1.4 times the rate of gonorrhea, and 2.1 times the rate of syphilis compared to their White counterparts (Centers for Disease Control and Prevention, 2021). In addition, the rate of HIV infection among Hispanic young adults is elevated relative to non-Hispanic White young adults (Centers for Disease Control and Prevention, 2019a). Relative to their White peers, Hispanic youth are more likely to engage in some sexual risk behaviors (e.g., early sexual debut), but not others (e.g., no condom use at last sex, multiple sex partners) (Centers for Disease Control and Prevention, n.d.), making it difficult to obtain an understanding of overall risk. Examining patterns of sexual behavior, as well as contextual factors that contribute to specific patterns of sexual behavior is essential, as certain combinations of behaviors may place individuals at greater risk than others (Pflieger et al., 2013; Vasilenko et al., 2015).

Sexual risk behaviors differ by context (e.g., World Health Organization, 2021). These differences are likely due to a confluence of factors, including access, knowledge, and beliefs around sexual activity and contraception (Wellings et al., 2006). Differences may also be related to a racial and ethnic group’s social standing in their context (i.e., whether they are part of the dominant or minoritized group). Minoritized groups often face the effects of structural racism, interpersonal and institutional discrimination, and oppression (Bailey et al., 2017; Garcia Coll et al., 1996), which may play a role in how they respond to their context. In particular, adults who identify as Hispanic, Black, or multiracial report more adverse childhood experiences (ACEs) than those who identify as White, as do Puerto Rican youth living in the South Bronx, New York (a context where they are part of the minoritized group) compared to those in their hometown of Puerto Rico (Giano et al., 2020; Ramos-Olazagasti et al., 2017). ACEs have been linked to multiple adverse health outcomes and risk behaviors (Anda et al., 2006; Dube et al., 2001; Felitti et al., 1998; Kalmakis & Chandler, 2015). In this study, we capitalize on a unique dataset that allows us to examine patterns of sexual behaviors among Puerto Rican youth who grew up in two contexts, those from the South Bronx (SBx), where they are minoritized compared to those from the metropolitan area in Puerto Rico (PR), where they reside as the majority group. We examine how exposure to ACEs relates to patterns of sexual behaviors, and whether they account for differences in patterns of sexual behaviors across the two contexts.

BACKGROUND

Sexual activity is multidimensional, consisting of different behaviors (e.g., vaginal sex, anal sex) that occur under different circumstances (e.g., with or without condoms, with one versus multiple partners). Sexual risk behaviors often cluster (Vasilenko et al., 2015); for example, early initiation of sexual activity is usually related with having multiple sex partners, having sex with high-risk partners, and having sex while intoxicated (Epstein et al., 2014; O’Donnell et al., 2001). Most research focuses on single indicators of sexual behavior and only a few studies have examined differences in patterns of behaviors.

Person-centered statistical approaches allow us to identify groups of individuals characterized by similar behavior patterns. For example, Vasilenko et al. (2015) identified five patterns (or classes) of adolescent sexual behavior ranging from no sexual activity to multiple risk behaviors. Rates of STIs varied by class membership. Pflieger et al. (2013) also found that certain behavior patterns (having non-monogamous multiple partners) were associated with greater risk for STIs than others (single, monogamous partner). This was the case even though individuals in relationships with multiple, non-monogamous partners used condoms more frequently.

Understanding these patterns can help design more effective interventions by identifying which aspects of the intervention to emphasize for whom. For example, if females are more likely to have sex without a condom with a single, main partner and exhibit no other risk behaviors, female-focused interventions that increase efficacy in safe sex negotiations would be indicated. If males engage in multiple risk behaviors simultaneously, they would benefit from comprehensive interventions targeting multiple behaviors.

While research on patterns of sexual behaviors is still emerging, evidence suggests patterns of behaviors vary by racial and ethnic group (Pflieger et al., 2013; Tubman et al., 2003; Vasilenko et al., 2015). Common value systems in the Hispanic culture that place a high premium in religiosity may favor unique patterns of behavior characterized by delays in the initiation of sex and monogamous relationships, but reduced condom use due to perceptions of sex as a forbidden behavior outside of marriage (Lescano et al., 2009).

Notably, the Hispanic population is diverse, with differences in sexual behaviors and reproductive health outcomes across country of origin, nativity, and acculturation level (Guilamo-Ramos et al., 2009; Guilamo-Ramos et al., 2005; Moreno & El-Bassel, 2007; Weiss & Tillman, 2009). For example, the Puerto Rican youth population in the continental U.S. is relatively more acculturated than youth from Mexican and Cuban origin (Guilamo-Ramos et al., 2005), and acculturation is related to sexual risk behaviors, albeit in complex ways (Guilamo-Ramos et al., 2005; Lee & Hahm, 2010). Puerto Rican adolescents in the U.S. engage in more sexual activity than youth of Mexican and Cuban origin (Guilamo-Ramos et al., 2005). A study using a convenience sample in New York found that Puerto Rican women were more heterosexually active, had more sex partners, and a greater incidence of STIs and HIVs than Dominican women, even when they were more likely to use condoms (Moreno & El-Bassel, 2007). In contrast, adolescents in Puerto Rico are less likely to be sexually active and to have multiple sex partners than Hispanic adolescents in the continental United States (Martin et al., 2017). To our knowledge, no population-based studies have documented patterns of sexual behaviors of Puerto Rican young adults specifically and compared these patterns across two contexts.

Among Puerto Rican young adults, sexual behaviors should be considered within the context where they live, and the risk factors present in that context. In Puerto Rico and in certain areas in the U.S. where Hispanics are concentrated, as is the case for Puerto Ricans in the SBx, poverty is widespread: 58% of children in PR (Kids Count Data Center, 2021) and 39% of those in the Bronx (U.S. Census Bureau, 2019) live in poverty. Poverty can affect sexual behaviors directly, for example, by limiting individuals’ ability to purchase condoms, but also, with poverty, often comes exposure to adversity (Blair & Raver, 2012). A previous study using the same dataset as the one used here showed that exposure to ACEs—including parental loss, child maltreatment, parental maladjustment—was elevated in the two communities relative to the general population (Sacks and Murphey, 2018), but particularly in the SBx (Ramos-Olazagasti et al., 2013). Several studies comparing the outcomes of youth in the two contexts have shown relatively less favorable outcomes in the SBx than in PR (e.g., Alegria et al., 2019; Bird et al., 2007; Canino et al., 2021; Duarte, Canino, Wall, et al., 2021; Ramos-Olazagasti et al., 2013).

Whether sexual behaviors of the same ethnic group varies by context remains unknown. Sexual behaviors vary around the world due to a combination of demographic, political, social, and cultural features that shape behavioral opportunities, rewards, and sanctions (Wellings et al., 2006). A greater exposure to adversity in one context may contribute to differences in patterns of sexual behaviors of the same ethnic subgroup in different contexts. Exposure to ACEs can impair a child’s social, emotional, and cognitive abilities and elevate risk-taking (Shonkoff et al., 2012). Recent studies have shown that exposure to ACEs and poverty-related stressors is related to individuals’ stress response, emotion regulation, executive functioning, and decision-making skills (Anda et al., 2006; Bruce et al., 2013; Carpenter et al., 2007; DePrince et al., 2009; Hart & Rubia, 2012; Lupien et al., 2009; Pechtel & Pizzagalli, 2011). These skills have been associated with increased impulsivity and risk taking (Khurana et al., 2013; Romer et al., 2009; Romer et al., 2011).

Specifically, ACEs are related to a number of sexual risk behaviors (SRB) such as early sex (Hillis et al., 2001; Nebbitt et al., 2017; Van Rosmalen-Nooijens et al., 2017; Wilson & Widom, 2008), reduced condom use (Senn & Carey, 2010; Van Rosmalen-Nooijens et al., 2017), sex with multiple partners (Heard-Garris et al., 2018; Hillis et al., 2001; Orgilés et al., 2015; Senn & Carey, 2010; Van Rosmalen-Nooijens et al., 2017), substance use before sex (Tubman et al., 2011; Voisin, 2005), and having sex for money (Wilson & Widom, 2008). Additionally, as the number of adversities increases, so does the likelihood of engaging in SRB (Alexander et al., 2018; Hillis et al., 2001). Most research, however, has focused on maltreatment, particularly sexual abuse. Other ACEs, like parental loss, parental incarceration, and parental mental health problems, which would likely require different intervention strategies, have received less attention. These ACEs can severely disrupt family dynamics and compromise the availability of parental support, monitoring, and supervision (Chilcoat et al., 1996; Goodman, 2007; Kjellstrand & Eddy, 2011).

To our knowledge, no study has focused on how ACEs are related to patterns of sexual behavior, beyond individual indicators. Past research on ACEs and sexual activity is also limited by a reliance on retrospective reports of ACEs and sexual behaviors over long periods of time, which can introduce bias and difficulties establishing directionality. Other limitations include the use of convenience or clinical samples (Abdala et al., 2016; Houck et al., 2009; Kogan et al., 2016; Tubman et al., 2011), samples of women only (Elliott et al., 2002; Hahm et al., 2010; Hillis et al., 2001; Senn & Carey, 2010), and samples of middle- to upper-class adults (Hillis et al., 2001). In addition, past studies have rarely examined the association between ACEs and patterns of sexual behavior among young adults for whom exposure to multiple stressors is common.

This study addresses some of the existing methodological limitations of other studies and uses a large, prospective, household probability sample of Puerto Rican youth from two contexts, one majority and the other a minoritized context, to identify distinct patterns of sexual behaviors, and prospectively examine the relationship between ACEs and these behavior patterns. Specifically, we sought to answer these questions:

1. What patterns of sexual behaviors are observed in Puerto Rican young adults, and do patterns vary by context? We hypothesized that young adults from the South Bronx would be more likely to exhibit sexual behavior patterns typically associated with risk for STIs/HIV. Because there are well-established differences in sexual behaviors between males and females but not enough information about how patterns of behaviors cluster by gender, we also report on gender differences.

2. How do ACEs, prospectively examined, relate to patterns of sexual behaviors, and do they account for any observed difference in patterns of sexual behaviors by context? We hypothesized that ACEs would be related to patterns of sexual behaviors typically associated with risk for STIs/HIV, and that if found, contextual differences in patterns of sexual behaviors would be accounted for by ACEs.

Methods

Sample Characteristics

The Boricua Youth Study (BYS) is a longitudinal study of Puerto Rican youth living in the South Bronx, NY (SBx, N=1,138) and in the Standard Metropolitan Area of San Juan and Caguas, Puerto Rico (PR, N=1,353) (Bird et al., 2006; Bird et al., 2007). Samples are multi-stage probability samples representative of households with at least one child between the ages of 5 and 13 who had a parent who identified as Puerto Rican. Up to three randomly selected children participated in the study per household. Data collection began in year 2000 (Wave 1, M_age=9 years). Children and their parents were interviewed annually for three consecutive years (Waves 1–3). A fourth wave was completed 11 years later, when youth were on average 22 years old (range 15–29) with high retention rates (83% in SBx, 82% in PR, N=2,004) (Alegria et al., 2022; Duarte, Canino, Alegria, et al., 2021).

Study procedures and sample characteristics are described elsewhere (Bird et al., 2006; Bird et al., 2007; Duarte, Canino, Alegria, et al., 2021). Briefly, youth and a primary caregiver (usually the mother) were interviewed in their homes or a convenient location by trained bilingual interviewers. The interview was computerized and administered by study staff, except the section that inquired about sexual behaviors at Wave 4, which participants completed independently to yield more reliable reporting and reduce social desirability bias. All procedures were approved by the IRB at the University of Puerto Rico Medical Sciences Campus and the New York State Psychiatric Institute.

This study focuses on sexual behaviors assessed at Wave 4. Analyses were restricted to young adults who were 15–24 years old at Wave 4 (N=1,480), the highest risk period for STIs (Satterwhite et al., 2013), and those who were sexually experienced (vaginal, anal, or oral) at Wave 4, N=1,203 (weighted percent=82.93%; mean age =21.38, SD=2.05). More young adults in the minoritized context, SBx, were sexually experienced (87.85%) relative to Puerto Rico (78.73%), but there was no difference by gender (82.29% for males, 83.62% for females).

Measures

Sexual behaviors (Wave 4).

Sexual behaviors were measured using an adaptation of the Adolescent-Risk Behavior Assessment (ARBA) (Donenberg et al., 2001), a computerized structured interview that assesses self-reported sexual behavior and injection drug use in youth, has good test-retest reliability (Vanable et al., 2009), and has been translated into Spanish (Robles et al., 1992). Seven indicators informed the latent classes, all referring to behaviors in the past three months, except age at first sex. The indicators and the coding of the responses were: age at first vaginal or anal sex (0=never had vaginal/anal sex, 1=vaginal/anal sex before age 14, 2=vaginal/anal sex on or after age 14); number of sex partners (0, 1, 2 or more); sex with a high-risk partner, defined as a partner whose sexual history is unknown, sex with a sex worker, sex with someone with HIV/AIDS, sex with an intravenous drug user, or sex with someone the participant met online (0=not currently active, 1=no, 2=yes to any); condom use during anal or vaginal sex (0=not currently active in vaginal/anal sex, 1=always, 2=sometimes, 3=never); vaginal or anal sex while intoxicated (0=not currently active in vaginal/anal sex, 1=no, 2=yes); oral sex (0=no, 1=yes), and anal sex (0=no, 1=yes).

Context.

Context was defined as majority (PR) or minoritized context (SBx) according to the recruitment site.

Adverse childhood experiences (Waves 1–3).

We included ACEs from the initial ACEs study (Anda et al., 1999; Felitti et al., 1998) and others identified in the literature (McLaughlin et al., 2010): parental loss (parental death, divorce/separation), child maltreatment (neglect, physical abuse, sexual abuse, emotional abuse), and parental maladjustment (intimate partner violence, parental emotional problems, parental incarceration, parental substance use problems). We added exposure to violence for its relevance among racial and ethnic minorities and its robust association with negative outcomes in youth (Cronholm et al., 2015) (a detailed description of ACEs is available elsewhere, Ramos-Olazagasti et al., 2017).

To examine dimensionality among the ACEs, we performed an exploratory factor analysis within context. Each ACE was considered dichotomous (absent or present); thus, factor analyses were based on tetrachoric correlations, and robust-weighted least squares estimators were used. We compared candidate factor models (1–4 factors) in each context by examining model fit indices: RMSEA, CFI, and TLI. Overall model-data consistency was evaluated using the chi-square goodness-of-fit test statistic. In both contexts, a 2-factor model was the most parsimonious well-fitting model. A confirmatory factor analysis with two factors on the entire sample fit the data well (RMSEA=0.026, 90% CI [0.020–0.033], CFI=0.963, TLI=0.953). The first factor represents child maltreatment (neglect, physical abuse, sexual abuse, emotional abuse) and exposure to violence. The second factor represents family and parental dysfunction ACEs: intimate partner violence, parental incarceration, parental substance abuse, parental emotional problems, and parental divorce/separation. These two factors load onto a second-order factor of overall ACEs. Parental death did not load onto either of the two sub-factors but loaded modestly onto the overall ACEs factor. We used the score on this second-order factor (herein total ACE score) to measure overall adversities, and factor scores for each of the sub-factors to analyze the associations of these two types of adversities–child maltreatment/violence-related and parent-related–with patterns of sexual behaviors.

Covariates.

Gender was self-reported by young adults at Wave 4. Participants reported on their sexual orientation by indicating whether they identified as heterosexual/straight, gay, lesbian, bisexual, or other. We combined all non-heterosexual responses and created a dichotomous variable indicating heterosexual or non-heterosexual orientation. We also adjusted for whether the participant was married or living with their partner.

Analytic Approach

We conducted a latent class analysis (LCA) (Collins & Lanza, 2009) in Mplus to examine patterns of sexual behaviors. We first conducted an LCA on the indicators of SRB, specifying 1–7 classes (Table 2). We selected the best model based on goodness-of-fit measures, prior research, and interpretability. To examine the distribution of latent classes by the covariates, we extracted the predicted probabilities of class membership for each individual and assigned them into their most likely class membership (Table 4). Next, we tested whether context and ACEs were associated with class membership using multinomial logistic regression in SAS. To test whether ACEs could account for the association between context and class membership, we compared the coefficient for context in models that included the ACEs to a model that did not include the ACEs. Models adjusted for age, gender, sexual orientation, and whether the participant was married or living with a partner. Analyses applied the BYS context-specific sampling weights that adjusted for non-response at Wave 4 (Duarte, Canino, Alegria, et al., 2021) and accounted for clustering of youth in the same family.

Table 2.

Model fit statistics for classes 1 through 7

Number of Classes	Entropy	BIC	AIC	−2LogL	p *
1	-	13670.12	13603.91	13577.91	-
2	0.99	10237.86	10090.76	10032.76	<0.001
3	0.98	10150.08	9941.29	9859.29	<0.001
4	0.80	10116.93	9836.84	9726.84	0.018
5	0.79	10155.88	9804.49	9666.49	0.357
6	0.80	10220.26	9797.57	9631.57	0.957
7	0.80	10297.86	9803.88	9609.88	>0.999

^*Likelihood ratio test: testing if current number of classes (k) fits data better than the previous number of classes (k-1).

Table 4.

Latent class prevalence as a function of context, gender, sexual orientation, and marital/cohabitation status

	Class 1: Currently inactive	Class 2: Single partner, low activity	Class 3: Single partner, inconsistent condom use	Class 4: Single partner, sex without a condom	Class 5: Multi- risk
Overall class prevalence	16.51%	13.49%	32.19%	27.65%	10.16%
Context
South Bronx	14.10%	13.08%	27.73%	32.76%	12.33%
Puerto Rico	18.80%	13.89%	36.44%	22.79%	8.08%
Gender
Male	19.10%	12.43%	30.69%	23.51%	14.27%
Female	13.78%	14.61%	33.77%	32.01%	5.83%
Sexual orientation
Heterosexual	16.46%	11.16%	34.52%	28.20%	9.67%
Not heterosexual	18.18%	35.75%	10.86%	22.07%	13.14%
Married or cohabitating
No	22.90%	16.46%	27.96%	21.54%	11.14%
Yes	1.94%	6.85%	41.65%	42.38%	7.18%

Note. Percentages are weighted.

Results

Distribution of sex behaviors

Table 1 shows the distribution of sexual behaviors overall and by context and gender. Approximately 9% of young adults indicated that they first had vaginal or anal sex before age 14. Most young adults reported having sex without condoms in the past three months: 24.55% never used condoms and 30.05% used them inconsistently. Of note, 30.32% of the sexually experienced sample had not had anal or vaginal sex in the past three months. Most young adults (67.82%) had only one sex partner in the past three months, and 17.25% had two or more partners. Oral sex was common (60.04%), but not anal sex (15.02%). Close to one third of young adults (30.60%) were having sex with a partner identified as high risk, and 20.88% had sex while intoxicated.

Table 1.

Weighted distributions of sexual behaviors across contexts and separately by site and gender

Sexual Risk Behavior	Total	SBx	PR	Females	Males
	%(SE)	%(SE)	%(SE)	%(SE)	%(SE)
Age at first vaginal or anal sexa
Never had vaginal or anal sex	2.98(0.52)	2.69(0.70)	3.25(0.77)	3.05(0.73)	2.92(0.74)
<14 years olda	8.70(0.88)	11.67(1.37)	5.87(1.09)	6.82(1.10)	10.50(1.35)
≥14 years olda	88.32(1.00)	85.63(1.50)	90.87(1.31)	90.13(1.29)	86.58(1.50)
Condom use during anal or vaginal sex*ab
Not active in vaginal or anal sexa	30.32(1.46)	27.38(1.91)	33.15(2.19)	28.57(2.01)	32.00(2.11)
Always used a condomb	15.08(1.14)	17.40(1.59)	12.85(1.63)	13.07(1.47)	17.00(1.72)
Never used condomsb	24.55(1.35)	24.92(1.83)	24.20(1.98)	29.63(2.00)	19.69(1.80)
Inconsistently used condoms	30.05(1.43)	30.31(1.93)	29.80(2.09)	28.73(1.97)	31.31(2.05)
Number of partners*b
0b	14.94(1.16)	12.77(1.45)	17.02(1.81)	12.35(1.53)	17.41(1.74)
1b	67.82(1.49)	68.27(1.99)	67.38(2.20)	76.82(1.89)	59.19(2.22)
≥2b	17.25(1.18)	18.96(1.66)	15.60(1.68)	10.82(1.34)	23.40(1.89)
Oral sex*a
No	39.96(1.57)	36.29(2.06)	43.47(2.33)	37.41(2.17)	42.38(2.25)
Yes	60.04(1.57)	63.71(2.06)	56.53(2.33)	62.59(2.17)	57.62(2.25)
Any anal sex*
No	84.98(1.12)	83.68(1.56)	86.23(1.60)	87.10(1.43)	82.95(1.69)
Yes	15.02(1.12)	16.32(1.56)	13.77(1.60)	12.90(1.43)	17.05(1.69)
Sex with a high-risk partner*
Not sexually active	15.09(1.18)	13.06(1.47)	17.04(1.82)	12.62(1.56)	17.46(1.76)
No	54.31(1.59)	53.30(2.14)	55.27(2.33)	55.42(2.21)	53.25(2.27)
Yes	30.60(1.44)	33.64(2.03)	27.68(2.04)	31.96(2.06)	29.29(2.02)
Vaginal or anal sex while intoxicated*ab
Not active in vaginal or anal sexab	30.29(1.46)	27.47(1.91)	32.99(2.20)	28.70(2.02)	31.80(2.11)
Noab	48.83(1.58)	41.65(2.09)	55.72(2.31)	54.34(2.19)	43.56(2.24)
Yesab	20.88(1.25)	30.88(1.94)	11.28(1.46)	16.95(1.58)	24.64(1.90)

Note. Analyses are restricted to those <25 years old and who ever had sex.

^*In the past 3 months

^aSite difference, p<0.05

^bGender difference, p<0.05

Bivariate analyses revealed differences by context and gender. Early initiation of sex was more prevalent in the SBx (11.67%) than in PR (5.87%), as was oral sex (63.71% versus 56.53%) and having sex while intoxicated (30.88% versus 11.28%). Females were more likely than males to always have unprotected sex (29.63% versus 19.69%, respectively), but less likely to have multiple partners (10.82% versus 23.40%, respectively) or have sex while intoxicated (16.95% versus 24.64%).

Latent classes

We compared models with one through seven classes (Table 2) and determined that a five-class solution model fit the data well and allowed for the most meaningful classes. We selected it over the four-class model because it identified a group that engaged in multiple risk behaviors, which is of substantive interest. We did not select the six-class model due to low class prevalence for one class (under 5.0%) and redundancies between classes.

Table 3 shows the latent class prevalence and item-response probabilities for the five-class model. We interpreted the classes as: (1) Currently inactive (16.51%), characterized by no sexual activity in the past three months; (2) Single partner, low activity (13.49%), distinguished from the currently inactive class for their high probability of currently having one partner. Young adults in this class had not engaged in any recent vaginal or anal sex and had a relatively low probability of engaging in oral sex. The single partner, inconsistent condom use (32.19%) class (3) was characterized by high probabilities of having a single partner who is not high-risk; inconsistent use of condoms with similar probabilities of using condoms consistently, sometimes, or never; low probabilities of having sex while intoxicated or having anal sex; and high probability of having oral sex. The single partner, sex without a condom (27.65%) class (4) is differentiated from the previous class for its high probability of using condoms inconsistently, or never. While not the norm, young adults in this class also had relatively higher probabilities of having sex with a high-risk partner, having sex while intoxicated, and having anal sex. Last, the multi-risk (10.16%) class (5) was characterized by a high probability of having multiple partners, high-risk partners, using condoms inconsistently, having sex while intoxicated, having oral sex, and relative to other classes, higher probability of having anal sex.

Table 3.

Latent class prevalence and item-response probabilities for five class model of Puerto Rican youth’s sexual activity and HIV/STI risk behaviors

	Class 1: Currently inactive	Class 2: Single partner, low activity	Class 3: Single partner, inconsistent condom use	Class 4: Single partner, sex without a condom	Class 5: Multi-risk
Class Prevalence	16.51%	13.49%	32.19%	27.65%	10.16%
Item Response Probabilities
Age at 1st vaginal or anal sex
Never had vaginal or anal sex	0.10	0.10	0.00	0.00	0.00
Vaginal or anal sex before age 14	0.09	0.07	0.03	0.11	0.19
Vaginal or anal sex on or after age 14	0.81	0.84	0.97	0.89	0.81
Number of sex partners*
0	1.00	0.01	0.00	0.00	0.00
1	0.00	0.84	0.91	0.89	0.04
2+	0.00	0.15	0.09	0.11	0.96
Sex with a high-risk partner*
Not sexually active	1.00	0.00	0.00	0.00	0.00
No	0.00	0.71	0.80	0.55	0.42
Yes	0.00	0.29	0.20	0.45	0.58
Condom use during anal or vaginal sex*
Not sexually active in vaginal or anal sex	1.00	1.00	0.00	0.00	0.00
Always	0.00	0.00	0.32	0.09	0.40
Sometimes	0.00	0.00	0.35	0.45	0.59
Never	0.00	0.00	0.33	0.46	0.01
Vaginal or anal sex while intoxicated*
Not sexually active in vaginal or anal sex	1.00	0.99	0.00	0.00	0.00
No	0.00	0.01	0.92	0.63	0.34
Yes	0.00	0.00	0.08	0.37	0.66
Oral Sex*
No	1.00	0.70	0.35	0.09	0.16
Yes	0.00	0.30	0.65	0.91	0.84
Any anal sex*
No	1.00	1.00	1.00	0.64	0.71
Yes	0.00	0.00	0.00	0.36	0.29
Covariates
Mean Age(SD)	20.65(2.14)	20.83(2.11)	21.33(2.08)	21.57(1.92)	21.29(1.99)
Not heterosexual	10.48%	25.36%	3.23%	7.66%	12.59%
Male	59.34%	47.27%	48.90%	43.62%	72.05%
South Bronx	41.68%	47.29%	42.04%	57.80%	59.24%
Married/Cohabitating	3.41%	14.74%	38.22%	44.98%	21.11%

^*In the past three months

Note.Class membership is adjusted for by marital/cohabitation status and age at Wave 4.

Table 4 shows the distribution of class membership by context and other covariates. The most common behavior pattern among young adults in SBx (the minoritized context), and young adults who were married or cohabiting was the single partner, sex without a condom class (class 4). For young adults in PR (the majority context), young adults from both genders, those who identified as heterosexual, and those who were not married or cohabiting, the most common behavior pattern was having a single partner, with inconsistent condom use (class 3), whereas the most common behavior pattern for those who identified as gay, lesbian, or bisexual (not heterosexual) was one characterized by single partner, low activity (class 2). The prevalence of behavior patterns characterized as “risky” (class 5) was relatively low for females (5.83%), but not for males (14.25%).

Patterns of Sexual Behaviors by Context and Exposure to Adverse Childhood Experiences

Using the latent classes as outcomes in multinomial logistic regressions, we added predictors with covariates, using the single partner, inconsistent condom use class (class 3) as the reference (Table 5). Young adults in SBx were more likely than those in PR to be in the multi-risk class (class 5) (OR=1.81[1.14;2.88], p<.05), and in the single partner, sex without condoms class (class 4) relative to the referent class (class 3) (OR=1.89[1.36;2.63], p<.001). Males were more likely than females to be in the multi-risk class (OR=2.64[1.62;4.31], p<.001) than in the single partner, inconsistent condom use class (class 3).

Table 5.

Odds ratios and confidence intervals showing the association between context, gender, and ACEs and latent class membership

	Class 1: Currently inactive	Class 2: Single partner, low activity	Class 3: Single partner, inconsistent condom use	Class 4: Single partner, sex without a condom	Class 5: Multi-risk
Context (SBx=1)	0.77[0.51;1.16]	0.85[0.54;1.32]	REF	1.89[1.36;2.63]***	1.81[1.14;2.88]*
Gender (Male=1)	1.43[0.96;2.15]t	1.03[0.54;1.32]	REF	0.91[0.65;1.27]	2.64[1.62;4.31]***
Child maltreatment/violence ACEs score	1.05[0.84;1.31]	1.22[0.97;1.52]t	REF	1.16[0.96;1.40]	1.28[1.00;1.64]*
Family/parental dysfunction ACEs score	1.10[0.89;1.38]	1.02[0.81;1.28]	REF	1.06 [0.89;1.26]	0.99[0.76;1.28]
Total ACE score	1.09[0.87;1.36]	1.08[0.86;1.35]	REF	1.12[0.94;1.33]	1.09[0.85;1.42]

Notes. Models adjust for gender, context, age, sexual identity, and cohabitation/marital status. ACEs factors were tested in separate models.

^tp<0.10

^*p≤0.05

^**p<0.01

^***p<0.001

With regards to ACEs exposure, we found that children with higher scores on the child maltreatment/violence ACEs factor had greater odds of being in the multi-risk class (class 5) and the single partner, low activity class (class 2) (OR=1.28[1.00;1.64], p=.05 and OR=1.22[0.97;1.52], p<.10, respectively) than class 3 (the referent group), although the latter finding was only a trend. Adding the child maltreatment/violence ACEs factor score into the model did not reduce the size of the coefficient for the context effect (OR=1.93[1.35;2.77], for class 4, and 1.78[1.09;2.91], for class 5). Neither the parent-related ACE factor nor the total ACE score showed significant associations with the outcome.

Post-hoc, we examined child maltreatment/violence ACEs individually to determine whether specific adversities were driving the observed association. The marginally significant effect on the single partner, low activity class appeared to be driven by neglect (OR=1.90 [1.10;3.28]), p<.05 and the association with the multi-risk class was driven by both neglect (OR=1.87[1.02;3.44]), p<.05 and exposure to violence (OR=1.84[1.08, 3.14], p<.05). Ancillary analyses showed that associations between context and class membership did not vary by gender and that the relationship between ACEs and class membership was similar across genders and contexts (available upon request).

Discussion

This study describes patterns of sexual behaviors among Puerto Rican young adults living in two contexts during a developmental period when sexual activity and risky behaviors are at their peak. We identified five sexual behavior patterns ranging from current inactivity to multiple, co-occurring SRB (currently inactive; single partner, low activity; single partner, inconsistent condom use; single partner, no condom use; multi-risk). Young adults living in a minoritized context (SBx), males, and those who experienced greater ACEs related to child maltreatment and exposure to violence were more likely to engage in behavior patterns that included behaviors usually associated with risk for adverse outcomes (e.g., multiple sex partners, sex while intoxicated). Interestingly, we found that early age at first sex did not differentiate patterns of sexual behavior and that none of the groups were characterized by consistent condom use. Even though ACEs reflecting maltreatment and violence exposure were associated with risky sexual behavior patterns, they did not explain differences in patterns of sexual risk by context.

In the current sample of Puerto Rican young adults with prior sexual experience, the most prevalent behavior pattern (32.19%) was one characterized by being sexually active in the past three months with a single partner who is not considered to be high-risk, not engaging in risk behaviors, but having similar probabilities of using condoms “consistently”, “always”, or “never”. This class was common among those who were married or cohabiting. In low-risk environments, this group could be perceived as low risk. However, in contexts where STIs/HIV are elevated, as is the case for the Bronx, which has some of the highest rates of people living with HIV/AIDS and reported cases of chlamydia, gonorrhea, and syphilis (Centers for Disease Control and Prevention, 2018; New York State Department of Health, 2017), any unprotected sex can be risky. This is especially true if there is a concurrent sexual partner without the other partner’s knowledge (Morris et al., 2009).

None of the groups were characterized by consistent condom use. This may reflect a lack of use of protection, or a reliance in methods that protect against pregnancies but not STIs/HIV. Of note, only a few participants reported not using condoms because they were trying to conceive (n=37). Overall, there has been an increase in the use of highly effective contraceptive methods (e.g., long-acting, reversible contraceptives) in the U.S. (Kavanaugh et al., 2015). While such an increase is positive insofar as it can effectively reduce unintended pregnancies, these methods do not protect against STIs/HIV. The use of dual methods in this population seems particularly important, given the high rate of both teen births and STIs/HIV.

Unexpectedly, early age at first sex did not differentiate the classes. Early sexual debut has been linked to greater number of sex partners, sex with high-risk partners, and sex while intoxicated (Epstein et al., 2014; O’Donnell et al., 2001). For this reason, early sex is considered an indicator of future sexual risk. Others have found that age at initiation of sex distinguished youth’s behavior patterns, with early initiation of sex co-occurring with other SRB (Heywood et al., 2015; Pflieger et al., 2013; Vasilenko et al., 2015). It is possible that in this population, starting to have sex at a young age is not necessarily maladaptive but rather an adaptive response to a context of adversity (National Academy of Sciences, 2019). The age at first sex varies considerably across contexts and nations (Wellings et al., 2006). In a population-based sample of Puerto Rican adults on the island, 38% of men and 21% of women reported first having sex by age 15 (Ortiz et al., 2011). In the U.S., 18% of men and 13% of women have had sex by that time (Martinez & Abma, 2015). It may be that as a group, Puerto Ricans start having sex at an earlier time, making this is a normative behavior and thus, not a risk indicator.

This study extends previous work by examining how context can shape which behavior patterns are most likely to be observed in a population. Compared to young adults from Puerto Rico, young adults from the SBx were more likely to be in the “multi-risk” class and the “active, sex without condoms class”, which include behaviors usually associated with risk for STIs/HIV. This finding is consistent with previous studies of the BYS sample showing that youth in the SBx are exposed to more risks and are more likely to engage in problematic behaviors (e.g., antisocial behaviors (Bird et al., 2007), bullying (Morcillo et al., 2015)) than those in PR. Youth in SBx also have higher levels of sensation seeking than those in PR (Martins et al., 2015), which can contribute to greater risk-taking.

In addition to identifying contextual differences in the prevalence of different behavior patterns, we found that ACEs reflecting maltreatment and violence exposure were associated with risky sexual behavior patterns. Previous studies had identified associations between maltreatment and several SRB examined independently (Hahm et al., 2010; Hillis et al., 2001; Oshri et al., 2011; Senn & Carey, 2010). An emerging literature also documents links between exposure to violence and SRB (Voisin et al., 2011). We expand extant research by showing that exposure to these adversities relates to more complex sexual behavior patterns. The child maltreatment/violence ACEs factor score was associated with sexual behavior patterns characterized by multiple, co-occurring risks. It has been theorized that children exposed to traumatic experiences like maltreatment, develop insecure attachment styles that can be transferred to their romantic relationships later on (Briere, 2002; Cicchetti & Toth, 2005). Individuals with insecure attachment styles can either have difficulty engaging in steady relationships and move from one brief encounter to the next, or they might desperately seek acceptance from others and fail to negotiate safe sex practices with their partners to avoid rejection, or both (Fraley & Shaver, 2000).

We expected ACEs to account for contextual differences in patterns of sexual behaviors, but this was not the case. It is still possible that differences in patterns of sexual behaviors are related to greater exposure to other stressors among young adults in the SBx. As members of a minoritized group in the continental U.S., Puerto Rican young adults in the SBx experience minority-related stressors like discrimination (Ramos-Olazagasti et al., 2013) that may elevate risk-taking. A higher prevalence of mental health problems among young adults in the SBx may also explain the observed differences. A previous report indicated that youth in the SBx had higher levels internalizing problems than those in PR, and that this difference was accounted for by greater exposure to violence and discrimination (Ramos-Olazagasti et al., 2013). In young adulthood, young adults in the SBx had a higher prevalence of past year depression, generalized anxiety disorder, and substance use disorder (Duarte, Canino, Wall, et al., 2021). Mental health problems are associated with SRB (Elkington et al., 2010; Lehrer et al., 2006) and with patterns of behaviors characterized by multiple co-occurring risks (Tubman et al., 2003). Lastly, differences may be related to a loss of cultural values that are protective against risk behaviors among young adults in the SBx. Risk taking usually increases with acculturation to the Anglo culture (Afable-Munsuz & Brindis, 2006) and the loss of certain Hispanic values like familism (or a strong family orientation) (Guilamo-Ramos et al., 2009), religiosity, simpatia, and collectivist values (Ma et al., 2014; Schwartz et al., 2011).

Family/parental dysfunction ACEs were not related to young adults’ patterns of sexual behavior in the sample, neither did an overall score reflecting the totality of ACEs. Evidence linking family/parental dysfunction ACEs and the accumulation of ACEs with SRB is not as strong as the research on maltreatment. While some had found associations between an ACEs count or an overall ACEs score and SRB (Hillis et al., 2001; Kogan et al., 2016; VanderEnde et al., 2018), those studies had limited generalizability due to the sampling characteristics of the studies (e.g., only women (Hillis et al., 2001), rural black men (Kogan et al., 2016)). There may be specificity in how ACEs relate to outcomes, with stronger evidence for an association with maltreatment. Alternatively, in a sample where family/parental dysfunction ACEs were common given the high prevalence of divorce and separation, only ACEs that affect children more directly have effects on their sexual behaviors.

Study limitations include the absence of information about other contraceptives, a low prevalence of anal sex that made it impossible to distinguish between protected and unprotected sex, a rather restrictive definition of gender identity (male versus female) and sexual orientation (heterosexual versus not heterosexual), and limited information about the timing and chronicity of exposure to ACEs. Our sample is also restricted to two geographic areas in the two contexts. Notwithstanding, to our knowledge, this is the first study to describe patters of sexual behaviors in a well-defined Hispanic subpopulation. However, even the Puerto Rican population is diverse in terms of phenotype, race, and sexual identity. Further research is needed to explore within-group variation in how ACEs are experienced and their relationship to sexual behaviors across subgroups. Moreover, this study was only a first step at exploring differences in patterns of sexual behaviors by social context. Future research should further examine the interrelationship between sexual behaviors, sexual orientation, cultural factors (e.g., religiosity, acculturation), and other contextual influences.

Our findings have clear implications for intervention. Several sexual health interventions have been applied to Hispanic populations, with some success in reducing SRB (Cardoza et al., 2012; Sutton et al., 2014). Findings suggest that one-size-fits-all approaches may not be as effective as tailored solutions. Young adults in the currently inactive and single partner, low activity classes would benefit from primary prevention programs that target knowledge, attitudes, beliefs, and intentions as they were somewhat younger and therefore might be less sexually experienced. A higher proportion of young adults who identified as gay, lesbian, or bisexual belonged in that class, relative to those who identified as heterosexual. We do not know if they were currently inactive or having oral sex only as a preventive measure, or because they had not had the opportunity to engage in other behaviors. Prevention strategies that prepare youth for sexual behaviors that might arise are likely appropriate for these groups.

Groups of people who were sexually active with a single, seemingly low-risk partner and were using condoms inconsistently would benefit from interventions that reinforce current behaviors and support healthy sexual behavior decision-making. Because women were more likely to be in this class, a focus on gender-specific interventions that empower women and increase their self-efficacy in negotiating safe sex practices could be successful. This might be especially the case for Hispanic women as cultural values around machismo often deter women from having an active voice in sexual decision making (Pulerwitz et al., 2002). At the same time, women empowerment interventions should be careful not to place the burden exclusively on women by not explicitly addressing machismo in couples’ relationships. The intervention Cuidate! presents participants with an alternative definition of machismo that includes a responsibility of caring for others and keeping the family safe by practicing safe sex behaviors. This intervention capitalizes on traditional female gender roles (e.g., prioritizing the family) to reinforce the importance of fidelity and normalizing the use of condoms to protect the family (Villarruel et al., 2005).

More information about the use of contraceptives is needed to fully understand the sexual behaviors of young adults in relationships with a single, seemingly low-risk partner, but who tend not to use condoms ever. It is possible that young adults view condoms only as a pregnancy prevention method and do not consider it necessary during anal sex, or when they are using other contraceptives. Interventions should focus on the importance of dual methods to protect against pregnancy and STIs/HIV. This is particularly important if there is any suspicion of infidelity (Macauda et al., 2011). Young adults from the SBx were more likely to be in the multi-risk class, as were those exposed to child maltreatment/violence ACEs, and those who identified as gay, lesbian, or bisexual albeit this was the least common behavior pattern. These young adults would benefit from intensive behavioral intervention along with pre-exposure prophylaxis (PrEP). Interventions for this group should also reinforce the importance of condom use consistency to ensure protection.

Interventions that focus on strengthening family relationships, as is the case of Familias Unidas (Prado et al., 2007), have been successful at reducing some SRB among Hispanic youth, as have some interventions directly targeting Hispanic men who have sex with men, like Sin Buscar Excusas (O’Donnell et al., 2014) (for reviews of interventions to reduce HIV among Hispanic populations, see Herbst et al., 2007; Pérez et al., 2018). However, in families where youth have experienced child maltreatment and violence exposure, additional supports such as providing trauma-informed care (Ko et al., 2008; SAMHSA, 2014), could help alleviate youths’ suffering, rebuild parental relationships, and develop strategies to engage in healthy romantic relationships. Universal prevention programs like the Triple-P Positive Parenting Program that have been shown to reduce maltreatment (Prinz et al., 2009) are also needed.

In sum, Puerto Rican young adults in two sociocultural contexts displayed different patterns of sexual behaviors. Riskier sexual behavior patterns occurred in the SBx compared to PR and were related to child maltreatment/violence ACEs. Interventions for young adults displaying different types of sexual behaviors are needed and need to be tailored to specific needs and contexts.