Sexual Health Programs for Latinx Adolescents: A Meta-analysis

We conducted a comprehensive search of PsycINFO, Cumulative Index to Nursing and Allied Health Literature, and PubMed databases to extract relevant studies published through January 2019. We used the following key words: youth, adolesc* or teen*; sexual health or safe* sex or sexually transmitted disease or sexually transmitted infection or STD or STI or HIV or AIDS or pregnancy or reproductive health or condom* or contracept* or unprotected sex or abstinence; intervention or program or education or prevention or promotion or trial; latino* or latina* or latinx* or minorit* or ethnic* or hispanic or african american* or black* or race or racial or biracial. As is evidenced by our key words, we initially planned to do one meta-analysis on interventions for Latinx and black adolescents but ultimately determined it would produce more meaningful results to understand each population separately.⁴³  Other studies were located by examining previous reviews and meta-analyses^{18,28–32,44–46}  and examining the reference lists of all included articles. This initial search produced 3068 different articles.

Studies were included if they met the following criteria: (1) included a US-based sample; (2) sampled Latinx adolescents (ie, at least 95% of the sample was Latinx, no participants were >24, and the mean sample age was ≤18); (3) evaluated the effects of a primary prevention intervention with a component aimed at promoting abstinence and/or safer sex behavior (as is common in similar meta-analyses,^34,47  studies focused on secondary prevention were excluded^48,49 ); (4) evaluated effects of an intervention by using an experimental or quasiexperimental design; (5) included at least 1 of 4 behavioral outcome measures: abstinence, condom use, number of sexual partners, or pregnancy; (6) were published in English; and (7) provided sufficient statistics to calculate effect sizes. When a study had a mixed-ethnicity sample but included subgroup analyses that evaluated intervention effects among only Latinx participants, the results of the subgroup analyses were included.^24,26,50  We selected the sexual health intervention that was most comprehensive when studies had more than one intervention group. If researchers of multiple studies reported findings using the same data, the study with adequate data to calculate effect sizes and reporting results for the longest-term follow-up was included as the most conservative estimate of treatment effects. All other studies using the same data were excluded. Similarly, for studies with multiple follow-up points, we used the longest-term follow-up with adequate data to calculate effect sizes. When studies included more than one indicator for an outcome (eg, multiple indicators of condom use), we used a random number generator to select one outcome to eliminate potential bias.⁵¹ 

These selection criteria resulted in a final sample of 12 articles (Fig 1). From these, we calculated 9 independent effect sizes for abstinence, 11 for condom use, 7 for number of sex partners, 3 for sexual health knowledge, 3 for sexual health intentions, and 2 for sexual health self-efficacy. Only researchers of 1 study examined pregnancy as an outcome,²³  so we were not able to meta-analyze the effectiveness of pregnancy prevention across studies. See Table 1 for examples of outcome variables from primary studies that were used in analyses.

Two authors independently coded each of the studies during the month of June 2019, which involved reading the full text of each article included in the meta-analysis and recording study characteristics. The following data were extracted: (1) demographic and sample characteristics (eg, mean age, sex), (2) intervention characteristics (eg, setting, dose), and (3) methodologic characteristics (eg, length of follow-up, retention rate). The mean percentage agreement across all coding categories was 94%. Discrepancies between coders were resolved through group discussion until a consensus was reached. Risk of bias was calculated by using criteria adapted from the Cochrane risk-of-bias tool.⁵² 

As the indicator of effect size, the standardized mean difference, Cohen’s d, was used. Effect size d can be interpreted as small (0.20), medium (0.50), or large (0.80).⁵³  If d’s and confidence intervals (CIs) were reported in an article, they were extracted. Otherwise, other statistics that could be converted to d’s (eg, summary statistics, odds ratios) were calculated, using Comprehensive Meta-Analysis V2.0⁵⁴  and the Practical Meta-Analysis Effect Size Calculator.⁵⁵  When no statistics in the study could be converted to a d, appropriate data were requested from study authors. To ensure consistency and interpretability of effect sizes, higher values for the following outcomes always indicate that the sexual health intervention group performed better than the control: abstinence, condom use, sexual health knowledge, sexual health intentions, and sexual health self-efficacy. Lower values for number of sex partners always indicate that the sexual health intervention group performed better than the control.

To allow for the possibility of differing variances across studies, random-effects meta-analytic procedures were used for the primary analyses across all independent effect sizes.⁵¹  As recommended by the Cochrane Collaboration,⁵⁶  sensitivity analyses were conducted for behavioral outcomes: studies with high risk of bias were excluded to determine if they were effecting the results.

The Q statistic and I² were used to determine if significant statistical heterogeneity existed among effect sizes. Effect sizes and 95% CIs for hypothesized categorical moderators were calculated, and those effect sizes were compared by using the Q_b statistic. For categorical moderators, mixed effects models were used for these analyses to allow for the possibility of differing variances across subgroups. To test continuous moderators, random-effects meta-regression was used. The Q-value model statistic was used to determine if there was significant moderation for these models. Analyses were conducted by using Comprehensive Meta-Analysis V2.0.⁵⁴ 

A summary of the studies included in this meta-analysis is presented in Table 2. A total of 4673 Latinx adolescents (weighted mean age = 14.34) were included across 12 studies. Researchers of most studies targeted mixed-sex samples (10), whereas 2 included only female adolescents, and no studies included only male adolescents. Substantial cultural tailoring, defined as the incorporation of Latinx-specific practices and values (eg, familialism) into intervention materials, was included in 8 of the 12 studies. Intervention dose varied widely across studies: 3 studies included <10 hours of program content, and 4 studies included >20 hours of content. Only researchers of one study specified that all participants received the full program dose; many did not provide information about the dose participants received.

Risk of bias was low across studies (see Supplemental Table 6). Researchers of >80% of studies randomly assigned participants to study conditions, and 11 out of 12 studies had <30% of participants in the full sample drop out or become lost to follow-up. However, in many studies, it was unclear if other relevant data were gathered that were not reported.

Individual study effect sizes for abstinence ranged from d = −0.12 (95% CI: −0.69 to 0.45) to d = 1.39 (95% CI: 0.28 to 2.51), with an overall weighted mean effect size across studies of d = 0.15 (95% CI: 0.02 to 0.28; P = .02) (Fig 2). This suggests that, on average, sexual health interventions have a small but significant effect at improving rates of abstinence among Latinx adolescents (Table 3). We calculated a fail-safe N to determine the robustness of this finding: 12 additional nonsignificant studies would have to exist for the combined P value to exceed .05. There was not significant statistical heterogeneity among studies (Q = 10.13, df = 8, P = .26, I² = 21.06); thus, we did not examine moderators of intervention success for abstinence.

Individual study effect sizes for condom use ranged from d = −0.31 (95% CI: −0.59 to −0.02) to d = 1.26 (95% CI: 0.69 to 1.83), with an overall weighted mean effect size of d = 0.44 (95% CI: 0.18 to 0.70; P = .001) (Fig 3). On average, sexual health interventions have a significant effect of moderate size on condom use among Latinx adolescents (Table 3). A fail-safe N indicated that there would have to be 121 additional nonsignificant studies for the combined P value to exceed .05.

There was significant statistical heterogeneity among studies for the condom use outcome (Q = 48.33, df = 10, P < .001, I² = 79.31); thus, we examined moderators of intervention success (see Tables 4 and 5). Cultural tailoring moderated the association between sexual health interventions and condom use. Effect sizes were larger for studies that evaluated interventions that were culturally tailored (d = 0.57, 95% CI: 0.33 to 0.81; P < .001), compared with those that evaluated interventions that were not culturally tailored (d = 0.01, 95% CI: −0.39 to 0.42; P = .68). Effect sizes did not significantly differ across the following moderators: participant age, parent involvement in the intervention, intervention completion, inclusion of communication skills training in the intervention, inclusion of condom skills training in the intervention, dose of the intervention, intervention setting, year of study publication, and length of follow-up. We examined sex as a moderator and found no significant differences between the effectiveness of interventions with mixed-sex samples and those with girls only; however, there were no studies examining the effectiveness of a sexual health intervention among boys only. This limited our ability to evaluate sex as a moderator.

Individual study effect sizes for number of sex partners ranged from d = −1.55 (95% CI: −2.59 to −0.51) to d = −0.03 (95% CI: −0.55 to 0.50), with an overall weighted mean effect size of d = −0.19 (95% CI: −0.37 to −0.001; P = .049) (Fig 4). This indicates that compared with control conditions, across studies, the sexual health interventions were associated with a reduction in the number of sex partners among Latinx adolescents (Table 3). There was not significant heterogeneity among studies (Q = 9.25, df = 6, P = .16, I² = 35.11); thus, we did not examine moderators of intervention success. We calculated that there would have to be 8 additional nonsignificant studies for the combined P value to exceed .05 (fail-safe N = 8).

Individual study effect sizes for sexual health knowledge ranged from d = 0.13 (95% CI: 0.01 to 0.26) to d = 0.58 (95% CI: 0.33 to 0.84), with an overall weighted mean effect size across studies of d = 0.40 (95% CI: 0.09 to 0.70; P = .01), indicating that overall, sexual health interventions were efficacious in promoting sexual health knowledge. However, there was no significant association between sexual health interventions and sexual health intentions or sexual health self-efficacy. Effect sizes for sexual health intentions ranged from d = −0.13 (95% CI: −0.33 to 0.07) to d = 0.39 (95% CI: −0.14 to 0.92), with an overall weighted mean effect size of d = −0.001 (95% CI: −0.18 to 0.18; P = .995). Effect sizes for sexual health self-efficacy ranged from d = −0.05 (95% CI: −0.17 to 0.08) to d = −0.08 (95% CI: −0.28 to 0.11), with an overall weighted mean effect size of d = −0.06 (95% CI: −0.16 to 0.05; P = .30). We did not test for heterogeneity in these outcomes because of the small number of studies for each (knowledge k = 3; intentions k = 3; self-efficacy k = 2 [k refers to the number of studies]).

Three studies were determined to have high risk of bias.^25,50,58  When these studies were excluded from analyses, there were no substantial differences in our findings for the 3 behavioral outcomes: abstinence, d = 0.16 (original d = 0.15); condom use, d = 0.36 (original d = 0.44); and number of sex partners, d = −0.19 (original d = −0.19). All effects remained significant.

In this meta-analysis, we synthesized >20 years of research on the efficacy of sexual health interventions among Latinx adolescents. Overall, results from 12 experimental or quasiexperimental studies with 4673 Latinx adolescents reveal that youth who participated in a sexual health intervention, compared with a control group, were significantly more likely to abstain from sexual activity, use condoms, have fewer sex partners, and report greater sexual health knowledge. Latinx adolescents are disproportionately burdened with unplanned pregnancy and STIs.¹  In this meta-analysis, it is shown that sexual health interventions can play a role in combatting these health disparities.

Although sexual health interventions had a significant protective influence on Latinx adolescents’ sexual behavior, these effects were, on average, small to moderate. The overall effect sizes for the abstinence and sex partners outcomes were particularly small (abstinence d = 0.15; number of sex partners d = −0.19). This could be because there are factors that impact the sexual behavior of Latinx adolescents that are outside of the scope of most sexual health interventions. Most of the interventions included in this meta-analysis targeted individual or family-level risk and protective factors (eg, sexual health knowledge; parent–child sexual communication); however, there are factors at community and society levels that also impact Latinx sexual health (eg, access to sexual health care; stigma).⁶¹  If sexual health interventions target factors beyond individual and interpersonal levels of influence, they may have an even larger effect on the sexual behavior of Latinx adolescents.⁶²  Researchers may consider developing and evaluating interventions with components that increase the cultural understanding of clinicians that work with Latinx communities or that provide low-cost, ongoing sexual health services in Latinx communities.

Although the overall effect of sexual health interventions on condom use was protective, there was a significant amount of statistical heterogeneity; thus, this finding must be considered in light of moderation analyses that elucidate which interventions were most effective: namely, those with cultural tailoring. Most interventions (8) were culturally tailored. For example, in the ¡Cuıdate! program,⁵⁹  salient aspects of Latino culture, specifically familialism and gender role expectations, were incorporated. The ¡Cuıdate! program highlighted Latino cultural values and targeted behaviors related to health issues affecting Latino individuals. Cultural tailoring moderated the association between sexual health interventions and condom use, such that interventions that were specifically tailored to meet the needs of Latinx adolescents were more effective than those with no explicit cultural tailoring. Indeed, 2 studies that found interventions had a significant, large effect on condom use included cultural tailoring,^21,25  and in the study with the largest effect size, researchers collaborated closely with community members to develop the intervention being tested. In the other study with large effects,²¹  the program was specifically designed for Hispanic families. Researchers of many studies with moderate to large effect sizes evaluated the Familias Unidas intervention,^19–21  which was grounded in ecodevelopmental theory, a theory that highlights the influence of family and culture on individuals.⁶³  These findings are consistent with previous meta-analyses, which found that culturally sensitive programs are effective in reducing externalizing behaviors among Latino youth.⁶⁴  Researchers of the current study extend this line of work by illustrating that sexual health interventions, specifically, are more effective in promoting condom use among Latinx adolescents when they are tailored. This may be because culturally tailored interventions are more congruent with the identities and lived experiences of Latinx adolescents, which increases the likelihood that these programs will modify adolescent behavior.⁶⁵  Additionally, culturally tailored programs may promote cultural identity, which is tied to healthier lifestyles among Latinx adolescents.⁶⁶  Interventionists should consider the importance of cultural relevance to intervention success.

Regarding abstinence and number of sex partners, readers should note that our fail-safe n was small, indicating that only 8 to 12 nonsignificant studies would be needed for the overall weighted effect of sexual health interventions to be nonsignificant. Thus, readers should take caution in interpreting these findings.⁶⁷ 

In conducting this review, limitations and future directions related to intervention design, evaluation, and reporting became clear. First, although we evaluated the association between sexual health interventions and 3 potential psychological mediators of intervention effects on sexual behavior (knowledge, self-efficacy, intentions),^40,41  only improvements in sexual health knowledge were significantly associated with sexual health interventions among Latinx adolescents. This is likely because few studies could be included in this analysis: for example, researchers of only 2 studies examined sexual health self-efficacy. If researchers evaluating behavior change also more frequently evaluated psychological mediators, this might provide a better understanding of how the most effective programs are working. In this case, future sexual health programs could target the mediators that more frequently lead to behavior change. Additionally, there may not have been significant changes in self-efficacy and intentions because oftentimes adolescents start “high” on these protective constructs at pretest. If adolescents are already high in self-efficacy and have strong intentions to practice safer sexual behaviors before they take part in a sexual health program, then there is not much room to improve in response to the intervention.

Second, there was methodologic heterogeneity in the measures used to evaluate sexual health outcomes across studies. For example, in evaluating abstinence, some researchers asked about “vaginal intercourse,”²⁵  whereas others asked about “vaginal, oral, and anal sex.”^26,57  Also, the time frame participants were asked to report on varied. Some researchers asked participants about their condom use the last time they had sex,^24,26,50  whereas others asked about condom use in the last 2 to 3 months.^21–23  Because there is currently no “gold standard” for assessing adolescent sexual behavior, variability in outcome assessments across studies is expected. Yet, more meaningful synthesis of research would be possible if the methods used to assess adolescent sexual health outcomes were standardized, as the World Health Organization is currently proposing in their development of a population-representative sexual health survey instrument.⁶⁸ 

Third, data regarding many important sample characteristics were missing from studies included in this meta-analysis. We hoped to more thoroughly report and analyze participant characteristics, thus highlighting the clinical heterogeneity that exists across groups of Latinx adolescents in the United States. However, inconsistent and brief reporting of participant characteristics across studies made this difficult. For example, researchers of only 6 studies reported the percentage of participants born in the United States. More thorough reporting of this participant characteristic could have been informative; however, we recognize that this could be a sensitive topic with respect to immigration issues, and researchers may have decided to omit questions regarding country of origin and citizenship to protect participants.

Fourth, initially we planned to analyze the association between sexual health interventions and pregnancy among Latinx adolescents; however, only 1 study that met our inclusion criteria evaluated pregnancy as an outcome.²³  This is unfortunate considering Latinx adolescents experience unplanned pregnancies at high rates¹ ; thus, programs that are successful at reducing unplanned pregnancy are urgently needed. Researchers of future studies should evaluate biological outcomes, such as pregnancy and HIV transmission.^33,69 

Fifth, programs must evaluate and improve their implementation strategies. Of the 12 studies included in this review, only 1 indicated that all participants completed the entire intervention. Research on general educational opportunities suggests Latinx adolescents face barriers that may affect their intervention attendance⁷⁰ ; however, if these barriers are not identified through implementation evaluation, they will be impossible to eliminate. Emerging literature on intervention science insists that to maximize the sustained success of evidence-based interventions, the appropriateness and feasibility of health programming must be evaluated.⁷¹  Researchers evaluating sexual health interventions might consider including a comprehensive process or implementation evaluation to understand why some Latinx adolescents are not receiving the full program dose. Although researchers of some studies included in this meta-analysis evaluated attendance,²⁴  attrition,^20,59  and/or intervention fidelity,^19,20,57  feasibility and cost were only assessed in 1 study.²³ 

Finally, none of the interventions we identified in this review specifically targeted Latinx boys. Intersectionality theory contends that we cannot address disparities by focusing solely on race or ethnicity and must also consider within group experiences across sex and class.⁷²  Thus, Latinx boys’ experiences of health and well-being and intervention may be distinct from Latinx girls.^72,73  It follows then that sexual health programs must target the unique needs of this group specifically to be most effective, and effects of these programs should be considered separately.

Results from 12 studies with 4673 Latinx adolescents reveal that sexual health interventions are associated with improvements in abstinence, condom use, number of sex partners, and sexual health knowledge. Although effect sizes are small to moderate, sexual health programs do improve some sexual behaviors among Latinx adolescents. Findings from this meta-analysis highlight the following important future directions for those invested in improving the sexual health of Latinx adolescents: (1) sexual health interventions must be tailored specifically to the practices and experiences of Latinx adolescents, (2) there is a need for more thorough evaluation of sexual health outcomes and reporting of sample characteristics, (3) future research could evaluate the impact of sexual health interventions on biological outcomes, (4) more research is needed on how to successfully implement sexual health programs in real-world settings, and (5) additional programs aimed at improving the sexual health of Latinx boys are needed. Finally, these sexual health interventions are largely effective at improving the sexual health of Latinx adolescents; therefore, it is important that culturally tailored sexual health programs be available to Latinx communities across the United States.

CI

confidence interval

STI

sexually transmitted infection