Evaluating the Impact of a Youth‑Led Sexual Violence Prevention Program: Youth Leadership Retreat Outcomes

Abstract

Involving youth in developing and implementing prevention programs to reduce sexual violence (SV) has the potential to improve prevention outcomes. However, there has been little focus on youth-led SV prevention programs, and limited evaluation research to help guide efforts. The current study examined the effectiveness of Youth Voices in Prevention (Youth VIP) leadership retreats on SV victimization and perpetration, forms of violence related to SV (e.g., bullying), SV bystander behaviors and readiness, and perceptions of norms related to SV prevention. Results identified mixed findings for program impact, with variations in outcomes that can help guide future youth-led prevention program initiatives. Youth attending a large “kick-off” leadership retreat (that was less youth-led that subsequent smaller retreats) later reported more bystander behaviors, but also reported increased perpetration and victimization, compared to non-attending youth. However, youth attending smaller, more focused leadership retreats held during the school year, reported reductions in sexual harassment perpetration and improved bystander behaviors and attitudes compared to non-attending youth. Evaluation of moderator variables suggests that program impact was generally stronger for younger participants, sexual minority youth, and non-White youth (which were largely Native American youth in this sample). Findings suggest promise for youth-led prevention work but also highlight the need for testing the impact of different training structures and modalities. Clinical trials number: NCT03207386.

Sexual violence (SV) is a pernicious public health issue that disproportionally impacts middle and high school students (Centers for Disease Control & Prevention, 2020; Rinehart et al., 2008; Siller et al., 2020). To date, however, few primary prevention programs have been successful in reducing rates of SV victimization and perpetration (DeGue et al., 2014; Edwards & Banyard, 2018) although there are exceptions (Coker et al., 2017; Espelage et al., 2015; Foshee et al., 2004; Taylor et al., 2015). One possible solution to enhancing the effectiveness of SV prevention efforts is to include youth in the development and implementation of such efforts (Edwards et al., 2016). Indeed, research suggests that the effectiveness of prevention programming decreases as youth transition from childhood to adolescence (Finkelhor et al., 2014; Yeager et al., 2015). Traditional adult-developed and facilitated prevention efforts may clash with adolescents’ increasing desire for autonomy and increasing influence of friends and peers (Banyard et al., 2020; Maxwell, 2002), which could contribute to reduced effectiveness of SV prevention initiatives (Edwards et al., 2016; Finkelhor et al., 2014; Yeager et al., 2015). Further, most evaluated SV prevention takes place in educational contexts, and we have less understanding of how prevention outside of school may impact adolescents’ behavior.

To date, youth-led SV prevention efforts have not been rigorously evaluated, although there is documentation of youth being involved in leadership roles in SV prevention programs. In a study of adult practitioners involved in SV prevention efforts, researchers noted that “peers as educators” was the most common method of youth involvement, although youth also served on advisory boards related to SV prevention and were involved in community and social action-oriented projects, such as designing social marketing and media campaigns around violence prevention (Weisz & Black, 2009). There is also some limited discussion of the role of diffusion of innovation as a theory for prevention and how youth, as popular opinion leaders (POLs) among their peers, can spread positive norms and prevention messages (Cook-Craig et al., 2014).

Formal youth-led prevention has been implemented more extensively in other health-behavior fields, such as substance abuse and sexual health, with findings about the effectiveness of these efforts showing mixed results. For example, research in substance abuse prevention suggests that youth-led efforts are more effective than adult-led efforts (Cuijpers, 2002; Mellanby et al., 2000). Conversely, in the area of sexual and reproductive health, findings for the efficacy of youth-led efforts have been more mixed and less promising (Chandra-Mouli et al., 2015; Kim & Free, 2008; Tolli, 2012). Research documents that well-trained facilitators are one of several key components to effective prevention (Nation et al., 2003). Facilitator training may be even more important for youth who are less experienced in both content and delivery of prevention programs than adults. In fact, this lack of experience may, in part, explain the mixed findings for some youth-led prevention efforts in the areas of substance abuse and sexual health.

Also, most prevention curricula are implemented in school settings, a logical location given the amount of time youth spend in school. However, we also know that schools have many competing demands for their time and are not always willing or able to partner to implement the most effective, comprehensive, skills-based SV prevention programs. Thus, interest is growing in taking prevention outside the school, which may be especially important for youth with high levels of absenteeism and likely most at-risk for SV (Eaton et al., 2008; GarcĆa & Weiss, 2018; Hawkins et al., 2000). To date, however, very few SV prevention programs for youth have been implemented outside of the school setting, and none of them are youth-led. As such, the purpose of the current project was to evaluate the effectiveness of a youth-led, afterschool SV prevention program (i.e., Youth Voices in Prevention [Youth VIP]) that took place in community spaces (see next section for an overview of Youth VIP). For a detailed discussion of Youth VIP programming, see: https://www.wavi.org/assets/docs/uploads/youth-vip/youth-vip-toolkit.pdf

Youth Violence in Prevention Overview

Overview

Youth VIP was designed using empirically driven components for best practices in SV prevention (Basile et al., 2016a). These included training in bystander intervention, promoting healthy social norms, and diffusion of innovation skills (Basile et al., 2016a; Edwards et al., 2019; Orchowski, 2019). The program also drew from wider research on evidence-based practices in interpersonal violence prevention by incorporating aspects of social-emotional learning skills (Basile et al., 2016a; Taylor et al., 2017) as well as recent calls in the literature to enhance youth leadership in SV prevention (Edwards et al., 2016).

Programming Components

Youth VIP consisted of four distinct types of events: retreats, action events, project campaigns, and internships. In this paper, we report on outcomes associated with retreats, and in another paper, we report on outcomes associated with other time-limited events, specifically action events and project campaigns (Banyard et al., in press). The objective of Youth VIP retreats was to build youth capacity and leadership to recognize SV and SV prevention best practices (i.e., bystander intervention, social emotional skills, social norms) with the goal that they would share the information and skills that they learned with their peers (i.e., diffusion). Given the large presence of Native American youth in the community, cultural components (e.g., Lakota traditions and cultural practices) were infused throughout all program activities.

Given that the “kick off” training retreat was for the primary purposes of demonstrating acceptability and feasibility of the Youth VIP programming, this event was examined in separate analyses from subsequent training retreats. This large “kick off” retreat was conceptually different from the small retreats; the “kick off” retreat was much larger and did not include as much skills-based prevention content as the subsequent small retreats. Organizers learned many lessons during the large retreat, which resulted in higher quality among the subsequent smaller retreats.

Youth Leadership

Whereas at the first camp the programming was largely developed and facilitated by adults, youth leaders played a much larger and more meaningful role in working alongside adults to create and deliver the prevention content at subsequent smaller retreats. Youth received training prior to implementing prevention content at smaller retreats and support at smaller retreats from adult mentors in delivering prevention content. At all retreats, youth were encouraged to share the information that they learned with their peers, representing another opportunity for youth-leadership. Following retreats, youth created action events and project campaigns, often through paid internship, that were related to SV prevention. Youth created the focus and activities of the events and campaigns with the mentorship of adults. For outcomes associated with action events and project campaigns, see Banyard et al. (in press). Thus, although youth retreats had youth-led components, action events and project campaigns subsequent to the retreats were more truly youth led.

Youth Selection

Social network analyses performed on survey data where youth named up to seven best friends was used to identify popular opinion leaders (POLs) by school. Youth in the top 10% of nominated peers in each school were invited to attend the leadership retreats. When response rates were low, more youth based on their POL position were invited until the maximum number of youth for each camp was achieved. Youth who were not nominated as a POL were invited to attend although priority was given to POLs initially via special invitations and a reward ceremony that celebrated these youth as leaders. The purpose of retreats was to train youth in best practices SV prevention (see Supporting Information Appendix A and below for further details). Thus, a first component of the program being youth-led was providing youth with leadership and prevention training in a collaborative atmosphere.

Current Study

Research suggests that SV intersects with other forms of violence such as physical teen dating violence, bullying, and sexual harassment (Hamby & Grych, 2013; Ozer et al., 2004; Siller et al., 2020) and that prevention programs focused on one type can have carryover effects to reduce other forms (DeGue et al., 2021; Vivolo-Kantor et al., 2021). Thus, it is important to measure other forms of interpersonal violence as secondary outcomes, which we did as part of this project. As such, our first specific aim was to examine the short-term impact of a large “kick-off” leadership retreat on the primary outcomes (i.e., sexual perpetration, sexual victimization, overall perpetration [inclusive of SV, physical teen dating violence, bullying, and sexual harassment], and overall victimization [inclusive of SV, physical teen dating violence, bullying, and sexual harassment]; Aim 1a) and intermediary outcomes (i.e., social norms, denial of the issue of sexual violence, proactive bystander behavior, and four types of reactive bystander behavior [inappropriate touch, sexual violence, sharing sexual photos, spreading sexual rumors] Aim 1b). In addition, we examined variation in the short-term impact of a large “kick-off” leadership retreat for all primary and intermediary outcomes via testing for moderation by sex, sexual orientation, race, age, and denial (a prevention attitude related to bystander intervention and prevention engagement in other studies) (Aim 1c).

Our second aim was to examine the short-term impact of small leadership retreats on the same primary (Aim 2a) and secondary (Aim 2b) outcomes, as well as to test demographic moderators (Aim 2c). Finally, our third aim was to examine the long-term impact of the large leadership retreat on the same primary (Aim 3a) and secondary (Aim 3b) outcomes, as well as to test demographic moderators (Aim 3c). Of note, we could not examine long-term analyses for small leadership retreats because the end date of the project was before the one-year follow-up for small retreat participants. We hypothesized that program attendance would be associated with fewer rates of SV perpetration and victimization, higher perceptions of norms intolerant of SV, less bystander denial (i.e., low readiness to engage in situations of SV), more proactive behavior such as initiating conversations about prevention, and more bystander intervention in situations of SV. Moderation analyses were exploratory.

In this paper, we report short- and long-term outcomes associated with attendance at an initial “kick off” training retreat that included 125 youth and took place as an overnight event outside of the small city where youth lived (labeled large retreat), as well as short-term outcomes associated with attendance at subsequent smaller leadership retreats (both overnight and day retreats) that included anywhere from 28 to 49 youth (113 youth attended one or more small retreats). Although nominated youth were sent special invitations to participate, all youth in the district could attend as space allowed.

Method

Research Design and Setting

Data collection took place over 3 years in five waves: fall 2017 (W1), spring 2018 (W2), fall 2018 (W3), spring 2019 (W4), and fall 2019 (W5). The average number of days between W1 and subsequent waves was 180 to W2 (standard deviation [SD] = 6.8), 361 to W3 (SD = 15.5), 531 to W4 (SD = 19.9), and 733 to W5 (SD = 9.5). In the current study, we used waves based on timing of retreats. We used data from W3 (to measure the short-term impact of the large retreat) and W5 (to measure the short-term impact of the small retreats and long-term impact of the large retreat).

Participants

Participants were 2647 youth. At W1, they were in grades 7 to 10 and drawn from all public middle and high schools in the district (7 middle schools and 3 high schools), and the mean age at Wave 1 was 13.7 years (SD = 1.2 years). Including participants from all waves, the sample was 51.6% female participants (n = 1355) and 48.4% male participants (n = 1269).¹ Participants could identify as more than one race or ethnicity; the majority (76.6%, n = 1988) identified as White, 21.0% (n = 545) as Native American, 5.3% (n = 138) Black/African American, 3.2% (n = 82) Asian, and 2.4% (n = 61) Hawaiian/Pacific Islander. Moreover, 13.0% (n = 334) identified as Hispanic/Latino. Regarding sexual orientation, 89.2% (n = 2243) identified as heterosexual/straight and 10.8% (n = 272) identified as a sexual minority (e.g., bisexual, lesbian, gay). In regard to poverty, 37.2% (n = 901) of youth reported receiving free or reduced lunch, although this measure may be biased because some schools in the district had school lunches programs (e.g., everyone in the school received free lunch). We also found 12.9% (n = 223) did not have either a computer, internet access, or a clothes washer/dryer in their home, and that 40.1% (n = 692) thought their family felt somewhat or very stressed about money. Compared to data obtained from the school district in which the study took place, our sample was representative of the school district with the exception of Hispanic/Latino students being over-represented in our study.

Procedures

Written parental consent and student assent were required for youth to complete the study, consisting of five survey waves. Any youth in grades 7 to 10 (n = 4172) at the beginning of the fall 2017 semester were eligible and invited to enroll in the survey; the first survey occurred between October 2017 and December 2017. We used intensive recruitment procedures such that the consent forms were sent to parents in multiple ways (i.e., via their students from school, mailings, email), and we called and conducted home visits to households in which consent forms had not been returned. We also had multiple ways in which the consent forms could be returned (e.g., email, text, in person). At study initiation, of the 4172 eligible students, the majority (n = 3257; 78.0%) of youth returned the consent forms, and of those that returned the forms the majority (n = 2667; 81.8%) of guardians gave permission for their student to take the survey. Most students (n = 2232; 83.6%) with guardian permission took the survey. Figure 1 depicts eligibility and participation by wave. Past W1, we conducted ongoing study recruitment, such that we mailed consent forms to new students and followed up with calls and home visits, provided consent forms during in-school surveys, and offered consent forms at various community and school events. Figure 2 depicts the timing of programming activities and survey assessments.

Fig. 1.

Participation flow diagram. Reasons for not taking survey included absence, refusal, or inability to take the survey without assistance. We made an effort to contact students who left the district to take the survey

Fig. 2.

Timing of Youth Voices in Prevention programming. Note: The numbers in parentheses denote the total number of events, inclusive of retreats, action events, project campaigns (and youth campaign meetings in which these events were planned), and internship orientations and meetings that took place during that time period. This figure is also published in Waterman et al. (2021): https://journals.sagepub.com/doi/abs/10.1177/26320770211010817

The survey was administered on computers in school by trained research staff. All students had unique logins that were created in part so that students only with parental permission could access the survey. Students received a small incentive (e.g., fruit snack, pencil) and were entered to win one of twenty $100 gift cards which increased by $50 at each of the four subsequent surveys. Students who completed all surveys were entered into a drawing to receive a large prize equal to $1000 (e.g., tablets, pizza party). At each wave, students who missed the in-school survey (n =475–1289 across waves) were sent a letter in the mail requesting that they take the survey online; instructions were provided for how to take the survey online. Return rate of these out-of-school surveys ranged from 1.8 to 8.4%. Overall, retention from W1 ranged from 58.3 to 85.6% across waves. The highly transient nature of the community where data were collected was a large factor in participant attrition; the most common reason for not participating was absence from school, and many of those absences were from students who had left the district. If students who left the district were not considered in retention analysis (e.g., removed from the denominator), retention from W1 ranged from 87.9 to 98.7%. We identified inattentive responders (students taking the survey) using questions like “Do you have more than 10 kids?” By the last wave, 108 participants (4.1%) were identified as an inattentive responder and removed from analyses.

Participant Attrition Analysis

We conducted a series of chi-square and t-test analyses to understand patterns in attrition based on demographics and key study variables. First, we compared, among participants who completed W1, participants who completed each subsequent wave to participants who did not complete that subsequent wave. We completed this analysis for Waves 2, 3, 4, and 5. In general, younger students and White students were more likely to complete subsequent surveys. In general, male students, students of color, and youth who reported some forms of perpetration and victimization were less likely to take subsequent surveys.

Of the 125 students who attended the large retreat, 109 completed at least one wave, and 65 completed all five waves. Of the 113 who attended small retreats, 84 completed at least one wave, and 47 completed all five waves.

Measures

Demographics

A brief measure was included to assess sex, age, year in school, race, ethnicity, and sexual orientation.

Interpersonal Violence

We used several measures to assess for a wide range of interpersonal violence victimization and perpetration experiences during the past six months, all with response options 1 = yes or 0 = no. We used mirror items to assess for both victimization and perpetration experiences. Three items assessing SV were drawn from Cook-Craig et al. (2014) measure that assessed for sexual coercion (e.g., “You had sexual activities with someone because you either threatened to end your friendship or romantic relationship if they didn’t or because you pressured the other person by arguing or begging?”), physically-forced sex (e.g., “You had sexual activities with someone by threatening to use or using physical force [twisting their arm, holding them down, etc.]?”), and incapacitated sex (e.g., “You had sexual activities with someone because she or he was drunk or on drugs?”). Items adapted from the YRBS were used to assess physically forced sexual contact (e.g., “You forced someone to do sexual things that she or he did not want to do [count such things as kissing, touching, or physically forcing someone to have sexual intercourse]?”), sexual dating violence (e.g., “You forced someone you were dating or going out with to do sexual things that they did not want to do [count such things as kissing, touching, or physically forcing someone to have sexual intercourse]?”), physical dating violence (e.g., “You physically hurt someone you were dating or going out with on purpose [count such things as hitting, slamming into something, or injuring them with an object or weapon]?”), bullying on school property (e.g., “You bullied another person on school property?”), and electronic bullying (i.e., “You bullied another person electronically (count bullying through texting, Instagram, Facebook, or other social media)?”) (Centers for Disease Control & Prevention, 2017). We used two items from the AAUW (2001) to assess for homophobic bullying and sexual harassment (i.e., sexual comments [e.g., “You made sexual comments, jokes, gestures, or looks about/to a person?”], and sexual rumors [e.g., “You spread sexual rumors about a person?”]). Two items assessed homophobic bullying (e.g., “You said a person was gay or a lesbian, as an insult [as a put down or to make fun of someone?”]) (AAUW, 2001) and racial bullying (e.g., “You bullied another person or were mean to another person because of the other person’s race/ethnicity/skin color?”). A final item was added based on community input that measured aggravated sexting: “You threatened to share or actually shared a private picture of someone else the person did not want shared (including by texting, Instagram, Snapchat, Facebook or other way of sharing)?” Binary variables were created from the five items that measured sexual violence (1 = yes or 0 = no) to create the sexual violence perpetration/victimization outcomes (sexual harassment and aggravated sexting were distinct from sexual violence); binary variables were created from all 13 items (1 = yes or 0 = no) to create the any interpersonal violence perpetration/victimization outcomes.

Social Norms for Sexual Violence Prevention

Three items were used to assess youth’s perceptions of injunctive norms related to sexual violence prevention. These items were adapted for this study from earlier work with middle and high school samples (Edwards et al., 2017). The three items included the following: (a) “My friends think that it is important for adults to talk to students about healthy relationships,”(b) “My friends think that students should show that it is NOT okay to joke or make fun of people’s bodies,” and (c) “My friends think that students should talk about how to stop sexual assault (sexual assault is any sexual thing that happens when someone doesn’t want it to happen).” Students responded on a 4-point Likert scale ranging from 1 = strongly disagree to 4 = strongly agree, such that higher scores across all three items reflected more prosocial norm perception about prevention behaviors. Mean scores were calculated from all three items. Cronbach’s alpha for these items was 0.59–0.78 across W1, W3, and W5.

Bystander Denial

We used the Denial subscale of the Readiness to Help Scale (D-RHS; Banyard et al., 2014; Edwards et al., 2017) to assess the extent to which students rejected the role that they could play in preventing relationship abuse and sexual assault. The D-RHS consisted of four items (e.g., “There is not much need for me to think about relationship abuse and/or sexual assault among middle and high school students.”). However, one item “Doing something about relationship abuse and sexual assault is solely the job of the crisis center” was removed for this study given time constraints and youth reporting that the item was confusing during cognitive interviewing. Response options ranged from 1 = strongly disagree to 4 = strongly agree. Items are summed so that higher numbers are indicative of higher denial of responsibility in situations of relationship abuse and sexual assault. In the current study the Cronbach’s alpha was 0.59–0.69 across W1, W3, and W5.

Reactive Bystander Behavior

Items measured both opportunity to help as a gateway to questions about actual bystander intervention across four peer sexual violence situations based on previous work (Coker et al., 2011; Edwards et al., 2017). The items included (a) “Saw or heard a student grabbing or touching another student sexually (like on their butt or breasts),” (b) “Saw or heard about a student using physical force or alcohol or drugs to make/force another student to have sex,” (c) “Saw or heard about a student sending a naked photo of another student without that person’s permission,” and (d) “Saw or heard about a student spreading sexual rumors about another student.”

Participants first answered an opportunity item asking how many times they saw or heard the behavior. Participants who answered “0” did not get the follow up behavior questions. For each of the four reactive behavior items in which participants responded affirmatively with the opportunity to take action, we asked participants how they responded. Participants were presented with the following types of behavior and asked to select all of the things they did in response to witnessing the experience: (a) “Did nothing/ignored what was happening”; (b) “Laughed, took a video, or showed that you did not think what was happening was a big deal”; (c) “Tried to make the situation stop by using distraction, such as dropping something to make a noise; starting a random conversation”; (d) “Get help from another teen, parent, and/or adult”; (e) “Said something or tried to stop the person doing the hurtful behavior”; and (f) “Said something or tried to help or support the person who was being hurt.” For each of the response behaviors, students’ responses were recoded to either 0 = no or 1 = yes. Each of the four situations is scored and used as a separate behavior outcome. Scores were created to represent the degree to which, in each situation, participants acted positively. A point was subtracted from the score if participants acted negatively (behaviors a and b) and added to the score if participants acted positively (behaviors c, d, e, and f). Thus, final scores for each situation ranged from − 2 to 4.

Proactive Bystander Behavior

Two questions asked students about taking proactive steps to prevent sexual assault from happening in the first place (referred to as proactive bystander behavior); one that assessed how much students “talked during the past six months with their friends or parents, teacher, minister, elder, etc. about things you all could do that might help stop sexual assault” and a second question that inquired about the “use of social media (like Facebook, Twitter, etc.) or texting to show that sexual assault is not okay.” The questions used the following response options for each of the items: 0 = 0 times, 1 = 1–2 times, 3 = 3–5 times, 6 = 6–9 times, or 10 = 10 or more times. We used a mean of the two items; the Cronbach’s alpha for these two items was 0.65–0.71 across W1, W3, and W5.

Data Analyses

All analyses were estimated in Stata 15. In all analyses, we used multiple imputation (MI; 20 imputed datasets) to address attrition and other missing data (Lang & Little, 2018). To address Aim 1a, we conducted a series of logistic regression analyses, one for each of the dichotomous W3 outcomes (i.e., sexual perpetration, sexual victimization, overall perpetration, overall victimization; dependent variables). These analyses examined the short-term effects (approximately 3 months post) of the large retreat. The independent variable (predictor) in each model was large retreat attendance (0 = did not attend; 1 = attended). Covariates were sex (1 = male), ethnicity (1 = Hispanic/Latinx), race (1 = White), sexual orientation (1 = sexual minority), age (higher = older), school (dummy coded with the largest school as the comparison group), and the W1 outcome score (as applicable to each model). To address Aim 1b, we conducted a series of linear regression analyses. Each outcome (i.e., social norms, denial, proactive bystander behavior, reactive bystander behavior—touch, reactive bystander behavior—sexual violence, reactive bystander behavior—share, and reactive bystander behavior—rumors) was the dependent variable in a separate model. The independent variables and covariates were identical to aim 1a analyses.

To address Aim 1c, we added interaction terms (i.e., moderator by independent variable) to the analyses described above. We tested moderation for three dichotomous (i.e., sex, sexual orientation, and race) and two continuous (i.e., age [higher = older] and denial [higher = more denial]) variables for each of the four main outcomes (i.e., sexual perpetration, sexual victimization, any perpetration, any victimization) and each of the seven intermediary outcomes (i.e., social norms, denial, proactive bystander behavior, reactive bystander behavior—touch, reactive bystander behavior—sexual violence, reactive bystander behavior—share, and reactive bystander behavior—rumors). We estimated separate models for each moderator. To follow-up significant interactions, we probed to understand the directionality of the interaction, holding retreat attendance constant with attenders as the reference group (Aiken et al., 1991). For categorical moderators, we switched the reference group in the model to obtain estimates for the different levels of the moderator (e.g., females and males). For continuous moderators, we re-centered the moderator to obtain estimates + 1 SD above the mean or − 1 SD below the mean.

Analyses for Aims 2a-2c and 3a-3c followed the same pattern as the prior description except the outcomes were from W5 (which allowed for a long-term analysis of large retreat effects and short-term analysis of small retreat impact). However, because large and small retreat attendance (e.g., youth could attend both retreats) may co-vary and, thus, need to be included in the same models, models for Aims 2a-2c and 3a-3c were combined. That is, for Aims 2a and 3a, logistic regression models examined the effects of both large retreat attendance (0 = did not attend; 1 = attended) and small retreat attendance (0 = did not attend; 1 = attended) on each primary outcome included in the model. For Aims 2b and 3b, linear regression models examined the effects of large and small retreat attendance on each intermediary outcome. These analyses examined the long-term effects (approximately 1 year post) of the large retreat and short-term effects (six months post) of the small retreats. For Aims 2c and 3c, we again tested each moderator separately, in the model with both retreats. For example, the first model for Aims 2c and 3c contained the Sex x Large Retreat Attendance interaction and the Sex x Small Retreat Attendance interaction as related to sexual perpetration.

To summarize the use of waves, we used W1 as a baseline. W3 was the short-term follow-up for the large retreat since W3 occurred shortly after the large retreat. The small retreats occurred after W3 and W4, so W5 was used as a short-term follow-up for the small retreats (as well as a long-term follow-up for the large retreat. It was not feasible to analyze the small retreats separately given the small amount of youth that attended each one (attendance ranged from 28 to 49 youth).

In regard to effect sizes, for significant main effects for dichotomous variables (e.g., perpetration and victimization), we present the odds ratio as a measure of effect size. For significant main effects for continuous variables, we calculated unadjusted Cohen’s d using the descriptive statistics from the imputed datasets.

Finally, given this design was not a randomized controlled trial, selection effects likely played a role in which students attended. We explore selection and attendance in-depth in another publication (Banyard et al., in press). In all analysis, we control for previous behaviors and experience (that is, the outcome at W1).

Results

Aim 1a (Short‑term Effects of Large Retreat on Primary Outcomes)

Large retreat attendees were more likely to report subsequent overall victimization (odds ratio [OR] = 1.87; p = 0.02) than non-attendees. Because overall victimization comprised several types of IV victimization, we conducted post hoc analyses to understand whether a certain type of violence victimization was driving this finding. We found that when run as separate outcomes, only homophobic bullying (OR = 1.71, p = 0.05) was significant, suggesting that this type of victimization may have driven the findings. We did not find evidence of significant associations between large retreat attendance with subsequent sexual victimization or overall perpetration, and only with overall victimization (Table 1).

Table 1.

Aim 1a: Short-term effect of Youth Voices in Prevention large retreat on primary sexual violence outcomes (N = 2538)

Variable	OR	Std. err	p	95% CI		OR	Std. err	p	95% CI
	Sexual perpetration					Sexual victimization
Outcome at T1	–	–	–	–	–	7.14	2.20	.00	3.85	13.26
Male	–	–	–	–	–	0.27	0.07	.00	0.17	0.44
Hispanic/Latinx	–	–	–	–	–	1.62	0.46	.09	0.92	2.85
Sexual minority	–	–	–	–	–	1.32	0.34	.27	0.80	2.18
White	–	–	–	–	–	2.37	0.82	.02	1.18	4.75
Age	–	–	–	–	–	1.25	0.20	.17	0.91	1.72
Large retreat	–	–	–	–	–	1.37	0.52	.41	0.65	2.90
	Overall perpetration					Overall victimization
Outcome at T1	6.59	0.95	.00	4.94	8.79	8.02	0.99	.00	6.29	10.23
Male	1.74	0.23	.00	1.34	2.26	0.64	0.08	.00	0.51	0.81
Hispanic/Latinx	1.40	0.28	.10	0.94	2.09	1.36	0.24	.09	0.95	1.93
Sexual minority	1.46	0.29	.06	0.99	2.15	1.72	0.37	.01	1.12	2.64
White	1.86	0.30	.00	1.35	2.55	1.17	0.20	.34	0.84	1.64
Age	0.92	0.09	.39	0.76	1.12	1.04	0.10	.66	0.86	1.27
Large retreat	1.21	0.32	.46	0.72	2.04	1.87	0.49	.02	1.11	3.14

All analysis included school as a covariate; to preserve space, we did not include these estimates. Data were collected from 2017 to 2019. We could not examine the effects on sexual perpetration at W3 due to perfect prediction stemming from a low number of students reporting sexual perpetration

OR odds ratio, CI confidence interval

Aim 1b (Short‑term Effects of Large Retreat on Intermediary Outcomes)

Large retreat attendees reported significantly less subsequent denial (b = − 0.32; d = 0.33; p = 0.00) and more subsequent proactive bystander behavior (b = 0.91; d = 0.26; p = 0.00) than non-attendees. Large retreat attendees reported more subsequent reactive bystander behavior to address knowledge of unwanted touching against a peer (b = 0.47; d = 0.23; p = 0.00) than non-attendees. Large retreat attendees reported more subsequent reactive bystander behavior in response to knowing about unwanted photo sharing (b = 0.14; d = 0.10; p = 0.04) and more subsequent reactive bystander behavior when aware of peers spreading sexual rumors (b = 0.37; d = 0.23; p = 0.00) than non-attendees. We did not find evidence of associations between large retreat attendance with subsequent social norms or reactive bystander behavior to address sexual violence (Table 2).

Table 2.

Aim 1b: Short-term effect of Youth Voices in Prevention large retreat on intermediary sexual violence-related outcomes (N = 2538)

Variable	b	Std. err	p	95% CI		b	Std. err	p	95% CI
	Social norms					Denial
Outcome at T1	0.28	0.03	.00	0.23	0.33	0.26	0.03	.00	0.20	0.31
Male	− 0.29	0.03	.00	− 0.35	− 0.23	0.22	0.03	.00	0.16	0.29
Hispanic/Latinx	0.01	0.05	.83	− 0.09	0.11	− 0.02	0.04	.62	− 0.10	0.06
Sexual minority	0.06	0.05	.22	− 0.04	0.15	− 0.06	0.04	.20	− 0.14	0.03
White	− 0.01	0.04	.85	− 0.08	0.07	0.03	0.04	.37	− 0.04	0.11
Age	0.03	0.03	.21	− 0.02	0.08	0.00	0.02	.96	− 0.05	0.05
Large retreat	− 0.08	0.07	.24	− 0.21	0.05	− 0.32	0.06	.00	− 0.45	− 0.20
	Proactive bystander behavior					Reactive bystander behavior—touch
Outcome at T1	0.26	0.02	.00	0.21	0.31	0.19	0.02	.00	0.15	0.23
Male	− 0.42	0.08	.00	− 0.58	− 0.25	− 0.21	0.04	.00	− 0.29	− 0.13
Hispanic/Latinx	0.15	0.12	.24	− 0.10	0.39	− 0.03	0.06	.64	− 0.14	0.09
Sexual minority	0.24	0.14	.08	− 0.03	0.51	0.00	0.06	.94	− 0.12	0.13
White	− 0.17	0.09	.05	− 0.35	0.00	0.03	0.05	.53	− 0.07	0.13
Age	0.04	0.05	.40	− 0.06	0.15	− 0.02	0.03	.41	− 0.08	0.03
Large retreat	0.91	0.16	.00	0.60	1.23	0.47	0.10	.00	0.28	0.66
	Reactive bystander behavior—sexual					Reactive bystander behavior—share violence
Outcome at T1	0.08	0.02	.00	0.04	0.12	0.09	0.02	.00	0.05	0.12
Male	− 0.08	0.03	.00	− 0.13	− 0.03	− 0.12	0.03	.00	− 0.17	− 0.07
Hispanic/Latinx	0.03	0.04	.46	− 0.05	0.11	− 0.01	0.04	.72	− 0.10	0.07
Sexual minority	0.01	0.04	.82	− 0.07	0.09	− 0.04	0.04	.31	− 0.13	0.04
White	0.01	0.03	.77	− 0.05	0.07	0.02	0.03	.64	− 0.05	0.08
Age	0.00	0.02	.99	− 0.04	0.04	0.01	0.02	.53	− 0.03	0.05
Large retreat	0.06	0.06	.37	− 0.07	0.19	0.14	0.07	.04	0.00	0.27
	Reactive bystander behavior—rumors
Outcome at T1	0.15	0.02	.00	0.11	0.18
Male	− 0.21	0.04	.00	− 0.28	− 0.14
Hispanic/Latinx	0.00	0.05	.99	− 0.10	0.11
Sexual minority	− 0.07	0.06	.20	− 0.19	0.04
White	0.02	0.04	.59	− 0.06	0.11
Age	0.00	0.03	.91	− 0.05	0.05
Large retreat	0.37	0.09	.00	0.20	0.54

All analysis included school as a covariate; to preserve space, we did not include those estimates. Data were collected from 2017 to 2019

OR odds ratio, CI confidence interval

Aim 1c (Moderators of Short‑term Large Retreat Effects)

There was an interaction with age for two outcomes, overall victimization and reactive bystander behavior to address unwanted photo sharing. Regarding overall victimization, there was an interaction between age and large retreat attendance. In particular, there was an overall positive effect of retreat for younger students (i.e., students at one standard deviation below mean age; b = 1.30, p = 0.00), but not for older students (i.e., students at one standard deviation above mean age; b = − 0.12, p = 0.77). Regarding reactive bystander behavior to address unwanted photo sharing, there was an overall positive effect of retreat for older students (b = 0.34, p = 0.00) but not for younger students (b = − 0.03, p = 0.79). We did not find evidence of significant moderation by sex, sexual orientation, race, and denial on the effects of the large retreat on main or intermediary outcomes (see Supporting Information Appendix B).

Aim 2a (Short‑term Effects of Small Retreat on Primary Outcomes)

Small retreat attendees were less likely than non-attendees to report subsequent overall perpetration (OR = 0.30; p = 0.01). Because overall perpetration was comprised of several types of IV perpetration, we conducted post hoc analyses to understand whether a certain type of violence perpetration was driving the findings. We found that when run as separate outcomes, small retreat attendance was associated with lower odds of sexual harassment (OR = 0.15, p = 0.00), suggesting that this type of perpetration may have driven the findings. We did not find evidence of significant associations between small retreat attendance with subsequent sexual perpetration, sexual victimization and overall victimization (Table 3).

Table 3.

Aims 2a and 3a: Short-term effect of Youth Voices in Prevention small retreat and long-term effect of large retreat on primary sexual violence outcomes (N = 2538)

Variable	OR	Std. err	p	95% CI		OR	Std. err	p	95% CI
	Sexual perpetration					Sexual victimization
Outcome at T1	4.84	2.97	.01	1.39	16.80	3.32	1.21	.00	1.59	6.93
Male	0.51	0.15	.02	0.29	0.89	0.27	0.07	.00	0.16	0.44
Hispanic/Latinx	1.10	0.46	.83	0.47	2.53	1.77	0.48	.04	1.03	3.04
Sexual minority	0.84	0.41	.73	0.31	2.26	1.78	0.47	.03	1.04	3.02
White	1.29	0.50	.51	0.59	2.81	1.72	0.61	.13	0.85	3.50
Age	1.09	0.25	.72	0.68	1.74	0.89	0.14	.46	0.65	1.22
Large retreat	1.30	0.66	.60	0.48	3.53	2.52	0.80	.00	1.34	4.71
Small retreats	0.75	0.60	.72	0.16	3.60	1.20	0.62	.73	0.43	3.36
	Overall perpetration					Overall victimization
Outcome at T1	3.88	0.49	.00	3.02	4.99	3.86	0.49	.00	3.00	4.97
Male	2.33	0.31	.00	1.79	3.04	0.87	0.10	.24	0.69	1.10
Hispanic/Latinx	0.88	0.19	.55	0.57	1.35	1.08	0.21	.71	0.72	1.60
Sexual minority	2.52	0.55	.00	1.63	3.89	2.22	0.46	.00	1.47	3.35
White	1.58	0.28	.01	1.10	2.26	1.39	0.24	.07	0.97	1.98
Age	0.93	0.09	.45	0.76	1.13	1.00	0.10	.99	0.82	1.23
Large retreat	2.48	0.70	.00	1.42	4.32	1.40	0.36	.19	0.84	2.34
Small retreats	0.30	0.14	.01	0.12	0.75	0.89	0.26	.70	0.51	1.58

All analysis included school as a covariate; to preserve space, we did not include these estimates. Data were collected from 2017 to 2019

OR odds ratio, CI confidence interval

Aim 2b (Short‑term Effects of Small Retreat on Secondary Outcomes)

Small retreat attendees reported less subsequent denial (b = − 0.30; d = 0.33; p = 0.00) and more subsequent proactive bystander behavior (b = 0.56; d = 0.20; p = 0.00) than non-attendees. Small retreat attendees reported more subsequent reactive bystander behavior in relation to sexual violence (b = 0.27; d = 0.18; p = 0.00) than non-attendees. We did not find evidence of significant associations between small retreat attendance with subsequent social norms, or reactive bystander behaviors to address unwanted touching, photo sharing, or sharing of sexual rumors (Table 4).

Table 4.

Aims 2b and 3b: Short-term effect of Youth Voices in Prevention small retreat and long-term effect of large retreat on intermediary sexual violence-related outcomes (N = 2538)

Variable	b	Std. err	p	95% CI		b	Std. err	p	95% CI
	Social norms					Denial
Outcome at T1	0.20	0.03	.00	0.14	0.25	0.27	0.03	.00	0.22	0.33
Male	− 0.23	0.03	.00	− 0.30	− 0.17	0.25	0.03	.00	0.19	0.31
Hispanic/Latinx	− 0.05	0.06	.40	− 0.17	0.07	0.01	0.05	.75	− 0.08	0.10
Sexual minority	0.02	0.06	.75	− 0.10	0.14	− 0.07	0.05	.11	− 0.17	0.02
White	− 0.06	0.06	.27	− 0.18	0.05	0.06	0.04	.16	− 0.02	0.13
Age	− 0.02	0.03	.54	− 0.07	0.04	− 0.02	0.02	.33	− 0.06	0.02
Large retreat	− 0.11	0.07	.12	− 0.25	0.03	0.03	0.07	.61	− 0.10	0.17
Small retreats	0.01	0.09	.93	− 0.16	0.18	− 0.30	0.07	.00	− 0.44	− 0.17
	Proactive bystander behavior					Reactive Bystander behavior—touch
Outcome at T1	0.18	0.03	.00	0.11	0.25	0.11	0.02	.00	0.07	0.14
Male	− 0.19	0.07	.01	− 0.33	− 0.04	− 0.16	0.03	.00	− 0.23	− 0.09
Hispanic/Latinx	− 0.26	0.12	.03	− 0.50	− 0.03	0.00	0.05	.93	− 0.10	0.10
Sexual minority	0.20	0.12	.12	− 0.05	0.45	− 0.01	0.06	.89	− 0.12	0.10
White	− 0.16	0.11	.14	− 0.38	0.06	0.08	0.04	.08	− 0.01	0.16
Age	0.10	0.05	.03	0.01	0.20	0.02	0.03	.37	− 0.03	0.08
Large retreat	− 0.11	0.15	.48	− 0.40	0.19	0.15	0.08	.08	− 0.02	0.32
Small retreats	0.56	0.17	.00	0.23	0.89	0.18	0.10	.07	− 0.01	0.36
	Reactive bystander behavior—sexual					Reactive bystander behavior—share violence
Outcome at T1	0.05	0.02	.00	0.02	0.08	0.05	0.01	.00	0.02	0.08
Male	− 0.07	0.02	.00	− 0.11	− 0.03	− 0.03	0.02	.20	− 0.08	0.02
Hispanic/Latinx	0.07	0.03	.02	0.01	0.13	0.03	0.04	.33	− 0.03	0.10
Sexual minority	− 0.08	0.03	.02	− 0.14	− 0.01	− 0.02	0.04	.62	− 0.09	0.06
White	0.03	0.03	.29	− 0.02	0.08	0.00	0.03	.98	− 0.06	0.06
Age	0.02	0.02	.29	− 0.01	0.05	− 0.01	0.02	.64	− 0.04	0.03
Large retreat	0.07	0.05	.18	− 0.03	0.17	0.16	0.06	.01	0.04	0.27
Small retreats	0.27	0.06	.00	0.15	0.38	0.13	0.07	.06	0.00	0.26
	Reactive bystander behavior—rumors
Outcome at T1	0.09	0.02	.00	0.06	0.11
Male	− 0.14	0.03	.00	− 0.20	− 0.08
Hispanic/Latinx	0.04	0.05	.33	− 0.04	0.13
Sexual minority	− 0.11	0.05	.03	− 0.20	− 0.01
White	0.02	0.04	.59	− 0.05	0.09
Age	0.00	0.02	.82	− 0.04	0.05
Large retreat	0.14	0.08	.06	− 0.01	0.29
Small retreats	0.11	0.09	.21	− 0.06	0.27

All analysis included school as a covariate; to preserve space, we did not include those estimates. Data were collected from 2017 to 2019

OR odds ratio, CI confidence interval

Aim 2c² (Moderators of Short‑term Effects of Small Retreat)

We found significant moderation of the effects of small retreat on reactive bystander behavior to address sexual violence and reactive bystander behavior related to unwanted photo sharing. We did not find evidence of significant moderation of the effects of small retreat on main outcomes or on the intermediary social norms, denial, prosocial bystander behavior, or the other two forms of reactive bystander behavior (see Supporting Information Appendix B).

For bystander behavior to respond to sexual violence, there was an overall positive effect of small retreat attendance for sexual minority students (b = 0.37, p = 0.00), but not for heterosexual students (b = − 0.03, p = 0.82). For bystander behavior reacting to unwanted photo sharing, there was an overall positive effect of small retreat attendance for female students (b = 0.34, p = 0.00), but not for male students (b = 0.01, p = 0.88). Also, for bystander behavior responding to unwanted photo sharing, the interaction with race was significant for small retreat attendance. There was an overall positive effect of small retreat attendance for non-White students (b = 0.23, p = 0.01), but not for White students (b = − 0.05, p = 0.63). Finally, the interaction between age and small retreat attendance was significant. There was an overall positive effect of small retreat for younger students (b = 0.22, p = 0.01), but not for older students (b = 0.08, p = 0.40).

Aim 3a (Long‑term Effects of Large Retreat on Primary Outcomes)

Large retreat attendees were more likely than non-attendees to report more subsequent sexual victimization at the longer follow-up (OR = 2.52; p = 0.00). Large retreat attendees were more likely than non-attendees to report subsequent overall perpetration (OR = 2.48; p = 0.00). Because overall perpetration was comprised of several types of IV perpetration, we conducted post hoc analyses to understand whether a certain type of violence perpetration was driving the findings. We found that when run as separate outcomes, the large retreat attendance was associated with higher odds of sexual harassment (OR = 2.38, p = 0.01), suggesting that this type of perpetration may have driven the findings. We did not find evidence of significant associations between large attendance with subsequent sexual perpetration and overall victimization (Table 3).

Aim 3b (Long‑term Effects of Large Retreat on Secondary Outcomes)

Large retreat attendees reported more subsequent reactive bystander behavior to address unwanted photo sharing (b=0.16; d=0.12; p=0.01) than non-attendees. We did not find evidence of significant associations between large retreat attendance with subsequent social norms, denial, proactive bystander behavior, reactive bystander behavior—touch, reactive bystander behavior—sexual violence, and reactive bystander behavior—rumors (Table 4).

Aim 3c³ (Moderators of Long‑term Large Retreat Effects)

We found that demographics were significant moderators of long-term retreat outcomes of all four forms of reactive bystander behavior. We did not find evidence of significant moderation for sex, sexual orientation, race, age, and denial on the effects of retreats on main outcomes or on the intermediary social norms, denial, or prosocial bystander behavior (see Supporting Information Appendix C).

For reactive bystander behavior to address unwanted touch, there was an interaction between age and large retreat attendance. There was an overall positive effect of large retreat for older students (b = 0.34, p = 0.00) but not for younger students (b = −0.03, p = 0.79). For reactive bystander behavior in relation to sexual violence, there was an interaction between sex and large retreat attendance. There was an overall positive effect of large retreat for both female (b = 0.44, p = 0.00) and male students (b = 0.17, p = 0.02), but the effect was stronger for female students than male students. There was also an interaction with sexual orientation for large retreat for reactive bystander behavior toward sexual violence. There was an overall positive effect of large retreat attendance for heterosexual students (b = 0.45, p = 0.01) but not for sexual minority students (b = 0.02, p = 0.65). Finally, also for reactive bystander behavior in relation to sexual violence, there was an interaction between denial and large retreat attendance. There was a positive effect of large retreat attendance for students who were lower on denial (i.e., students at one standard deviation below mean denial; b = 0.15, p = 0.02), but not for students who were high on denial (i.e., students at one standard deviation above mean denial; b = − 0.09, p = 0.32).

For reactive bystander behavior in relation to unwanted photo share, there was an interaction with sex for large retreat attendance. There was an overall positive effect of large retreat attendance for male students (b = 0.45, p = 0.01) but not for female students was (b = − 0.03, p = 0.79). The interaction with race was significant for large retreat attendance for reactive bystander behavior—share. There was an overall positive effect of large retreat attendance for White students (b = 0.42, p = 0.00) but not for non-White students was (b = 0.09, p = 0.16). For reactive bystander behavior—rumors, there was an interaction between race and large retreat attendance. There was a positive effect of large retreat attendance for White students (b = 0.49, p = 0.01) and not for non-White students (b = 0.06, p = 0.46). The interaction between age and large retreat attendance was also significant for reactive bystander behavior—rumors. There was an overall positive effect of large retreat for younger students (b = 0.36, p = 0.00), but not for older students (b = − 0.11, p = 0.39).

Discussion

To our knowledge, this is the first study to examine the impacts of a youth-led SV prevention initiative, specifically focusing on both short- and long-term outcomes for youth associated with attending prevention leadership retreats. Overall, there were some promising findings associated with retreat attendance, especially at the smaller retreats, including reductions in later sexual harassment behaviors, reduced denial of bystander responsibility, and increased proactive bystander behavior following retreat attendance. Although the effect sizes were small to medium, small effects are meaningful for perpetration and bystander behaviors. For the “kick off” retreat that was relatively larger in size, the findings were mixed. Although there were some positive outcomes demonstrated (e.g., increases in some forms of bystander behavior), there were some concerning and potentially iatrogenic effects found as well. For example, results suggested that youth who participated in the large “kick off” retreat self-reported subsequently higher rates of both perpetration and victimization compared to youth who did not participate in this retreat.

Although these findings may be spurious, there are alternative explanations to consider. First, compared to subsequent smaller leadership retreats, the first “kick off” retreat included less skill-based prevention content, which was challenging to deliver with so many youth attendees. Second, there were problematic student behaviors (e.g., homophobic bullying) observed at the retreat by adult facilitators. Youth retreat participants were pre-identified as leaders and came from across the city; thus, youth were less likely to know other youth attending, which may have led to more problematic behaviors as well as less efficacy among adult facilitators in addressing these problematic behaviors. With additional adult facilitator training and smaller groups of youth, these behavioral challenges were resolved. Indeed, anecdotal feedback suggests that smaller leadership retreats allowed for faster engagement with materials, fewer problematic behavior disruptions, and more opportunities for all youth to participate and lead.

The finding that victimization was higher among retreat attenders, compared to non-attendees, is likely explained by the fact that at both the large and small retreats, there were a number of youth, predominantly girls, who disclosed experiences of interpersonal violence. Although some of these disclosures happened in private with an adult, some of the disclosures occurred during small and large group discussions. As disclosure became normative, others were likely to reflect on that experience and share their own disclosures, similar to how #MeToo has shown increases in SV disclosures across the country (Palmer et al., 2021). Indeed, as mentioned before, the content included bystander training to help youth better recognize SV and leadership exercises to encourage youth to use their voices to talk about prevention. Such activities may have allowed some youth to break silence and report victimization experiences.

Smaller retreats led to more favorable outcomes, such as increases in bystander action, decreases in bystander denial, and reductions in subsequent perpetration. In addition to the ideas mentioned above, the smaller retreats had more prevention content. Further, small retreats provided youth with more intimate opportunities to engage in discussion with one another and adults helping to facilitate the prevention content. Also, unlike the large retreat that was solely adult facilitated, in subsequent smaller retreats, youth took more leadership roles in facilitating the content alongside adult mentors, which may have increased the believability and salience of the material.

Some interesting moderation effects emerged in the current study. Although results were not totally consistent, retreats were more often effective for younger versus older youth, which is consistent with other research (Waterman et al., 2021), suggesting that older youth likely need additional prevention components (e.g., alcohol use prevention) not addressed in the Youth VIP program. Smaller retreats also worked better for sexual minority youth than heterosexual youth which is an interesting finding in light of research documenting that interpersonal violence prevention programs for youth are more effective for heterosexual youth compared to sexual minority youth (Coker et al., 2020; Waterman et al., 2021). Smaller retreats included break-out sessions, one of which was for sexual and gender minority youth to come together to discuss how topics of interpersonal violence related specifically to their lives. These discussions were facilitated by sexual and gender minority adults who answered questions about coming out, dating, and how to help other sexual and gender minority youth in abusive relationships. Although speculative, these discussions, alongside scenarios that were inclusive of sexual and gender minority youths’ experiences, may have helped to reduce proximal minority stress, a robust predictor of SV among LGBTQ + youth (Decker et al., 2018; Edwards & Sylaska, 2013; Scheer, 2020).

Smaller retreats also worked better for racial/ethnic minority youth (which were largely Native American youth) than for White youth. This is likely because smaller retreats had a strong and intentional focus on connecting Lakota virtues to prevention content as well as engaging in Lakota traditions (e.g., smudging, traditional prayer). Research suggests that connection to culture is a protective factor against experiencing among violence among Native American youth (Edwards et al., 2021).

Despite the important information gleaned from the current study, there are a few limitations. First, we did not use an experimental design to examine the impact of retreat attendance on outcomes, thus limiting the causal inferences of relationships documented herein. Selection effects certainly played a role in student attendance (masked for review). Second, the number of retreat attendees who completed surveys, compared to non-attendees who completed surveys, was small. Third, effect sizes were small to medium, consistent with previous research indicated the challenges of changing interpersonal violence perpetration and victimization; we did not adjust the p-values for the number of tests. Also, despite a large presence of Native American youth in our sample, there were few youth representing other racial/ethnic minority groups and the majority of the sample identified as white, which limits the generalizability of the findings of this study. There was also likely selection bias in who participated in both the surveys and retreats; for example, youth who reported some forms of perpetration and victimization were less likely to take subsequent surveys. Thus, future research is needed to replicate these findings using larger, more diverse samples of youth with more rigorous outcome evaluation methodologies. Further, while the Youth VIP program was designed to incorporate many foundational best practices for SV prevention, length of the evaluation instrument precluded measuring all of these aspects. Surveys were done in schools and had to be short to accommodate class periods and to work against survey fatigue. Thus, we were unable to measure key variables like aspects of social-emotional learning or connection to culture that may also have been affected by programming. These will be important outcomes to measure in future studies. Finally, recruitment and outcomes were based on surveys administered in schools. Thus, although this was an outside-of-school program, there were still many links to the school context. Thus, the ability of the current study to contribute knowledge about programs in which recruitment also happens outside of school is limited.

The current study has several important implications for prevention. First, the findings of this study, especially regarding the smaller retreats, are promising, and suggest that youth-led prevention programs that use smaller retreats may be an important component of comprehensive SV prevention. Second, these data underscore the feasibility but also the need for prevention programs to be inclusive of youth occupying marginalized social identities including LGBTQ + youth as well as racial/ethnic minority youth to enhance programming impact. The current study also reinforces the utility of prevention efforts with youth that extend beyond the walls of schools. Although in need of refinement to enhance program impacts, Youth VIP aligns with the best available evidence in SV prevention (Basile et al., 2016b; David-Ferdon et al., 2016) and demonstrated that this type of approach (especially smaller-based retreats) may be one critical component to comprehensive SV prevention among youth.