Abstract
This study evaluated the effectiveness of a 6-session (12-hour) empowerment self-defense classroom delivered curriculum (i.e., IMpower) among American Indian girls. Girls (N = 74) in one middle school and two high schools on an Indian Reservation in the Great Plains region of the United States received the intervention and completed a pre-test and a post-test six months following the final program session. The surveys administered assessed hypothesized intermediary (i.e., efficacy to resist a sexual assault, self-defense knowledge), primary (i.e., sexual violence victimization), and secondary (i.e., physical dating violence, sexual harassment) outcomes. Native American girls (N = 181) in five middle schools and three high schools in a nearby city where there was no sexual assault prevention occurring completed surveys assessing sexual violence, physical dating violence, and sexual harassment victimization approximately six months apart, thus serving as a comparison to girls in the treatment condition on primary and secondary outcomes. Girls exposed to the IMpower program reported significant increases over time in efficacy to resist a sexual assault and knowledge of effective resistance strategies. Furthermore, propensity score analyses suggested that girls who received the IMpower program reported significantly fewer types of sexual assault and sexual harassment at follow-up compared to girls in the control condition. However, no effect was found for physical dating violence. These data suggest that empowerment self-defense is a promising approach in preventing sexual assault and sexual harassment among American Indian girls.
Effectiveness of a Sexual Assault Self-defense Program for American Indian Girls
Sexual assault, which ranges from unwanted touching to completed rape, is a pernicious issue in the United States, and disproportionally impacts American Indian youth (Robin et al., 1998; Rosay, 2016; Rutman et al., 2008; Warne et al., 2017). For example, in a nationally representative sample of urban American Indian high school students, researchers documented that 16.4% of American Indian youth reported being physically forced to have sex during the past year (compared to 6.6% of White youth; Rutman et al., 2008). Research suggests that sexual assault is related to a host of deleterious outcomes in American Indian youth, including (but not limited to) depression, suicidal ideation, and binge drinking (Edwards et al., 2021). Thus, the prevention of sexual assault among American Indian girls is of critical public health importance.
Sexual assault also co-occurs at high rates with sexual harassment (e.g., unwanted sexual comments) and physical dating violence (e.g., pushing, slapping; Hamby & Grych, 2013; Sessarego et al., 2019). These forms of violence share overlapping risk and protective factors (Hamby & Grych, 2013; Wilkins et al., 2014) and are therefore often concurrently targeted in programs that seek to prevent sexual assault (Coker et al., 2017; Edwards et al., 2019). Even if programs are primarily targeting the prevention of sexual assault, it is important to measure other outcomes, such as sexual harassment and physical dating violence, as secondary outcomes.
To date, research suggests that a handful of prevention initiatives lead to reductions in rates of sexual assault (DeGue et al., 2014). The Centers for Disease Control and Prevention has identified several strategies that research suggests are important in preventing sexual assault (Basile et al., 2016). For example, there are several studies that suggest that bystander intervention training reduces sexual assault, sexual harassment, and dating violence among high school students (Coker et al., 2019; Edwards et al., 2019). Moreover, research suggests that engaging boys and men in prevention efforts via encouraging healthy masculinity and/or correcting misperceptions of social norms are effective in preventing sexual assault (Gidycz et al., 2011; Miller et al., 2012). Research also suggests that teaching adolescents healthy relationship skills (e.g., healthy ways to manage conflict) can help to reduce sexual assault and dating violence (Foshee et al., 1998; Foshee et al., 2004; Foshee et al., 2000; Niolon et al., 2019).
Additionally, empowerment self-defense training in which girls and women are taught to recognize risk and respond using verbal and/or physical resistance strategies to thwart a sexual assault is effective at reducing sexual assault (Orchowski et al., 2018). Although perpetrators are always to blame for sexual assault, it is important that girls and women be equipped with tools to effectively resist a sexual assault (Orchowski et al., 2018). Indeed, research suggests that when girls and women respond assertively to a sexual assault (e.g., scream, run away), they are more likely to thwart a sexual assault than if they use passive resistance strategies (e.g., cry, freeze; Gidycz et al., 2006; Gidycz et al., 2008; Orchowski et al., 2018; Turchik et al., 2007). Several rigorous research studies have found that empowerment self-defense training leads to reduced likelihood to experience a sexual assault (Gidycz et al., 2001; Gidycz et al., 2006; Orchowski et al., 2018; Senn et al., 2015). Moreover, despite some criticisms that self-defense training for girls is victim blaming, research suggests that girls and women who participate in empowerment self-defense training, compared to girls and women who do not, report less self-blame following a sexual assault (Gidycz et al., 2015; Mouilso et al., 2011; Orchowski et al., 2018; Senn et al., 2008).
Most research to date has evaluated empowerment self-defense training in samples of college women. An exception to this is the IMpower program, implemented by No Means No Worldwide. IMpower is a 6-session (12-hour) empowerment self-defense classroom delivered curriculum for girls ages 10–20 (refer to Table 1 for an overview of the six classes). The IMpower program teaches girls to identify risk, say “no” and talk their way out of trouble, and if “no” is not respected, girls learn physical skills to resist a sexual assault. Additionally, girls engage in chants and other activities to reinforce the messages that they are worth defending. To date, the IMpower program—which originated in Nairobi, Kenya—has been implemented and evaluated (via cluster randomized control trials) in high-risk environments (e.g., informal settlements) in Kenya and Malawi. Results from these trials suggest that the IMpower program leads to reductions in sexual assault victimization and increases in efficacy to resist a sexual assault and self-defense knowledge (Baiocchi et al., 2017; Decker et al., 2018; Sarnquist et al., 2014).
IMpower Class Outline.
The effectiveness of the IMpower program has not been evaluated yet in the United States, and to our knowledge, no studies have evaluated the impact of sexual assault prevention programming on American Indian youth. The purpose of the current study was to explore these gaps in the literature using a sample of American Indian middle and high school girls, some of whom received IMpower and some of whom did not. Moreover, given the high co-occurrence of sexual assault with sexual harassment and physical dating violence (Hamby & Grych, 2013; Sessarego et al., 2019) in addition to the fact that refusal skills may also help to prevent these other forms of victimization, we examined sexual harassment and physical dating violence as secondary outcomes. Specific hypotheses were as follows:
H1: Girls who received IMpower would report increases in efficacy to resist a sexual assault and self-defense knowledge (intermediary outcomes)
H2: Girls who received IMpower, compared to girls who did not, would report fewer types of SA victimization (primary outcome).
H3: Girls who received IMpower, compared to girls who did not, would report less physical dating violence and sexual harassment ictimization (secondary outcomes).
Method
Design
A quasi-experimental design was used such that American Indian girls residing on an Indian Reservation in the Great Plains region of the United States were compared to American Indian girls living in a city in the same state. Girls were not randomly assigned to receive the program on the Indian reservation given that there could be contamination effects in addition to the fact that our school and community partners felt that it was inappropriate to withhold the program from some girls in the school. Prior to the implementation of the program, several community members residing on the Indian reservation, including youth, provided feedback (via informal conversations) on the programming in order to make several adaptations to ensure its community relevance. This included the infusion of local language into the curriculum and adapting scenarios to be relevant to the local context. All research procedures were approved by the University of New Hampshire Institutional Review Board as well as Tribal Council and the Health Board on the Indian reservation. Finally, the President of the Tribe approved the publication of these data.
Participants
Participants were American Indian middle and high school girls in grades 6 through 12 residing either on an Indian Reservation in (N = 74; treatment group) or in a city in in the same state (N = 181; comparison group) who completed both baseline and follow-up surveys. The girls in the reservation (treatment) sample were on average 14.0 years old (range = 12 to 18; SD = 1.90), in grades 6 to 12 (median = 9th, SD = 2.03), and the majority identified as heterosexual (66.7%). The girls in the city (comparison) sample were on average 13.7 years old (range = 12 to 18; SD = 1.29), in grades 6 to 12 (median = 8th, SD = 1.07), and the majority identified as heterosexual (80.1%). The girls in the city sample come from a larger multiple baseline study examining the impact of a youth-led initiative on reducing rates of sexual assault (Edwards et al., 2021); all data collected in the city sample that is presented in this article occurred before the implementation of the initiative.
Data Collection
Written guardian consent procedures were used for participants under 18 years of age. Intensive recruitment procedures were used that included students receiving the consent forms in school, mailing the consent forms to guardians, calling, home visits, tabling at community events, and hosting community feeds. Figure 1 presents participant enrollment and retention data. Relative to attrition, there were no significant differences among the treatment participants instead of control participants between those who completed both pre- and post-tests compared to students who did not complete the post-tests with respect to age, grade, sexual orientation, mattering, bystander opportunity, binge drinking, grades, and victimization experiences. However, attrition analyses revealed that among treatment participants, those who endorsed alcohol use at the pre-test were less likely to participate in the post-test survey. For the control group, there were no significant differences among Time 1 participants and Time 2 participants relative to sexual orientation, mattering, alcohol use, binge drinking, bystander opportunity, and victimization experiences. However, attrition analyses revealed that among control participations, those with higher grades and who were older at Time 1 were less likely to take the time two survey. Ultimately, these variables that were related to attrition were included in the group of covariates considered in the propensity score analyses that aimed to equate condition groups at baseline. In combination with modern missing data techniques and the use of intent-to-treat, our analyses adequately addressed attrition (Enders, 2010).
Consort flow diagram.
Students completed surveys in school and received a small incentive (i.e., fruit snack, pencil, and the chance to win a gift card). Surveys were completed prior to and six months after the IMpower programming for the treatment sample and approximately six months apart for the control sample (all behaviorally oriented questions inquired about past six months for both samples). The girls’ surveys in the treatment sample were matched over time using a self-generated ID (i.e., subject number calculation form) which led to 94% of surveys being matched and allowed for anonymity of data. The girls’ surveys in the control sample were matched over time using a broker system which led to 100% of surveys being matched. A broker system is a method to match surveys over time by which an individual maintains the master list who never has access to the behavioral data (and creates individual log-on cards for participants that do not display the participant ID), and the researchers never have access to the master list, thus allowing the data to remain anonymous to researchers. Although the funds were not available to implement a broker system with the control sample, matching of surveys among girls in this condition was high.
Measures
Condition
Girls in the treatment group were coded as one, and girls in the control group were coded as zero.
Dependent variables.
Four items were used to assess sexual assault. Three items were taken from Cook-Craig et al.’s (2014) measure assessing sexual coercion (i.e., “Another student had sexual activities with you although you did not really want to because either the student threatened to end your friendship or romantic relationship if you didn’t or you felt pressured by the student’s constant arguments or begging”), physically forced sex (i.e., “Another student forced you to do sexual things that you did not want to do [count such things as kissing, touching, or being physically forced to have sexual intercourse]?”), and incapacitated sex (i.e., “You had sexual activities when you did not want to because you were drunk or on drugs?”). One item from the Youth Risk Behavior Survey (YRBS; Centers for Disease Control and Prevention, 2014) was used to assess for sexual dating violence (i.e., “Someone you were dating or going out with forced you to do sexual things that you did not want to do [count such things as kissing, touching, or being physically forced to have sexual intercourse]?”). Participants responded “yes” or “no” to whether the behavior had happened during the past six months. Each item was scored 0 (no experience in the past six months) or 1 (any experience in the past six months). All four items were summed with higher scores indicative of more categories of sexual assault victimization experiences. These items were administered to girls in both conditions.
Dating violence.
One item was drawn from the YRBS (2014) to examine physical dating violence (i.e., “Someone you were dating or going out with physically hurt you on purpose [count such things as being hit, slammed into something, or injured with an object or weapon]?”). Participants responded “yes” or “no” to whether the behavior had happened during the past six months. The item was scored 0 (no experience in the past six months) or 1 (any experience in the past six months). This item was administered to girls in both the reservation and city samples.
Sexual harassment.
Two items from the American Association of University (AAUW, 2001) women were used to assess for sexual harassment. Items included sexual comments (i.e., “Another student made sexual comments, jokes, gestures, or looks about/to you?”) and sexual rumors (i.e., “Another student spread sexual rumors about you?”). Participants responded “yes” or “no” to whether the behavior had happened during the past six months. The items were scored 0 (no experience in the past six months) or 1 (any experience in the past six months). The two items were then summed with higher scores indicative of more categories of sexual harassment victimization experiences. These items were administered to girls in both samples.
Efficacy to resist
An item to assess girls’ confidence to resist a potential attacker was modeled after previous research (Decker et al., 2018). The item read: “If I am attacked by a strong man, I feel confident that I can defend myself” with response options no (0) and yes (1). The items are displayed in Table 2. These questions were administered at pre-test and the 6-month post-test to girls only in the treatment condition.
Pre- to Post-test Changes in Self-efficacy to Resist a Sexual Assault and Self-defense Knowledge.
Note. SERSA = Self-efficacy to resist a sexual assault. SDK = Self-defense knowledge.
Sexual defense knowledge.
Four items assessed girls’ self-defense knowledge that was modeled after previous research (Decker et al., 2018). The items and response options are displayed in Table 2. These questions were administered at pre-test and the 6-month post-test to girls only in the treatment condition.
Covariates/cofounders.
Several additional surveys were included as measures of covariates/confounders for use in the propensity score analysis. These surveys were administered to girls in both the treatment and control conditions.
Mattering.
We used a single item from the YRBS (Centers for Disease Control and Prevention, 2014; Eaton et al., 2012) to assess for mattering. Participants responded to the item (i.e., “Do you agree or disagree that at your school you feel like you matter to people?”) on a four-point Likert scale ranging from 1 = Strongly disagree to 4 = Strongly agree.
Alcohol use.
We used two items from the YRBS (Centers for Disease Control and Prevention, 2014; Eaton et al., 2012) to assess for alcohol use (i.e., “During the past 30 days, on how many days did you have at least one drink of alcohol?” and “During the past 30 days, on how many days did you have 4 [if you are a female] or 5 [if you are a male] or more drinks of alcohol in a row within a couple of hours?”). Response options were as follows: 0 = 0 days, 1 = 1 or 2 days, 3 = 3 to 9 days, 10 = 10 to 19 days, and 20 = 20 to 31 days.
Grades.
A question the YRBS (Centers for Disease Control and Prevention, 2014; Eaton et al., 2012) inquired about grades in school during the past six months (i.e., “During the past 6 months, how would you describe your grades in school?”) with response options as follows: 1 = Mostly A’s, 2 = Mostly B’s, 3 = Mostly C’s, 4 = Mostly D’s, and 5 = Mostly F’s.
Bystander opportunity.
We used the work of Banyard (2015) as a base as well as adaptations of this work for high school samples by Coker et al. (2011). 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;” (d) “Saw or heard about a student spreading sexual rumors about another student.” Students were first asked the frequency of being in each situation in the past six months. Response options ranged from 0 = 0 times, 1 = 1–2 times, 3 = 3–5 times, 6 = 6–9 times, and 10 = 10 or more times. Items were scored to create a composite measure with higher scores indicative of more bystander opportunity.
Demographics.
We asked several demographic questions in order to describe our sample. We also included demographic variables as covariates/confounders in the propensity score analyses. Demographics assessed age, grade in school, and sexual orientation (i.e., recoded to reflect identification as a sexual minority [i.e., bisexual, something else, lesbian/gay], 0 = No, 1 = Yes). Only girls who identified as American Indian were included in the analyses.
Data Analysis
To assess changes in efficacy to resist a sexual assault and self-defense knowledge, the McNemer test, which is used for non-parametric paired nominal data, was used. For analyses examining differences in rates of sexual assault, sexual harassment, and physical dating violence (dependent variables [DV]) between the condition groups, propensity score weighting was used given our quasi-experimental design. Propensity score analysis is a rigorous quasi-experimental approach that aims to statistically equate two or more “treatment” groups on a set of observed confounders (i.e., covariates) as a means for minimizing selection bias in the evaluation of treatment effects, thus, producing more valid causal inferences related to questions of treatment efficacy (Rosenbaum & Rubin, 1983). Propensity score analysis methods reduce several covariates into a single variable (i.e., propensity score or balancing score), which represents the probability of being assigned to the treatment group given the vector of covariates. Conditional on the propensity score, the distribution of observed covariates is theoretically balanced across treatment groups. Following recommendations by Lanza et al. (2015), the propensity score analytic approach included several steps, detailed below.
Step 1 included the creation and estimation of the models that generated the propensity score weights using SAS 9.4. In this step, we first identified the potential covariates from pre-test (prior to the intervention; i.e., age, sexual minority status, grades, alcohol use, mattering, reactive bystander, as well as, pre-test indicators of our outcomes, sexual assault, sexual harassment, and physical dating violence) using theory and prior literature. We evaluated which covariates to include in the final propensity score models using bivariate tests to evaluate the magnitude of condition group differences on the covariates. Using t-tests (for continuous covariates) and chi-square analyses (for categorical covariates) using condition (0 = Comparison, 1 = Treatment) as the DV, covariates that resulted in absolute standardized mean differences >.20 (as an indication of selection bias; Stuart, 2010) were included in the model estimating the propensity score. This produced a final set of covariates that were included in the propensity score model (i.e., binge drinking, grades, bystander opportunity, sexual harassment). Second, we estimated the propensity score model parameterized as a logistic regression model. This model estimated the probability of receiving IMpower as predicted from our final set of covariates, resulting in an estimate of the propensity score for each participant (comparison and treatment group). Third, we checked whether the weighting was successful in improving covariate balance between conditions using each covariate’s absolute standardized mean differences (i.e., difference in means between conditions divided by the standard deviation) resulting from a model that included the propensity score weight. We used the .20 cut-point to determine if the differences were either balanced or still undesirable (Stuart, 2010).
In Step 2, we used the propensity scores as weights (inverse probability of treatment weighting; IPTW; Hirano & Imbens, 2001) to adjust for confounding in the causal analysis estimating the average treatment effect (ATE; i.e., the average effect that would be seen if individuals in both conditions received the intervention compared with those individuals not receiving the intervention; Harder et al., 2010). In estimating our ATE, we used the rigorous intent-to-treat approach to the analysis, in combination with recommended estimation procedures to address missing data; all individuals who completed Time 1 data collection were retained in the analysis (Enders, 2010). We compared sexual assault outcomes (post-test) for young women who received IMpower versus those in the comparison group, controlling for pre-test values. The outcome analysis estimated in SAS 9.4 included the condition indicator as the independent variable and the IPTW as an ATE weight (for details refer to Lanza et al., 2013). We estimated a Poisson regression for the count outcome variables of sexual assault and sexual harassment, and a logistic regression for the physical dating violence outcome variable. The Poisson regression coefficient of the exposure variable from this model, exponentiated, is the difference between conditions in the incidence rate ratio (IRR) for the sexual assault and sexual harassment outcomes, conditional on the covariates. The logistic regression coefficient of the exposure variable from this model, exponentiated, is the odds ratio (OR) of the outcome associated with difference in conditions in physical dating violence exposure, conditional on the covariates.
Results
As demonstrated in Table 2, girls who received the IMpower program increased significantly over time in their efficacy to resist a sexual assault. Girls who received the IMpower program also increased significantly over time in all domains of self-defense knowledge (H1).
Table 3 depicts raw descriptive statistics for outcome variables by condition prior to the propensity score analysis. Table 4 depicts the covariates that were retained in the propensity score model (i.e., binge drinking, grades, bystander opportunity, sexual harassment; met criteria absolute standardized mean difference >.20) and how the IPTW weighting improved covariate balance between conditions. After weighting, balance was improved on all covariates, as shown in the smaller absolute standardized mean differences (<.20).
Raw Descriptive Statistics for Outcome Variables (Prior to Propensity Score Weighting).
Covariate Balance Before and After Propensity Score Weighting.
Notes. RD = Raw group mean difference. ASD = Absolute standardized mean difference.
†p < .10. *p < .05.
Using the propensity score matching results in the context of Poisson regression, we tested whether girls who received IMpower, compared to girls who did not, reported lower rates of sexual assault victimization (primary outcome; H2; refer to Table 5). After IPTW weighting, the model for sexual assault suggested that for girls in the IMpower group the incident rate in sexual assault was .20 (95% CI = [.13, .31]) times the incident rate in the comparison group. This suggests that for girls in the IMpower group, compared to those in the comparison group, the incidence of sexual assault decreased by 80%.
Results of Post-intervention Outcome Models After Propensity Score Weighting.
Notes. CI = Confidence interval.
Condition coded: 0 = Comparison, 1 = Treatment.
Incidence rate ratios (IRR) provided for sexual assault and harassment from Poisson regression models.
Odds ratio (OR) provided for dating violence from logistic regression model.
We tested whether girls who received IMpower, compared to girls who did not, reported lower rates of sexual harassment and physical dating violence victimization (secondary outcomes; H3; refer to Table 5). After IPTW weighting, the model for sexual harassment suggested that for girls in the IMpower group the incident rate in sexual harassment was .74 (95% CI = [.55, .98]) times the incident rate for the comparison group. This suggests that for girls in the IMpower group, compared to the comparison group, the incidence of sexual harassment decreased by 26%. The physical dating violence outcome model estimated that for girls who received IMpower, compared to the comparison group, the estimated effect was modest and not statistically significant.
Discussion
The purpose of the current study was to examine how the IMpower program impacted intermediary (H1), primary (H2), and secondary (H3) outcomes in a sample of American Indian middle and high school girls. Consistent with previous evaluations of the IMpower program, there is evidence to support the effectiveness of the IMpower program in improving efficacy to resist a sexual assault and self-defense knowledge among participants (Baiocchi et al., 2017; Decker et al., 2018; Sarnquist et al., 2014). This finding extends previous research to a sample of American Indian girls demonstrating that knowledge and skills increase following participation in a empowerment self-defense program. Given the program’s focus on teaching girls about effective resistance strategies, it makes sense that we observed increases in this domain. Moreover, one of the unique aspects of the IMpower program is that it not only teaches girls skills to use in a potential sexual assault situation, but there is an emphasis on empowering girls to believe that they are worth defending, which may help to explain the self-efficacy finding. Indeed, as documented in another paper from this same project that reports on exit interviews with youth following the program, a theme that emerged is that girls reported that the program made them realize that “I am worth defending” (Siller et al., under review). The importance of girls finding their voice and realizing they are worth fighting for relates to the Tribe’s values of bravery and perseverance. This finding is especially important for American Indian girls who experience high rates of sexual assault, poverty, and parental absence, all of which are rooted in historical trauma and ongoing oppression of and discrimination against American Indian populations.
Perhaps the most promising finding is that compared to girls in the control group, American Indian girls who received the IMpower program reported lower rates of sexual assault, and effect sizes were large (i.e., 80% decrease in rates of sexual assault for girls in the treatment condition). This finding extends previous research with girls in East Africa (Baiocchi et al., 2017; Decker et al., 2018; Sarnquist et al., 2014) and young women in college in the United States (Gidycz et al., 2001; Gidycz et al., 2006; Orchowski et al., 2018; Senn et al., 2015), demonstrating that empowerment self-defense training is effective in reducing sexual assault among American Indian girls.
Furthermore, findings suggest that exposure to the IMpower program reduced sexual harassment. To date, we are not aware of any research examining the impact of empowerment self-defense training on reducing sexual harassment. It is possible that girls who received the IMpower training were perceived by boys as more assertive, which resulted in a lower likelihood of experiencing sexual harassment. Alternatively, the presence of the IMpower program in schools could have helped to shift social norms to be intolerant of sexual harassment and sexual assault, thus reducing the likelihood that boys would perpetrate sexual harassment. More research is needed to better unpack and understand these promising findings.
Whereas the IMpower program reduced sexual assault and sexual harassment, we found that the program did not have an effect on reducing physical dating violence. This finding may be an artifact of the single-item indicator we used to assess physical dating violence. It is also possible that an initial act of physical assault may not be thwarted, but that girls could put skills into action to stop the violence from escalating. For example, a boy may slap his girlfriend and then she may use resistance strategies (e.g., run away) to thwart additional violence. Given our limited measurement of physical dating violence, this nuanced response would not have been possible to detect. Alternatively, it is possible that the skills learned in IMpower may not translate to situations of physical dating violence. Although there is no empirical literature on this topic, the American Indian co-authors of this article believe that to effectively address physical dating violence, programs may need to include trauma-informed intervention components to help girls cope with high rates of exposure to domestic violence within the household that often co-occurs with substance abuse.
Limitations and Future Research
The current study’s limitations indicate opportunity for future research. First, we had a relatively small sample size limiting the generalizability of our findings, suggesting that future research should use larger sample sizes of American Indian girls when examining the effects of empowerment self-defense training. Second, we were unable to randomly assign schools/girls to a treatment or control condition, and our control and treatment groups were not perfectly matched given differences in age and geographic location, which may have impacted findings although rigorous propensity score analyses were used. Along these lines, we compared American Indian girls who resided on a reservation to American Indian girls in a nearby city. It is important to note that American Indian youth in this region of the country are transient, spending time in both the city and on the nearby reservation. Although this could have presented potential contamination issues, we would expect that if contamination were an issue, we would have found less robust intervention effects. Nevertheless, future research would benefit from using cluster randomized control trials in which Indian reservations are randomly assigned to treatment or wait-list control conditions, as long as this is acceptable to tribal communities. Third, due to time constraints, we were limited in what constructs we could measure, and measures of some other potential intermediary variables (e.g., collective empowerment) were not included. Thus, the impact that the program had on other outcomes is unknown. Also, although all girls regardless of condition received victimization questions, other questions like efficacy to resist a sexual assault were only administered to girls in the treatment condition given the dataset associated with the control community was pre-existing prior to the launch of the project and thus causal relationships about the impact of the programming on these variables cannot be determined given our design limitations. Along these lines, several of our measures, although used in previous research, do not have documented psychometric properties. Similarly, the fact that we used single-item indicators for some of our variables is also a limitation and future research should use more comprehensive measures. Fourth, there were less than ideal enrollment and retention rates, which is not uncommon for highly impoverished communities seeking to do public health research. Thus, the extent to which the program worked for youth not enrolled in the research procedures is unknown. Fifth, although the analyses that we conducted controlled for differences in the treatment and control conditions on numerous variables, it is noteworthy that 33.3% of girls in the treatment condition identified as a sexual minority (compared to 19.9% of girls in the control condition).
Conclusions
The current study represents the first-ever evaluation of a sexual assault prevention program in a sample of American Indian girls. These findings show strong support for the implementation of the IMpower program with this population and should be considered when sexual assault and sexual harassment prevention efforts are being developed that specifically target American Indian girls. Although other evidence-based components such as bystander intervention training (Orchowski et al., 2018) are also needed as part of comprehensive prevention practices, these data provide preliminary evidence of one effective method to reduce the public health burden of sexual assault and harassment among American Indian girls.
Footnotes
Acknowledgments
We owe a great deal of gratitude to our school and community partners and the research assistants and program facilitators. Without these agencies and individuals, this project would not have been possible.
Declaration of Conflicting Interests
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for the collection of the control group data was provided by the U. S. Centers for Disease Control and Prevention’s (CDC), National Center for Injury Prevention and Control, Grant #U01-CEO02838. Disclaimer: The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the CDC.
