Be Under Your Own Influence: Effectiveness of a Culturally-Adapted Drug Prevention Campaign Targeting Reservation-Dwelling American Indian Youth

Abstract

Aims:

Be Under Your Own Influence (BUYOI) is a previously validated school-based intervention designed to delay adolescent substance use (SU) initiation. This study examined the effectiveness of a culturally-adapted version of BUYOI in delaying SU initiation among reservation-dwelling American Indian (AI) youth.

Methods:

Five reservation-based middle schools participated. Three schools were randomly assigned to receive BUYOI-AI (N=321), and two schools served as controls (N=176). Beginning in 7^th grade, all participating students completed 4 assessments over the study period. Discrete time hazard models estimated the effects of BUYOI on students’ risk of initiating alcohol, alcohol intoxication and marijuana before the end of 8^th grade.

Results:

AI students exposed to BUYOI had a lower risk of initiating alcohol use or intoxication, though sex moderated the effect on intoxication.

Conclusion:

These findings provide preliminary support for the effectiveness of a culturally-adapted version of BUYOI in delaying AI youth’s first-time alcohol use and intoxication

American Indian (AI) adolescents report the highest levels of substance use compared to other U.S. youth, and recent epidemiological findings have shown that AI adolescents who reside on or near reservations demonstrate significantly higher lifetime and past month substance use rates relative to comparable national rates across all measured substances, with the exception of tranquilizers and amphetamines (Swaim & Stanley, 2018). Recent findings have also demonstrated that reservation-based AI adolescents start using substances like alcohol, marijuana, cigarettes, and inhalants at significantly earlier ages relative to the general population (Stanley & Swaim, 2015). These substance use disparities have persisted for decades among reservation-based AI youth, with little progress made towards reducing them (Stanley, Swaim, Kaholokula, Kelly, Belcourt & Allen, 2017; Stanley, Harness, Swaim, & Beauvais, 2014).

Reasons for these disparities are varied, but one important factor is the lack of effective, culturally appropriate interventions for this unique population. In response to a National Institute of Health (NIH) initiative to investigate health promotion and disease prevention among Indigenous populations, we adapted and tested the intervention “Be Under Your Own Influence” in a randomized controlled trial conducted in several reservation communities (Stanley, Kelly, Swaim, & Jackman, 2018). This paper presents our findings.

BUYOI.

Be Under Your Own Influence (BUYOI) is a media-based substance use prevention program developed through formative research and evaluation across various adolescent populations (Kelly, Comello, & Slater, 2006; Slater, Kelly, Edwards al., 2006; Slater, Kelly, Lawrence, Stanley, & Comello, 2011). The messages, targeted to middle-school youth, emphasize non-use as an expression of personal identity and autonomy—critical developmental challenges for middle-school youth, as well as the consistency of non-use with adolescents’ future aspirations. Although behavioral change theories such as the Theory of Reasoned Action contributed to the development of BUYOI, the intervention broadly relies on the communication theory of Re-framing (Slater, 2006), where campaign messages are designed to reframe substance use as inconsistent with individuals’ sense of personal autonomy and their future aspirations. BUYOI is also consistent with Self-Determination Theory (SDT), a broad framework defining intrinsic and extrinsic sources of motivation that contribute to individual well-being (Ryan and Deci, 2017). According to SDT, self-determination and intrinsic motivation are both supported through conditions that foster autonomy, competence, and relatedness (Ryan & Sapp, 2007). BUYOI is designed to bolster adolescents’ sense of personal autonomy and relatedness with others. Autonomy is bolstered by emphasizing that youth can and should make their own decisions about substance use, rather than having decisions made for them by peers and others (Slater, 2006), while relatedness is bolstered by capitalizing on the prosocial influence of older peers during a key developmental period for middle-school students (Karcher, 2005).

BUYOI-AI.

Using extensive Community Based Participatory Research (CBPR), both surface-structure and deep-structure adaptations were made to the original BUYOI intervention for the purposes of cultural adaptation (Stanley et al., 2018). The themes of autonomy and aspirations, along with the BUYOI slogan, were found to be especially culturally congruent and motivational by both AI youth and adults, and therefore, remained the campaign’s cornerstone themes. In partnership with community stakeholders, deep structure adaptations included developing campaign messaging related to finding strength in one’s tribal history, culture, and identity, and the use of high school role models to deliver campaign messages personally and through media (e.g., posters), often using personal narratives of strength and resilience. Surface structure adaptations consisted primarily of replacing original images with those reflective of AI youth, including their environment and culture.

Present study

This study examined the effectiveness of the culturally-adapted BUYOI intervention (BUYOI-AI) in reducing risk of first-time alcohol use, alcohol intoxication, and marijuana use among reservation-dwelling AI youth. Using event history analysis, the following hypotheses were tested:

• H1: Controlling for age at first-assessment, students attending BUYOI-assigned schools will have a lower risk of substance use uptake by the end of their eighth-grade year, relative to students attending control schools.

Additionally, BUYOI-AI was designed to increase students’ perceptions that substance use is inconsistent with future aspirations and personal autonomy. As such, we examined the main effects of students’ perceptions of substance use consequences for autonomy and personal aspirations on hazard of first-time use. Based on prior research findings and the underlying theoretical assumptions of BUYOI:

• H2: Higher levels of autonomy and future aspirations will predict lower hazard of first-time use.

Finally, there is limited empirical research examining the role of sex differences in response to substance use prevention efforts among AI youth. Those few studies that have investigated sex differences in the effectiveness of school-based substance use interventions have generally found mixed results (see Blake, Amaro, Schwartz, & Flinchbaugh, 2001 for a review; Vigna-Taglianti, Vadrucci, Faggiano, Burkhart, Siliquini, Galanti, & EU-Dap Study Group, 2009; Kulis, Yabiku, Marsiglia, Nieri, & Crossman, 2007). Because some of these studies suggest differential impacts by sex, we examined the potential moderating role of sex on the impact of BUYOI-AI.

Method

Design

BUYOI-AI efficacy was tested in a multi-year intervention study from 2016 through 2018, with six middle schools recruited for participation. Each middle school was randomly assigned to treatment (BUYOI-AI intervention) or control (no exposure to BUYOI-AI) conditions in a randomized cluster design. Due to a change in school personnel, one control school withdrew from the project. Two of the three treatment schools were located in the Northern Plains while one was located in the Southwest. Due to the potential for regional confounding in the intervention effects, we opted to exclude southwest school cases from analyses for the present study—see “recruitment and participants” section below. Both control schools were located in the Northern Plains. As part of the study, two consecutive cohorts of 7^th grade youth were followed for two years. For each treatment middle school, the associated local high school was asked to participate, and juniors were recruited to serve as role models to the 7^th grade students.

Intervention Description

At the beginning of each school year, local high school students in their junior year were recruited to be role models to the 7^th grade youth attending the treatment schools. Requirements for role models were a pledge to be drug and alcohol free throughout the intervention, along with active parental consent and student assent. Intervention activities consisted of distributing visual media (posters, banners, and business cards) in the middle schools, giving a presentation at a 7^th grade assembly about the personal meaning of being “under their own influence”, and distributing promotional items with the logo and tagline in classrooms or in small groups during lunch period. A school staff member advised each role model group, and research staff provided detailed instructions for each activity. Visual media (e.g., posters) used the same language across treatment schools, with some images localized by school (using role model photos) and other images that were identical across schools. Role models were encouraged to respond to questions asked by the younger students, using their personal experiences of choosing to be substance free while working toward their future aspirations. However, not all classroom presentations were specific to substance use— rather, role models often opted to discuss their general high school experiences, reinforcing the fact that the majority of high school students are autonomously making wise decisions to advance their futures.

Recruitment and Participants

Local staff were hired in each community to recruit 7th graders to participate in the study, with active parental consent and student assent required for participation. Recruitment was moderately high for each cohort, with 63% of 7^th grade students recruited for cohort 1, and 65% recruited for cohort 2. Retention rates were high because the research team worked directly with local school staff to ensure consented students completed their surveys at each measurement occasion. Approximately 84% of students who consented to participate in the study completed the final survey.

Across all participating schools, a total of 510 students were recruited for the original study, with 325 students attending intervention schools, and 185 attending control schools. Of those, 13 non-AI students were excluded from current analyses as the present study targets intervention outcomes for AI youth. Additionally, to mitigate the possibility of regional confounding of intervention outcomes, an additional 52 AI students recruited from a Southwest regional school and originally assigned to the intervention condition were excluded from current study analyses.

The final N included in present study analyses was 269 AI 7^th graders who received BUYOI-AI, and 176 AI 7^th graders who did not receive the intervention (445 students total). The average age of participants was 12.5 (SD = 0.6) at their first assessment, and 13.8 (SD = 0.56) at their last assessment. Participants were 50% female, and all identified as AI (see Table 1 for full demographic information).

Table 1.

Sample Descriptives

	Variable	Categories / Description	N (M)	% (SD)
Demographics	Sex
		Female	222	50.0
		Male	222	50.0
	Racial composition†	American Indian / Native American Only	297	66.7
		White	103	23.1
		Black or African American	27	6.1
		Other	61	13.9
	Ethnicity	Hispanic	59	13.3
		Non-Hispanic	386	86.7
	Age	Years	12.5	0.6
Predictors
	BUYOI	Intervention	269	60.4
		Control	176	39.6
	Autonomy - Marijuana	Perceived consequences of marijuana use for autonomy	3.5	0.8
	Autonomy - Alcohol	Perceived consequences of alcohol use for autonomy	3.7	0.6
	Aspirations - Marijuana	Perceived consequences of marijuana use for aspirations	3.1	0.9
	Aspirations - Alcohol	Perceived consequences of alcohol use for aspirations	3.3	0.9
Censored Criterion Variables ‡
	First time Alcohol use	Pre-Baseline	49	11.0
		Risk Period 1	110	24.7
		Risk Period 2	102	22.9
		Risk Period 3	152	34.2
	First time Alcohol intoxication	Pre-Baseline	23	5.2
		Risk Period 1	46	10.3
		Risk Period 2	76	17.1
		Risk Period 3	94	21.1
	First time Marijuana use	Pre-Baseline	92	20.7
		Risk Period 1	137	30.8
		Risk Period 2	179	40.2
		Risk Period 3	191	42.9

Note. With the exception of the criterion variables, all variables described in table represent variables at measured at baseline.

M = Mean; SD = Standard Deviation

^†Racial composition categories are not mutually exclusive, with the exception of “American Indian / Native American Only”. Though all students in this study identified as American Indian, students could select more than one racial category in the survey.

^‡Reflects the number of participants who reported lifetime substance use during a particular risk period.

Procedures

Each cohort was assessed at four time points, at the beginning and end of each school year, with the first assessment occurring prior to BUYOI-AI implementation. Using Qualtrics survey software, assessments were completed online during school hours or immediately after school. All procedures were approved by our university IRB. In addition, procedures were approved by the local school boards and the appropriate tribal IRBs. Each school received $500 per year for participating, as well as a comprehensive written report of their school’s survey results.

Measures

Age at baseline was included as a covariate in all models. In addition to the effect of the intervention, predictors included in each model were sex (male/female), perceptions of alcohol/marijuana use consequences for autonomy, and perceptions of alcohol/marijuana use consequences for future aspirations. Perception of impact on autonomy was measured with 4 items (e.g., “A way to be true to myself is to NOT drink alcohol”; α=.92) while perceptions of use on future aspirations was measured with 3 items (e.g., “One thing that could keep me from doing what I want to do is drinking alcohol”; α=.93). Each item was measured on a 5-point Likert scale (1=strongly disagree, 5=strongly agree), with larger scores reflecting stronger perceptions of consequences. Outcomes included hazards of first-time alcohol use, first-time marijuana use, and first-time alcohol intoxication (“Have you ever gotten drunk?”), dichotomized (0=never, 1=one or more times).

Statistical Analysis

Given the discrete characterization of time as interval-censored risk-periods, we used complementary log-log discrete time hazard models, implemented within R, to estimate the effects of BUYOI-AI and the covariates on the risk of first-time alcohol use, alcohol intoxication, and marijuana use during 7^th and 8^th grades. A discrete time hazard model is an optimal approach for analyzing longitudinal data characterized by non-continuous time (Singer & Willet, 2003). In this study, the discrete time hazard represents the conditional probability that a student will experience first time alcohol use, intoxication, or marijuana use (assessed as separate outcomes) during a measurement period, given that students did not use that substance during a previous period. Each variable in the model is considered time-invariant for the purpose of this analysis. Thus, for each model, we tested 1) the hypothesized main effects of BUYOI-AI and perceived consequences of use for aspirations and autonomy controlling for age and sex, and 2) the two-way interaction between sex and treatment.

Event Variable.

Event variables consisted of three dichotomized dependent variables calculating self-reported alcohol use, alcohol intoxication or marijuana use for the first-time in known participant history, where use was not previously reported in prior measurement occasions. That is, separately for each outcome, the event variable was first-time use reported at measurement occasion 2 after no use was reported at baseline, reported use at occasion 3 after none reported at occasion 2 or baseline, and reported use at occasion 4 with none reported prior to occasions 3, 2 or baseline. The same dichotomization was applied to intoxication, noting that those who reported first-time intoxication may have used alcohol previously without intoxication.

Episodes.

In this study, episodes represent the period of time between measurement occasions. As such, each episode represents the measured period of time between each wave of assessments. The average period of time between assessments was approximately 5 months.

Duration Variable.

The duration variable is the duration of participation in the media campaign intervention period across all four waves, beginning with the transition from first to second wave of data collection, with the entire duration lasting an average of 1 year and 4 months.

Censoring Indicator.

If no marijuana use, alcohol use, or alcohol intoxication was reported (separately for each model), the censoring indicator represents completion of study at wave 4. If substance use was reported, the censoring indicator indicates the risk period (the period of time between waves 1 and 2, 2 and 3, or 3 and 4) during which the participant experienced uptake of substance use. Approximately 11% (n = 49), 20.7% (n = 92) and 5.2% (n = 23) of participants reported lifetime alcohol use, marijuana use or alcohol intoxication at baseline, respectively (see Table 1). As such, these participants were censored at baseline, with censoring specific to each model.

Recanting and missing data.

If a student reported initiating marijuana use, alcohol use, or alcohol intoxication during a given risk period, but recanted their lifetime use in a subsequent risk-period, the student’s initial report was treated as the true report. Recanting is a common problem in school-based longitudinal surveys of lifetime substance use, with prior works documenting 25% to 84% of students recanting previously reported lifetime use depending on the reported substance. The percentages of students who recanted in the present study were comparatively lower, with 16%, 13% and 8% recanting prior lifetime alcohol use, intoxication, and marijuana use. Percentages of student non-response to the three dependent variables across all four time-points ranged from 11% to 19%. Person-periods with missing responses to a given dependent variable were excluded from the analysis including that dependent variable. All other missing data was treated as missing at random, with parameter estimates and standard errors estimated within a general linear modeling framework using full-information maximum likelihood.

Results

Descriptives.

Girls were disproportionately represented among students who reported first-time alcohol use (63% girls vs. 37% boys, χ² [1] = 28.6, p < .001), alcohol intoxication (66% vs. 34%, χ² [1] = 17.8, p < .001) and marijuana use(57% vs. 43%, χ² [1] = 7.6, p < .006 ), before the end of the study. Likewise, by the end of the study, a larger percentage of students in the control group than in the intervention group reported first-time use of alcohol (47% of control vs. 36% of intervention, χ² [1] =5.3, p = .021), while a marginally greater percentage of students in the control group than in the intervention group reported first-time intoxication, (30% vs. 22%, χ² [1] =3.7, p = .054). Interestingly, by the end of the study, percentages of first-time marijuana users were roughly equivalent in the control and intervention groups (53% vs. 51%, χ² [1] = 0.2, p = .659). With respect to covariates, AI girls and boys showed no difference in their baseline perceptions of the consequences of alcohol or marijuana use for their personal autonomy (M_alcohol = 3.71, SD = 0.58 (girls) vs. 3.67, SD = 0.69 (boys), t (371) = 0.66, p = .508; M_marijuana = 3.56, SD = 0.73 vs. 3.49, SD = 0.85, t (371) = 0.86, p = .391) or their future aspirations (M_alcohol = 3.35, SD = 0.83 vs. 3.24, SD = 0.99, t (372) = 1.15, p = .250; M_marijuana = 3.15, SD = 0.91 vs. 3.15, SD = 0.98, t (373) = 0.06, p = .949).

Finally, there were no significant differences between students attending control schools vs. those attending intervention schools in their baseline perceptions of alcohol use consequences for their personal autonomy (M_alcohol = 3.70, SD = 0.51 vs. 3.68, SD = 0.71, t(372) = 0.21, p = .835), though there was a marginal difference between the groups in perceptions of marijuana use consequences for personal autonomy (M_marijuana = 3.61, SD = 0.62 vs. 3.47, SD = 0.88, t(372) = 1.79, p = .074. There were no significant differences between groups in baseline perceptions of the consequences of either alcohol or marijuana use for future aspirations (M_alcohol = 3.33, SD = 0.90 vs. 3.28, SD = 0.92, t(372) = 0.58, p =.562; M_marijuana = 3.16, SD = 0.96 vs. 3.14, SD = 0.94, t (373) = 0.26, p = .796). Table 1 presents additional demographic descriptives for the sample.

Discrete Time Hazard Model.

To model the hazard of first-time substance use in AI youth, we assessed two discrete time hazard models for each substance use outcome (alcohol, alcohol intoxication, and marijuana). Model 1 examined the main effects of BUYOI-AI, baseline perceived consequences of alcohol or marijuana use for personal autonomy, and baseline perceived consequences of alcohol or marijuana use for future aspirations, on substance use initiation (main effects model), controlling for biological sex and age at baseline. Model 2 included the same predictors and control variables, plus the interaction of treatment group and sex (interaction model). Improvement in fit based on addition of the interaction terms over the main effects of treatment and sex was assessed by evaluating model AIC, with smaller AICs suggesting better fit. The interaction model was better or equal in fit relative to the main effects model across all analyses, with the exception of marijuana use initiation (See Table 2). The following descriptions of effects represent parameter estimates for the best fitting model.

Table 2.

Predictors of First-Time Alcohol Use, Alcohol Intoxication, and Marijuana use

Predictors		A. First-Time Alcohol Use					B. First-Time Alcohol Intoxication					C. First-Time Marijuana Use
		exp (β)	β	S.E.	p		exp (β)	β	S.E.	p		exp (β)	β	S.E.	p
Interval 1		0.20	−1.60	0.13	<0.001		0.12	−2.15	0.44	<0.001		0.20	−1.62	0.14	<0.001
Interval 2		0.18	−1.71	0.14	<0.001		0.25	−1.40	0.41	<0.001		0.29	−1.23	0.12	<0.001
Interval 3		0.35	–1.05	0.11	<0.001		0.38	−0.97	0.41	<0.001		0.42	−1.09	0.11	<0.001
Age		1.22	0.20	0.12	0.107		1.53	0.43	0.15	0.005		1.00	0.01	0.12	0.989
Sex (0 = Female)		0.47	−0.76	0.15	<0.001		0.52	−0.66	0.20	0.001		0.77	−0.26^	0.14	0.060
Future Aspirations		0.90	−0.11	0.09	0.233		0.80	−0.23^	0.12	0.053		0.80	−0.23	0.09	0.012
Personal Autonomy		0.72	−0.32	0.10	<0.001		0.65	−0.43	0.12	<0.001		0.62	−0.47	0.09	<0.001
Treatment Group (0 = Control)		0.66	−0.42	0.15	0.005		0.64	−0.44	0.20	0.023		1.15	0.15	0.14	0.294
Gender × Treatment Group		0.66	−0.41	0.30	0.164		0.47	−0.76^	0.39	0.052		0.77	−0.27	0.28	0.341
AIC -MEM		911.58					669.87					839.85
AIC - IM		911.53					667.98					840.94

Note. Parameters in bold are significant at p < .05.

^{^}indicates marginal significance at p < .10. All variables were centered prior to inclusion in the model.

MEM = Main Effects Model; IM = Interaction Model; exp = exponentiated

First-time alcohol use.

Model parameters are reported in Table 2a. Results indicated a main effect of sex, treatment condition, and perceptions of consequences of alcohol use for personal autonomy. AI boys had a 53% reduced risk of alcohol use uptake relative to AI girls before the end of the study period (β [0.15] = −0.76, p < .001). Controlling for sex and age at first assessment, students exposed to BUYOI-AI had a 34% reduced risk of first-time alcohol use by the end of the study period, consistent with our first hypothesis (β [0.15] = −0.42, p = .005). Partially consistent with our second hypothesis, greater levels of perceived consequences for autonomy at baseline were protective against later alcohol use initiation, with risk dropping 28% for one standard deviation positive difference in perceived consequences for autonomy between students (β [0.10] = −0.32, p =.001). Figure 1 plots the hazard of alcohol uptake by treatment group and sex.

Figure 1.

Hazard of alcohol uptake among AI youth by treatment group and sex

Note. h(t) = hazard function, where t = assessment point

First-time alcohol intoxication.

Model parameters are reported in Table 2b. There was a main effect of both intervention and sex, as well as perceptions of consequences of alcohol use for personal autonomy and future aspirations. Once again, AI boys had a significantly lower risk of first-time intoxication from their 7^th to 8^th grade year, with risk being 48% lower than AI girls (β [0.20] = −0.66, p < .001). As hypothesized, students exposed to BUYOI-AI had approximately 36% lower risk of first-time intoxication by the end of the study period than those in the control condition (β [0.20] = −0.44, p = .023). Furthermore, students who believed alcohol use would interfere with their personal autonomy and future aspirations demonstrated a significant 35% decreased risk (β [0.12] = −0.43, p < .001), and a marginally significant 20% decreased risk (β [0.12] = −0.23, p = .053), respectively, of first-time alcohol intoxication given a one standard deviation positive difference between students. Finally, though the interaction only trended towards significance at p = .053 and thus should be interpreted with caution, BUYOI-AI appeared to be marginally more effective at reducing risk of first-time intoxication for AI boys relative to girls (β [0.39] = −0.76, p =.052). Compared to boys in the control group, AI boys exposed to BUYOI-AI demonstrated an approximately 57% lower risk of first-time intoxication before the end of the study period. By contrast, AI girls exposed to BUYOI-AI did not show a significant reduction in risk of first-time intoxication relative to their peers in the control condition. Figure 2 plots the hazard of first-time alcohol intoxication by treatment group and sex.

Figure 2.

Hazard of alcohol intoxication uptake among AI youth by treatment group and sex

Note. H(t) = hazard function, where t = assessment point

First-time marijuana use.

Model parameters are reported in Table 2c. For first-time marijuana use, there was a main effect of consequences of use for personal autonomy and future aspirations; however, there was no effect of BUYOI-AI exposure. Departing slightly from the outcomes for alcohol use and intoxication initiation, the main effect of biological sex was marginally significant, with AI boys being 23% less likely to use marijuana for the first time before the end of the study period, relative to AI girls (β [0.14] = −0.26, p = .06). Likewise, risk of first-time marijuana use dropped 20% (β [0.09] = −0.23, p = .012) and 38% (β [0.09] = −0.47, p < .001) with a 1 standard deviation increase in aspiration and autonomy between students, respectively. Figure 3 plots the hazard of marijuana use uptake by treatment group and sex.

Figure 3.

Hazard of marijuana uptake among AI youth by treatment group and sex

Note. H(t) = hazard function, where t = assessment point

Discussion

Consistent with H1, all students exposed to BUYOI-AI had a reduced risk of first-time alcohol use by the end of the intervention period; however, our findings suggest that only AI boys exposed to BUYOI-AI demonstrated a reduced risk of first-time alcohol intoxication relative to their control group peers. Interestingly, BUYOI-AI did not appear to have a significant effect on girls’ risk of initiating alcohol intoxication before the end of the study period. Unexpectedly, BUYOI-AI did not significantly delay initiation of marijuana use among AI adolescents. This finding is inconsistent with prior studies evaluating the effectiveness of BUYOI within non-AI adolescent populations (Slater et al., 2006; Slater, Kelly, Lawrence, Stanley & Comello, 2011), which found stronger effects on marijuana initiation relative to alcohol initiation.

Marijuana use is significantly more prevalent than alcohol use among AI reservation-dwelling middle-school students (Swaim & Stanley, 2018) and initiation of marijuana use occurs at an earlier age than alcohol (Stanley & Swaim, 2015; Sapp, et al., 2012). Consistent with this, 1 in 5 AI middle-school students participating in this study reported lifetime marijuana use at baseline, compared with 1 in 10 reporting lifetime alcohol use. One potential reason underlying the disparate effects of BUYOI is the observed differences in perceived harm and negative consequences associated with alcohol versus marijuana use among participants who reported using either substance before the end of the study period. In their integrated model of communication, Eveland and Cooper (2013) note the impact of “priors” (e.g., beliefs and values) on communication exposure and processing, along with the ultimate effects of exposure and processing on resultant beliefs. For our study, “priors” consist, in part, of beliefs about the harmful effects of marijuana and alcohol use, including effects on autonomy and aspirational goals. Of note, students at time 1 perceived, on average, fewer adverse effects (i.e., perceived harm and negative consequences of use) of marijuana use as compared to alcohol use. Eveland and Cooper’s model suggests that messages communicating that substance use is incompatible with autonomy and achievement of aspirations may be more congruent with AI middle-school students’ beliefs about alcohol use and may be less applicable to marijuana or even counter-arguable. Indeed, Eveland and Cooper note that when individuals are aware of persuasive intent, they are more likely to counter argue the message if it is contradictory to their prior beliefs. Thus, persuasive messages targeted specifically to alcohol may be more noticed and attended to by students in the present study as compared to messages targeted to marijuana. With their model in mind, future works should examine the role of priors (that is, norms and values) in the effectiveness of BUYOI-AI across multiple substances.

Consistent with H2, AI students who reported stronger beliefs that substance use would interfere with their personal autonomy and future aspirations demonstrated a significantly lowered risk of initiation across substances, including first-time intoxication. These results are consistent with both Reframing (Slater, 2006) and Self-Determination Theories (Ryan & Deci, 2017) as applied to substance use prevention in AI youth. Future works should carefully examine the mediating roles of autonomy and future aspirations with respect to BUYOI-AI’s impact on substance use initiation. Furthermore, though the Be Under Your Own Influence Campaign messaging was originally designed to strengthen students perceptions that alcohol and marijuana use would interfere with their personal autonomy and future aspirations, the culturally-modified version of this intervention (BUYOI-AI) involved other key components that may have contributed to unmeasured differences in substance use behaviors between the control and treatment schools. For example, initial qualitative evaluations of BUYOI-AI (Stanley et al., 2018), indicated that the involvement of the role-models may have had a powerful impact on how the intervention was received by students. Future works should focus on quantitatively evaluating the effectiveness of each component of the intervention in this population (i.e., messaging content, role-model involvement, etc.) to better explicate the mediating processes involved in this cultural adaptation of BUYOI.

While the risk of initiating alcohol use was lower among AI boys and girls receiving BUYOI-AI compared to those in the control condition, only AI boys attending BUYOI-AI schools showed significant delays in first-time alcohol intoxication relative to their male peers attending control schools. While research examining sex differences in age of initiation or frequency of alcohol use or intoxication among AI youth is extensive (Stanley & Swaim, 2015; Whitbeck & Armenta, 2015), sex differences in the effectiveness of substance use prevention interventions are surprisingly under-examined in the substance use intervention literature (Blake, et al., 2001; Kumpfer, Smith, & Summerhays, 2008) in this population. Indeed, no studies that the authors are aware of explicitly address sex differences in the efficacy of interventions targeting alcohol and marijuana uptake among AI adolescents.

Limited research has examined the etiological mechanisms underlying sex differences in intervention effects, and findings have been mostly inconsistent and difficult to generalize across different populations, age-groups and prevention frameworks (Kumpfer, Smith, Franklin, & Summerhays, 2008).The scant studies that have examined sex moderation have yielded some evidence suggesting that adolescent girls respond better to interventions targeting family-related protective factors (e.g. family bonding, parental monitoring) while boys respond better to school and community-related protective mechanisms (Amaro, Blake, Schwartz, Flinchaugh, 2001; Kumpfer et al., 2008). Sex differences in effectiveness may be a particular concern for “selective” culturally-adapted intervention programs like BUYOI-AI, relative to “universal” substance use prevention programming, as the impact of culture is likely to intersect with sex (Kulis, Yabiku, Marsiglia, Nieri & Crossman, 2007). Thus, it is possible that AI girls and boys respond differently to BUYOI-AI messaging and delivery, particularly with respect to the themes of autonomy and future aspirations (Crabtree, Stanley & Swaim, 2020). Future works should systematically investigate sex differences among AI adolescents in the hypothesized mechanisms of BUYOI-AI, including variables related to personal autonomy and future aspirations.

Limitations

While compelling, the results of this study should be considered in light of several limitations. First, the sample is not a random sample of reservation-based AI middle-schoolers. Likewise, the participating reservation middle schools are not representative of all middle-schools located on or near reservations. Participation by schools was voluntary, and although we originally attempted to match control and treatment schools by region and size, this design was not possible due to fewer schools participating than originally recruited. In addition, one assigned control school dropped out of the study in year 1 due to a change in school administration, and one school was excluded from present analysis to avoid confounding region and treatment assignment, leaving two treatment and two control schools. Unfortunately, due to this small number of participating schools, this study could not employ multilevel analysis to properly account for clustering of students within schools.

Additionally, we cannot discount the possibility of unmeasured confounding effects, nor can we conclude that the results of this study are solely attributable to the effects of BUYOI-AI. Although we are not aware of other substance use prevention campaigns occurring during this study, there may have been differences between control and treatment schools that could explain, at least in part, differences in initiation rates found between students attending these schools. Similarly, although student recruitment rates for the longitudinal survey were above 60%, relevant differences may exist between participating and non-participating students, potentially influencing results. We attempted to minimize selection effects by contacting parents/guardians through local liaisons to explain the study and gain consent, rather than relying solely on the student to do so.

Another important limitation of these findings is student inconsistencies in lifetime substance use reporting. In particular, some students who reported lifetime alcohol use, intoxication, or marijuana use at an earlier time-point denied lifetime use at a subsequent time point (i.e. they recanted). Conversely, it is possible that some students failed to report lifetime use during the study. Unlike students who recanted, however, the number of students who underreported their use is not possible to identify. Student inconsistencies in reports of lifetime substance use are an extremely common issue in longitudinal studies (Fendrich & Rosenbaum, 2003), and can introduce measurement error in analysis dependent variables. So long as there are no systematic differences in students’ likelihood of recanting vs. under-reporting, however, such measurement error is unlikely to produce substantial estimation bias. Unfortunately, as it is not possible to identify students who under-reported their lifetime use in the current dataset, the degree of systematic measurement error due to reporting inconsistency is indeterminate.

Conclusion

Despite these limitations, our findings suggest BUYOI-AI may serve as an effective substance use prevention intervention, particularly targeting alcohol use and intoxication among reservation-dwelling AI adolescents. Moreover, the use of local older students to model drug-free behavior as a way to achieve goals and increase personal autonomy is a productive line of future investigation into reducing substance use among AI youth, in addition to encouraging alternative positive behaviors.