Trial design
The study design is a cluster randomized trial (CRT) with one baseline survey and six follow-up surveys (Fig. 1). The intervention condition involves exposure to three years of an integrated multi-level intervention within the school and community. The control condition communities will be offered the project intervention materials, resources and trainings after follow-up evaluation has been completed. Baseline assessment will take place among a cohort of youth during the fall semester of 10th grade before intervention initiation. Intervention activities will begin during the cohort’s 10th grade year, and will continue regardless of school status and for six months following on-time graduation. Students who are suspended, expelled or drop out of high school and remain within the community will continue to be tracked for follow-up surveys. Follow-up surveys will be conducted every six months over a three-year period measuring changes in proximal and distal outcomes as the intervention is initiated, ramps up, and becomes routine. The final follow-up survey will be conducted six months after on-time graduation, allowing assessment of changes in drug use outcomes during the important transition from high school to young adulthood.
Human subjects’ approval was received from the Cherokee Nation IRB, which is serving as the single IRB for this multi-site trial involving Cherokee Nation Behavioral Health and Emory University. All data collected will be de-identified and stored on secure servers accessible only to members of the research team. Any adverse events or protocol modifications will be tracked and, when indicated, reported in a timely manner to the institutional review board and funding agency.
Study setting, participants, and recruitment
The target population is students attending high schools in small rural towns in the 14 counties that partially or fully fall within the Cherokee Nation reservation. Figure 2 displays a planned flow chart for the study. Inclusion criteria for high schools include: (1) counties that partially or fully fall within the Cherokee Nation reservation, (2) town population of 3,000 or fewer, and (3) class size between 30 to 100 students. Exclusion criteria include: (1) metropolitan and micropolitan cores (Census Bureau Rural-Urban Commuting Area codes 1 and 4), and (2) existence of an established community drug prevention coalition. Based on the study inclusion and exclusion criteria, 28 of 60 schools met eligibility criteria. Eligible school/community units were selected for the first phase of recruitment based on geographic separation to reduce spillover risks of the community-level intervention. Twenty-four schools were invited to a recruitment meeting, 22 responded and attended a Zoom meeting with the project team. Twenty of the 22 agreed to participate in the study.
All 10th grade students enrolled in the participating high schools during the fall 2021 semester will be eligible to participate in the study. Based on historical records from the 20 schools, we will recruit approximately 1,000 students. Participating schools will either provide the researchers with a list of the most current home addresses for all students in the grade cohort in the participating high schools or be provided with postage paid consent letters to mail directly from school. Parent/guardian consent letters describe the study and provide a toll-free telephone number that parents/guardians can call with questions or concerns. Home visits will be made when requested or if the parent is unavailable via telephone. Parents who refuse permission for their child to participate will be asked to respond by returning an addressed postage-paid postcard or by calling Cherokee Nation Behavioral Health or Emory University using toll-free telephone numbers. Also, each student will be given an assent form and opportunity to decide not to participate or withdraw their participation at any time.
Constrained randomization
Following recruitment of 20 school-based clusters conducted by the PIs (KAK, TKK, JRS) and project manager (BJL), clusters will be allocated to either the intervention condition or delayed-intervention control condition using constrained randomization by the project statistician (MDL). Constrained randomization helps to ensure balanced cluster sizes as well as similar levels of risk between the intervention and control at baseline [24]. Constrained randomization begins by enumerating all possible randomization schemes (184,756 for 20 clusters with 10 conditions per cluster). Possible randomization schemes are then removed that do not achieve a priori balance conditions, while leaving enough candidate randomization schemes to maintain randomness. Existing guidelines indicate at least 1,000 candidate randomization schemes are sufficient for our study design [25, 26]. Finally, one of these randomization schemes is chosen at random. We constrained possible randomization schemes based on balance criteria for both high school and community characteristics (Table 1), resulting in 1,226 possible randomization schemes of which one was randomly selected. Following randomization, participating schools will be informed as to their status as intervention or delayed-intervention control schools.
Table 1
Variables included in constrained randomization.
Characteristic
|
Balance for Constrained Randomization
|
Source
|
High School
|
|
|
% American Indian
|
Within 5 percentage points
|
National Center for Education Statistics[27]
|
Absenteeism rate
|
Within 5 percentage points
|
Oklahoma School Report Cards[28]
|
% Free/reduced price lunch
|
Within 5 percentage points
|
National Center for Education Statistics[27]
|
Rate of substance use incidents per 100 students
|
Within 0.5 incidents per 100 students
|
Oklahoma Department of Education Annual Incident and Suspension Report[29]
|
Four-year graduation rate
|
Within 5 percentage points
|
Oklahoma School Report Cards[28]
|
Class size
|
Within 5 students or 10%
|
National Center for Education Statistics[27]
|
Community
|
|
|
Medical marijuana dispensary
|
No more than 1 discrepant community across conditions
|
OMMA Dispensary List[30]
|
City police
|
No more than 1 discrepant community across condition
|
Google search
|
Mental health counselors
|
Mean number per condition within 1
|
OK State Board of Behavioral Health-LPC License Search[31]
|
Preventive interventions
We will implement an integrated community intervention that efficiently combines components from evidence-based strategies at the individual, social network, and community levels, with innovations designed to strengthen and enhance effects. The multilevel intervention is designed to reduce “demand” for and “supply” of drugs. Demand reduction strategies include enhancing social connections and support, building self-efficacy to avoid drugs, and supporting no-use norms. Supply reduction strategies include reducing access to drugs and supporting community norms around protecting youth from alcohol and other drugs.
CONNECT. The CONNECT school-based intervention will be administered by the youth services branch of Cherokee Nation Behavioral Health and primarily focus on demand reduction by strengthening self-efficacy, drug-free norms, and social connections and support. Two to three Bachelor’s-level social workers (or closely related degree) will be hired to serve as CONNECT coaches in the 10 intervention schools. Their direct supervisor will be a Cherokee Nation Behavioral Health licensed clinical social worker who is trained and experienced using motivational interviewing (MI), specifically with American Indian youth. In addition, a PhD-level licensed clinical psychologist who is a member of the Motivational Interviewing Network of Trainers (MINT), will provide initial and ongoing training to the coaches. CONNECT coaches will spend one day per week in their assigned high schools. Each semester, the CONNECT coaches will meet individually with all students in the study cohort for approximately 10 to 15 minutes in a private school office or via Zoom. For students habitually absent from school, coaches will schedule sessions via electronic communications (e.g., Zoom, texts, calls, private messages via encrypted social media such as Whatsapp), and conduct face-to-face meetings in private community settings such as Cherokee Nation clinics.
During the CONNECT session, coaches will conduct screening, assess risk, and initiate MI. MI will be used with low-risk students to affirm the student’s decision to avoid drug use and develop personal goals to remain drug-free. For students assessed at moderate risk, MI will be conducted to enhance motivation to avoid riskier situations with access to drugs, and decrease drug use. For high-risk students, coaches will facilitate connection to substance use and mental health treatment facilities, and link students and their guardians to relevant social service resources. The full complement of Cherokee Nation health services are available to any AI youth. Non-Indian youth will be linked with other community services. Additionally, coaches will have a follow-up appointment within one month to monitor each moderate- and high-risk student, adjust goals, normalize setbacks, explore ambivalence about changing relevant behaviors, and provide ongoing support.
CONNECT coaches will implement skill training, Youth Mental Health First Aid (https://www.mentalhealthfirstaid.org/about/), for teachers, other school staff, parents and community members to expand school and community natural helpers and build a network of trusted support for youth. Youth Mental Health First Aid develops skills to identify youth with mental health and substance use concerns, and how to communicate and intervene to provide support.
Communities Mobilizing for Change and Action (CMCA). Ten part-time local community organizers will be hired, one in each intervention community, to focus on reducing the supply of drugs to youth. Other key goals of CMCA are to empower citizens to make change in their communities, enhance social connections, and support drug-free norms. CMCA organizers will be supervised, guided, and supported by team members with expertise in community organizing and opioid and other drug prevention and harm reduction. Community organizers will attend a series of project trainings, including ongoing skill-based training. Following initial training and baseline data collection, each organizer will initiate the CMCA organizing six-stage process in their community. The six-stage process includes (1) assessment of community conditions, norms, and practices; (2) building a citizen-led Action Team; (3) expanding the base, which involves understanding the power structure, finding additional passionate advocates, strategic thinking and planning, constituency building, and building a power base for taking action; (4) developing an action plan; (5) implementing the action plan; and (6) assessing initial results, celebrating accomplishments and refining planned next steps for maintaining effort and institutionalizing changes. Potential actions will focus on reducing easy access of alcohol and other drugs to youth and may include educational and advocacy campaigns, drug take-back initiatives, strategies to reduce social sources, and collaborations with law enforcement and other community institutions to prevent and reduce easy access of drugs to youth. The organizing approach is about empowering citizens, including parents, to influence relevant institutional leaders to take actions for structural and sustainable change. Strategic planning meetings with project leadership and organizers will occur multiple times a year and incorporate additional training for the community organizers.
Media. Media campaigns will support the CONNECT and CMCA interventions and will include: (1) earned media (aka “media advocacy” or generated news coverage); (2) strategically designed, timed, and placed paid media; and (3) social media venues such as Facebook, Twitter, Instagram and lesser known locally used platforms. Media campaigns will be designed for different segments of the community, including family and other social networks important to youth.
Primary and secondary outcomes
Data collection. Variables of interest will be measured with a self-report survey of the cohort of youth ages 15 to 17 at baseline and followed over three years through their transition out of high school (ages 18 to 20). The first survey will be conducted prior to intervention initiation during the fall of the cohort’s 10th grade. Six follow-up surveys will occur, one every six months, beginning in the spring of their 10th grade; follow-ups begin immediately following intervention initiation, and end six months following high school graduation. Youth surveys will be administered via a secure web-based survey using project tablets or personal devices. Location of the self-report survey administration will include home, school, or other private location. Survey data collection will be incentivized and vary by location.
Outcome measures. The survey includes common measures developed in collaboration with the HEAL Prevention Cooperative, ten research projects and a coordinating center funded by NIDA, and were heavily drawn from the PhenX Toolkit substance use patterns – adolescent module, measured with standard items from the Monitoring the Future (MTF) study [32]. Primary outcomes are past 30 days of alcohol, marijuana, and prescription opioid misuse (“without a doctor's prescription or differently than how a doctor or medical provider told you to use it”). Secondary outcomes of key mediators (i.e., mechanisms) include perceived availability of drugs, social support, social normative beliefs, normative estimates, and self-efficacy. In addition to key outcomes, we will include measures of substance use related problems (i.e., missed responsibilities, social problems), comorbidities (i.e., depression, anxiety, pain), and possible intervention effect moderators (e.g., age, gender, race/ethnicity, Native American tribal identity, measures of economic and food security).
Fidelity assessments
Field staff will use laptop computers loaded with research-team-designed management information systems that facilitate their work as well as provide ongoing measures of implementation and progress. Forms include easy-to-use drop-down menus that provide response options for each domain to minimize time burdens associated with completing paperwork and to improve overall efficiency. CONNECT coaches will complete interaction forms following each one-on-one encounter. Coaches will document risk level, behavioral goals, dates of follow-up appointments, and referrals made, if applicable. To assess MI skill among the coaches, during training sessions, coaches will complete mock CONNECT sessions with an actor who portrays a typical student. As in our previous trial, to measure fidelity, mock encounters will be video recorded, coded, and rated using the Motivational Interviewing Treatment Integrity Coding System [33]. We will survey high school teachers to assess confidence, skills, and practices of strategies covered in the Youth Mental Health First Aid training. CMCA organizers will document details of all community contacts, action team members, meetings, actions/events, and proximal organizing outcomes achieved. Data will be uploaded each week and reviewed by the project team and used in strategic planning meetings with leadership and field staff. Finally, we will survey the citizen Action Team members to assess changes in social capital. Additionally, we will assess other unrelated prevention activities as documented by a key school representative on an annual survey.
Cost assessments
In collaboration with the HEAL Prevention Cooperative (HPC) and coordinating center, we will document costs and collaborate on cost-effectiveness analysis. The cost instrument developed with the HPC economic workgroup will be used to capture activity-based costing. Implementation costs will be captured prospectively by including measures of resource use into intervention process evaluation forms. We will collect data on several measures, including: (1) minimum staff to implement the project; (2) local needs (e.g., office space, telephone, local travel); (3) training and technical assistance; and (4) opportunity and marginal costs (e.g. support from volunteers).
Measures
Measures evaluate community and school context, implementation outcomes, and youth outcomes including hypothesized mediators and moderators. Table 2 displays all study measures, and Supplementary File 1 provides additional detail for the youth outcomes measures.
Table 2
Study measures by construct.
Construct
|
Measure
|
Type
|
Informant
|
Timing
|
Context
|
|
|
|
|
School
|
School size, race/ethnic composition, etc
|
Record
|
Record
|
Year 1
|
Community
|
Library, churches, city government, police, etc
|
Record
|
Record
|
Year 1
|
Implementation
|
|
|
|
|
Connect implementation
|
Session attended, risk level, behavioral goals, use of MI
|
Session form
|
Connect coach
|
Year 2–4
|
Connect MI fidelity
|
Motivational Interviewing Treatment Integrity Coding System
|
Videotaped coding of Connect coach MI skills
|
Observer
|
Year 2–4
|
Teacher adoption of practices
|
Confidence, skills, and practices
|
Survey
|
Teachers
|
Year 2–4
|
CMCA implementation
|
Contacts, action team members, meetings, actions/events, and outcomes achieved
|
Access database forms
|
CMCA organizers
|
Year 2–4
|
CMCA implementation
|
Participation, social capital
|
Survey
|
Action Team members
|
Year 2–4
|
Other prevention activities
|
Curriculum, screenings, parent education, community prevention, media, etc
|
Survey
|
School representative
|
Year 2–4
|
Youth Measures
|
|
|
|
|
Demographics
|
Age, gender, race/ethnicity, tribal citizenship, tribal identity, free/reduced price lunch
|
Survey
|
Youth
|
Year 2–5
|
Mediators
|
Social support, perceived availability of drugs, social normative beliefs, self-efficacy, normative estimates
|
Survey
|
Youth
|
Year 2–5
|
Other
|
Depression, anxiety, pain, future aspirations
|
Survey
|
Youth
|
Year 2–5
|
Primary outcomes
|
Past 30 day use of alcohol, marijuana, prescription opioid misuse
|
Survey
|
Youth
|
Year 2–5
|
Data analytic plan
This trial consists of “doubly repeated” measures–repeated measures nested within each student, plus students are nested within schools. To account for both the clustering effect of students nested within schools and repeated measures nested within student, intervention effects on secondary and drug use outcomes will be estimated with generalized linear mixed models specifying a random intercept for each school and allowing within student residuals to be correlated over time. To estimate change in secondary (e.g, perceived drug availability, norms, social support) and primary (e.g., past month number of days of opioid use without a prescription, marijuana use) outcomes due to the intervention, we will begin by analyzing models of the following form: g(y)=β0 + β1TimeL + β2IntG + β3TimeL*IntG where g() represents the appropriate link function for outcome y, TimeL is a linear variable representing the survey wave, and IntG is an indicator for the intervention group. The parameter of primary interest is β3, the difference in slope between the intervention and control group. The link function g() will be chosen to properly model the distribution for each outcome. If it is found that a linear slope does not fit the shape of the intervention curve well, we will fit models of the following form: g(y)=β0 + β1−5TimeFE + β6IntG + β7−11TimeFE*IntG. This model is similar to the previous model except the linear variable for time has been replaced with a series of fixed effects (TimeFE) to allow for a non-linear pattern of change over time. For this model, the parameters β7−11 must be tested jointly using a likelihood ratio test comparing the proposed model and a model where β7−11 are fixed to 0. Attrition will be accounted for in all analyses by full information maximum likelihood [34].
Power
The proposed analytic approach is well-powered to achieve primary study aims. We estimated the statistical power to detect a change in the trajectory over time between intervention and control conditions using PROC GLMPOWER in SAS v9.4. To account for clustering within school and autocorrelation within student, we estimate power for a multivariate model equivalent to our planned “doubly repeated” mixed effects models previously described. Final multivariate correlation matrices are calculated by taking the Kronecker product of a compound symmetric matrix for students nested within schools (ICC = 0.01 to 0.05) and a LEAR matrix for survey waves nested within student (ρ = 0.3, δ = 0.2). The specified LEAR structure allows within student autocorrelation to decay more slowly than the AR(1) structure, which is appropriate for adolescent substance use behaviors [35]. To be conservative in our power estimates we treat time as a fixed effect in all power calculations. Should we determine that linear time trends provide an appropriate fit, power will exceed what is reported below. All power estimates assume a type-1 error rate of 0.05 and 20 balanced schools. While we expect low attrition rates due to planned follow-up procedures, we conservatively assume a 20% attrition rate by specifying the within school sample size as 46 (an average of 57 10th graders per school times 0.8). Our primary and secondary outcomes are continuous scales or count data. We approximate power for both by treating all outcomes as Gaussian. Assuming a linear intervention effect through post-high school follow-up, we are powered at the 0.8 level to detect a standardized mean difference of 0.32 at an ICC of 0.01 and 0.47 at an ICC 0f 0.05 at the end of follow-up.
Monitoring
The Principal Investigator will be responsible for monitoring the safety and efficacy of this trial, executing the Data and Safety Monitoring (DSM) plan, and complying with the reporting requirements to the funding agency and Cherokee Nation IRB. The annual DSM report will include the participants’ sociodemographic characteristics, expected versus actual recruitment rates, any regulatory issues that occurred during the past year, summary of adverse events and serious adverse events, and any actions or changes with respect to the protocol. Reporting guidelines will be followed for any unanticipated adverse events and serious adverse events. The investigators have developed a detailed plan for protection of data, mandatory reporting, and addressing a distressed participant that will be applied during the study. The Cherokee Nation IRB will review and approve all products for dissemination, including presentations and publications.