Feasibility, acceptability, and preliminary effects of the GROW Hopes for Life-Schools intervention on character strengths, resilience, substance use, risky sexual behavior, and school attendance and performance among Zambian adolescents: results of a pilot cluster-randomized controlled trial

Youth substance use preventive interventions are needed which are culturally adaptable, replicable, and designed for individuals with limited literacy. A previous trial of Global Resilience Oral Workshops (GROW), a 6-month spiritually-based character strengths/resilience training curriculum, was associated with increased hope and resilience and decreased alcohol use among Zambian youths ages 10–13. Herein, we describe feasibility, acceptability and preliminary signals of efficacy of GROW Hopes for Life-Schools (GROW-HFLS), a 12-month curriculum for older youths ages 13–17. GROW-HFLS combines a character strengths/resilience component with a substance use prevention component delivered in 36 weekly 90-min after-school sessions by trained community members. Leaders use storytelling and positive psychology exercises to teach 25 character strengths and 12 practices for preventing/overcoming unhealthy substance use. Cultural adaptation was accomplished through iterative consultation with cultural informants, local branding, and integrating community members as interventionists. Feasibility/acceptability were assessed using attendance and retention rates. Efficacy measures were changes in character strengths/virtues (CSV), substance use, risky sexual behaviors, and school attendance/performance, assessed using a multi-site, cluster-randomized, wait-list minimal intervention control design. Eighth-grade students from eight urban and rural secondary schools were cluster-randomized to waitlist control (WLC, n = 226) or intervention (INT, n = 234) arms. CSVs, substance use, and sexual behaviors were assessed using computerized audio-assisted self-report questionnaires at baseline, 6-months and 12-months. School records on student attendance and performance were collected pre- and post-intervention. Median GROW-HFLS attendance was high (34/36 classes) with little decline over time. After 12 months, INT vs. WLC participants demonstrated significantly greater increases in school attendance and performance and in measures of hope, self-regulation, resilience, and perceived self-efficacy to refuse offers of alcohol. After 12-months, past 6-month alcohol use rates were low (3.7% INT, 5.6% WLC) with adjusted relative risk ratios [95% CI] for alcohol use, other substance use, and any sexual partners of 0.57(0.16–2.12), 0.45(0.12–1.69), and 0.43(0.13–1.43), respectively. GROW-HFLS shows promise as a feasible, acceptable positive youth development program associated with improved school functioning and increases in multiple character strengths. Efficacy for substance use and risky sexual behaviors require further study. Training community members as group leaders offers a promising scalable prevention model.

Unhealthy use of alcohol and other drugs continues to be a major barrier to health and wellbeing, particularly among young people. These behaviors, if unaddressed, result in lifelong impact for individuals, their families and local communities. Africa, the world's youngest continent, is home to 475 million young people ages 15–35, with almost 60% of the population under the age of 25. A recent systematic review and meta-analysis of population studies since 2018 documented high levels of alcohol use, high-risk alcohol use and alcohol use disorder (AUD) in numerous countries in sub-Saharan Africa (SSA), including an overall AUD prevalence of 7.7% among adolescents (1). The situation is particularly acute in Zambia, which reported the highest weighted past-6-month prevalence of adolescent alcohol use (80.6%) and the highest prevalence of alcohol dependence (1). A significant proportion of alcohol use in Zambia and other SSA countries involves use of inexpensive unregulated non-industrial alcohol, which generates a high burden of disease and injury (2, 3). Health and social harms related to alcohol use among youths include unsafe sexual behavior, increasing the risk of transmission of HIV and other sexually transmitted illnesses, Fetal Alcohol Syndrome Disorder, injuries, gender-based violence, and mental health disorders (4, 5). The need for interventions, collaborations, and educational programs targeting young people in SSA is described as urgent (6).

While strengthening the prevention of unhealthy substance use is a key component of the United Nations’ (UN) Sustainable Development Goals 2030, as described in SDG 3.5 (7), efforts to address the issue are not prioritized. In many African nations, barriers to effective prevention of unhealthy substance use include significant cultural differences within their populations, lack of trained manpower, limited health system budgets, and limited literacy (7). Many countries in the region have limited infrastructure for prevention and education regarding substance use disorders. A study of secondary schools in Zambia found no consistent policy on substance use education, despite evidence that high numbers of individuals seeking substance use treatment were attending secondary schools (8). As part of the UN's 2030 Agenda for Sustainable Development, the UN has called on member nations to strengthen the prevention and treatment of substance use disorder (7), while the African Union (AU) has called for evidence-based prevention activities in an integrated partnership approach involving the government, private sector, and local communities (9).

Schools are the focus of many attempts to develop effective approaches to substance use prevention (10). A recent systematic review identified four studies in their quest to identify effective school-based alcohol use prevention interventions in Africa (11). Programs found to be effective in Africa were typically multi-component in nature, centered on participatory approaches, and targeted high school students before initiation of alcohol use (11). Furthermore, provision of leisure activities for students during times that they would otherwise engage in substance use (12), and the engagement of trained peer leaders in conducting the interactive activities, were key aspects of the programs (11).

Interventions from the field of positive psychology, which are increasingly being utilized in low and middle income countries, offer a novel approach to substance use prevention that has the potential to prevent or delay unhealthy substance use while also promoting adolescent flourishing and thriving (13, 14). While such an approach may fall under the more general category of positive youth development (i.e., the development of positive assets and strengths), a well-studied and widely-used approach to substance use prevention (14), the application of concepts from the field of positive psychology, which is relatively new, to addressing youth substance use is novel. Moreover, most positive youth development interventions are secular, while religion and spirituality play key roles in many cultures globally. Few studies have explored the synergistic impact of combining spiritually-based interventions with secular positive psychology interventions.

To that end, one of our study team (DS) developed Global Resilience Oral Workshops (GROW), a six-month, after-school, spiritually-based character strengths training curriculum for youth that provides weekly instruction using storytelling and other evidence-based practices to teach each of the 24 character strengths (CS) of positive psychology identified by Peterson and Seligman (15). In 2017, GROW was culturally adapted for the Zambian context (GROW Zambia), with input from our Zambian partners and key stakeholders (e.g., community health workers), and tested in an initial cluster-randomized controlled trial among 543 urban youths ages 10–13 attending 23 primary schools in Lusaka, Zambia (16, 17). Compared to students in the delayed-start control group, the intervention group showed increased hope and resilience and lower alcohol use rates post-intervention. Upon completion of the GROW Zambia trial, focus group discussions with parents, school staff, and GROW Leaders revealed strong support for expanding the program to older adolescents, for whom risk for substance use initiation and related problems is even greater. We subsequently developed the GROW Hopes for Life-Schools (GROW-HFLS), a one-year expanded GROW curriculum designed for older Zambian youth (ages 13–17) which combines CS training with substance use prevention education delivered in 36 weekly 90-min after-school sessions led by trained community members. Herein, we describe findings from a pilot cluster-randomized controlled trial of GROW-HFLS in which we assessed intervention feasibility/acceptability and efficacy for enhancing character strengths, reducing alcohol and related risk behaviors, and improving school functioning among urban and rural Zambian teens ages 13–17.

Study setting and design

The study was conducted in secondary schools in two Zambian districts, namely Lusaka and Sinda. Lusaka District is situated in Lusaka Province and served as the urban setting for the study. Sinda District is in the Eastern Province and served as the rural setting for the study. We used a prospective, cluster-randomized wait-list control group design, with school-level randomization, with participating students attending schools assigned to the control group being offered the program after students in the initial-start schools had completed the program.

Recruitment

School recruitment

We used community mapping to identify potentially eligible secondary schools within our urban and rural target communities. We met with representatives of churches, community-based organizations (CBOs), and school administrators to present the project and its aims of enhancing youth resilience and character strengths, decreasing alcohol and other risk behaviors, and improving school functioning. Our team created a local brand for the program, “Ziyembekezo Za Umoyo,” (English translation: Hopes for Life), to communicate its goal of enhancing hope and wellbeing among program participants. Through this process, we recruited eight secondary schools (four urban and four rural) where weekly after-school group sessions could be offered and which had no participation in positive youth development or substance use prevention interventions in the past three years.

Student recruitment

In each participating school, school headmasters were asked to identify eighth grade classes whose teachers were supportive of their students’ participation in an after-school character/resilience training and substance use prevention program (18). Eighth grade was selected as representative of secondary school students in our target age group since this grade included students from across the age range of concern to community and school leaders (13–17 years). Among the identified classes, eighth-grade students were eligible to participate if they were aged 13–17, attended an information session with a parent/guardian, agreed to complete all confidential study questionnaires and participate in 36 weekly 90-min after-school group sessions, provided written informed assent, and had a parent/guardian who provided written informed consent for their participation. Written informed consents (for parents) and assents (for students) described the project and its aims of enhancing youth resilience and character strengths, decreasing alcohol and other risk behaviors, and improving school functioning. An average of ∼60 students were enrolled per school. Students received a modest honorarium after completing each study questionnaire, in an amount consistent with standard honoraria for such research activities in Zambia.

Power analysis and sample size

We conducted a two-proportion power calculation with the following parameters: hypothesized reduction of alcohol use of 35% (the effect found in the previous GROW Zambia study among 10- to 13-year-olds) (17), power 80%, significance level 0.10, and the proportion of control subjects reporting recent (past 30-day) alcohol use 40% [in accordance with survey data from Zambian adolescents in the Global School-based Health Survey (19)]. The projected minimum total sample size required was 306 (153 per group). A limitation of the sample size calculation is that it did not account for the cluster sampling; specifically, no adjustment was made for potential intracluster correlation or the resulting design effect.

Randomization

Eight schools were cluster-randomized within the urban/rural settings to either an initial-start intervention (INT) group or the waitlist control group (WLC) in a 1:1 ratio (4 clusters in each arm). School randomization was conducted using a coin toss conducted by a team member (SKH) who was not involved in intervention delivery or data collection. There was no intervention masking nor masking of study arm assignment during data collection or analysis.

Ethics approvals

The study was approved by the ethics/institutional review boards of the University of Zambia (UNZABREC), Augusta University, and Boston Children's Hospital. UNZABREC, the primary local ethics committee, approved the study after full board review in accordance with international ethics standards. There was no data and safety monitoring committee for this study. The study was registered March 24, 2022 with the Pan African Clinical Trials Registry (ID# PACTR202203698019870).

Intervention description

GROW Hopes for Life-Schools (GROW-HFLS), delivered in 36 weekly after-school 90-min sessions, is a biopsychosocialspiritual recovery-resilience training program that uses spiritual and other evidence-based positive psychology practices to teach 24 character strengths (15) and 12 life practices that support avoidance of substance use, with the ultimate goal of preventing/reducing youth use of alcohol, tobacco and other drugs. Within the field of positive psychology, character strengths are positively valued personality traits associated with a broad array of important life outcomes, including life satisfaction, academic achievement, work performance, relationships, and health-related behaviors (15, 20). GROW-HFLS is an integrated curriculum which includes two curriculum components: GROW Strong and GROW Free, which are available in multiple languages online (growglobalresilience.com). GROW Strong, adapted from the curriculum tested in the first GROW Zambia project (16), comprises the first 24 sessions and focuses on teaching the 24 Values in Action character strengths (15) which are honored by many cultures across the world (21, 22). Each session is crafted around a Bible story which demonstrates the character strength being taught and highlights a biblical mentor practicing it. Evidence-based positive psychology activities, including song and dance (23), positive self-affirmation (24), drama (25), active constructive response (26), brainstorming (27), goal setting (28), meditation (29), role modeling (30), and gratitude exercises (31), are used to deepen understanding and practice of each character strength and enhance psychological resilience.

GROW Free, taught after completing GROW Strong, is a 12-week group curriculum developed specifically for low and middle income countries (LMIC) through 25 years of fieldwork in Latin America and Africa (15, 32, 33). Using the same positive psychology-based exercises as GROW Strong, GROW Free teaches character strengths and practices that explicitly address alcohol and other substance use prevention/reduction and build a commitment to abstinence (34). In particular, GROW Free focuses on strengthening five specific character strengths that have previously demonstrated associations with lower alcohol use, i.e., forgiveness (35), hope (36), self-regulation (37), spirituality (38), and prudence (39). GROW Free's lessons aim to help participants deepen their relationship with God (40) and other group members, learn media literacy skills and alcohol refusal skills (41), avoid risk-promoting social networks and select non-drinking friends (42), be able to cope with negative affect without using alcohol or other psychoactive substances, help others in their lives using substances achieve sobriety (43), celebrate life without alcohol/drugs, express gratitude daily (44), and practice resilience in the face of a temporary return to alcohol or drug use.

Homework booklets were created for both GROW Strong and GROW Free to encourage participants to accomplish the goal they set during each GROW session, memorize the week's scripture verse, practice the meditation, take three gratitude walks each week, and share what they were learning with someone else.

Each GROW-HFLS group was led by one male and one female group leader. Each school in the intervention arm had two groups of ∼30 students running concurrently, meeting on different days of the week, both led by the same group leader pair.

Cultural vetting and adaptation

The cultural appropriateness of the GROW-HFLS curriculum for the Zambian context was enhanced through an iterative process of review, feedback, and refinement conducted with Zambian collaborators who had previous experience providing alcohol prevention and intervention services to adolescents and young adults. The study welcomed individuals from all faiths and backgrounds. Based on pilot testing in communities that included people from both Christian and Muslim backgrounds, an interfaith curriculum had been developed that primarily utilized Bible stories from scriptures respected by both Christians and Muslims (the five books of Moses, Psalms, and the gospels). The interfaith curriculum was shared and found to be acceptable in this study's community stakeholder meetings which included a broad range of participants including village headmen, community members, church leaders from Zambia's four Christian motherboards, representatives of civil societies, community-based organizations, the Ministry of Education and Ministry of Health. The informed consent and assent described the spiritual aspects of the program including prayer and use of Bible stories. Parents and students were informed of their right to participate in the study or not to participate, their freedom to withdraw from the study at any point without penalty or explanation, and their freedom not to participate in specific activities and not to answer questions that made them uncomfortable. There was no secular or non-religious option offered for those who elected not to participate in this voluntary after-school program. This was not flagged as an ethical concern by any of the study's three ethics committees.

Selection and training of intervention group leaders

The research team collaborated with Zambian pastors, teachers and addiction specialists to invite 54 lay candidates from the community to participate in group leader (GL) training for this trial. Requirements for candidates included having a compassionate character, good interpersonal skills, willingness to pray aloud, a high school diploma, and a letter of character reference from their pastor or another leader from their community. Candidates first received two days of interactive training in GROW Strong, followed by two days of training in GROW Free. Candidates were trained to lead each of the activities delivered in the weekly 90-min group sessions. Senior trainers then chose four top-ranked trainees to serve as GL coaches, and 15 other highly ranked trainees to serve as GLs. Program managers, coaches and GLs received smartphones to facilitate communication and uploading of project documentation and a modest monthly honorarium for their contributions to the program and to pay for wi-fi bundles for project-related communication. Transportation to project training and implementation sites was provided; urban GLs had access to van transport and rural GLs were provided with bicycles. Program developers (JPS, DS) provided weekly internet-based training to coaches, who then provided GLs with weekly in-person coaching.

Fidelity monitoring

A fidelity checklist was created which listed all 15 activities to be included in each GROW session. The fidelity checklist was utilized at each GROW session. As one GL conducted the training, the other GL monitored the session using the checklist, ensuring that all activities were completed as prescribed. Completed checklists were submitted to the program manager each week. Coaches and managers conducted regular field visits at all sites to confirm fidelity.

Safety measures

Research team members, including program managers, coaches and GLs, were trained in a robust Child Protection Protocol during a three-day workshop and were taught how to identify and address signs of psychosocial distress, including referral for assessment and treatment. Program managers and coaches provided ongoing supervision of GLs via site visits. Participants and parents were informed during the consent process of limits to confidentiality in cases of danger to self or others or reports of physical or sexual abuse. Individuals with safety risks were referred by research team members to the project's clinical social worker and received assessments, appropriate intervention and treatment. The safety protocol was implemented with one participant during the study period; no adverse events were noted.

Data collection procedure and measures

Program feasibility/acceptability

We inferred program implementation feasibility/acceptability by examining 1) the degree to which trained program staff were able to be retained; 2) program enrollment among students invited; and 3) program attendance/completion rates among enrolled students. Student attendance at each session was recorded by GLs on paper or computerized attendance registers. If a student missed a session, they could attend a make-up session when available and then be recorded as having completed the session. We examined the median (and interquartile range) number of classes attended during the whole program and, to assess drop-off in attendance over time, separately for the first and second halves of the program. We examined the percentage of intervention participants starting the program (completed ≥1 class), failing to be retained in the program (completed <18 of 36 classes), and completing the program (≥30 classes). We computed these measures for the total sample and stratified by urban/rural setting.

After program completion, we also obtained qualitative feedback about program feasibility/acceptability and impact through focus group sessions with key stakeholder groups (e.g., school administrators, parents/guardians, students, GLs); these qualitative data are reported elsewhere (45).

Efficacy measures

All student participants were invited to complete confidential questionnaires on tablet computers at three data collection timepoints: at baseline (T1), prior to the first cycle of curriculum implementation; at ∼6 months (T2), the midpoint of the curriculum; and at ∼12 months (T3), immediately after the INT group completed the curriculum. All three timepoints included assessment of participant demographic characteristics, character strengths (CS), psychological resilience, alcohol and other substance use behaviors, and sexual risk behaviors (see below for details). All measures were available in English, the primary language of education used in all secondary schools in Zambia, and were also translated into Chinyanja, the primary local language at all project sites, and participants completed the questionnaire in the language of their choice. Small groups of students completed questionnaires individually using electronic tablets in school classrooms with a research assistant present to answer participants’ questions as needed. The questionnaire section with sensitive questions regarding substance use and sexual behaviors was administered with audio assistance and headphones (46). Given the sensitive nature of information gathered, strict confidentiality procedures were observed, including training of research assistants in privacy procedures, use of unique participant ID numbers rather than names, storing of data in secure, password-protected and encrypted computers, and de-identification of data for analysis purposes. School functioning measures were collected from school records for the spring term before the start of GROW-HFLS implementation, and the spring term a year later, after the initial-start group completed the program.

Character strengths/resilience

Character strengths (CS) were assessed using self-report measures developed for this study and previously validated in this Zambian school population, as described in detail elsewhere (46). Our primary CS measures of interest in this study were forgiveness, hope, self-regulation, spirituality, and prudence, due to prior evidence of their associations with substance use (35–40). These brief multi-item scales all used the same 0–10 response scale in which respondents indicated agreement/disagreement with each statement presented (e.g., “Even when things look bad, I stay hopeful”), with 0 meaning “completely disagree” and 10 “completely agree.” Psychological resilience was measured using the 10-item Connor-Davidson Resilience Scale (CD-RISC) (47).

We also examined as exploratory measures students’ level of hopefulness regarding their future expectations for themselves (e.g., their chance of “having a happy family life” on a scale from 0% to 100%) as the curriculum targets optimism, self-efficacy, and resilience which may produce more hopeful future outlooks, and their perceived ability to refuse alcohol when others are drinking which is a target of the GROW Free curriculum. The latter was assessed using a single item created for this study: “Even when others are drinking alcohol, I am able to say no” (0–10 completely agree to completely disagree scale).

Alcohol and other substance use

Alcohol use, and use of other psychoactive substances (tobacco, cannabis, inhalants, and “anything else you have used to get high” including cocaine, hallucinogens, stimulants, opioids, etc.), were assessed using items from the Global School-Based Health Survey (48) and a modified version of the WHO's Alcohol and Substance Involvement Screening Test (ASSIST) instrument (49) adapted to the local context. We examined lifetime, past-6-month, and past-3-month days of any alcohol use and of any other substance use (all other substances were combined due to small cell sizes). Only lifetime and past-3-month use were collected at baseline; past-6-month substance use was not collected at baseline due to an oversight.

Risky sexual behaviors

Risky sexual behaviors were assessed using three items from the World AIDS Foundation HIV Risk Survey regarding number of sexual partners in lifetime, and, during the past 2 months, having sex after having too much to drink or when high on drugs, or in exchange for a place to stay, money, drugs or other favors (transactional sex) (50).

Quality of life

We assessed health-related quality of life (HRQOL) using the 23-item Pediatric Quality of Life Inventory (PedsQL™) (51), which produces four subscale scores (Physical Functioning, Emotional Functioning, Social Functioning, School Functioning) and a Total Score. All items have a 5-point response scale ranging from 0 (Never) to 4 (Almost always). Following the official scoring guidelines, the scale scores were created by reversing and transforming the 0–4 response scores into a 0–100 scale, and then dividing the sum of each scale's items by the number of items answered, with higher scores indicating higher HRQOL (51).

School functioning

Our curriculum content was designed to make weekly sessions fun and engaging, as well as informative and character-building. We hypothesized that program engagement and increases in character strengths would translate into improved school functioning. We examined two school functioning measures during successive spring trimesters pre- and post-intervention (May-July, 2021, and May-July, 2022): 1) school attendance as measured by the percent of school days attended in the term, and 2) school performance as measured by the average score on term final exams in the five core subjects taken by all public secondary school students in Zambia (English, Math, Science, Social Studies, and Religious Studies). School administrators provided these data for participating students from existing school records.

Data analysis

We conducted all analyses using SUDAAN® v.11.0.3 with district and site (and site alone in district-specific analyses) as the nest variables to account for correlated error arising from our cluster-sampling design. We computed descriptive statistics to describe baseline sample characteristics and conducted bivariate analyses, using chi-square tests (for categorical variables) and t -tests (for continuous variables), to assess baseline equivalence of the two study arms. Statistically significant differences were defined by p -value <0.05. Baseline variables that differed significantly between the study arms were controlled for in subsequent outcomes analysis. We employed an “intent-to-treat” approach, analyzing intervention arm participants regardless of whether they completed the intervention. We conducted all analyses on the total sample. We subsequently performed exploratory analyses stratified by urban/rural setting to assess potential moderation by setting. The sub-analyses results are presented in the Supplementary Material.

Alcohol use frequency variables had highly skewed response distributions (right-skewed; e.g., at baseline, only 3.9% reported any alcohol use days and 2.8% any other substance use days in past 3 months); thus, past-3-month and past-6-month alcohol use at each follow-up timepoint were dichotomized into any/no use for outcomes analyses. ASSIST risk category variables were computed for each substance type, with scores <4 indicating low risk and 4 or higher indicating moderate/high risk. However, because of very small cell sizes in the moderate/high risk categories for all substances (e.g., 2.4% for alcohol at 6 months), we chose not to conduct further analysis of the ASSIST variables.

We also dichotomized several demographic variables to support adequate cell sizes for analysis. With whom they lived was collapsed into “lived with a parent/guardian at home vs. none”; number of income earners in their home into 0, 1, >2; number of rooms in their home into 1–2, 3–4, > 5; and usual number of meals eaten per day into 1–2 vs. > 3.

To analyze intervention effects, we conducted two types of adjusted modeling which controlled for group demographic differences and the cluster-sampling design. For our dichotomous substance use outcomes (any/none), we conducted multiple logistic regression modeling with generalized estimating equations (GEE) to compare use probabilities between study arms at each follow-up, with baseline substance use measures entered as a predictor variable in the model. Specifically, for analysis of the intervention effect on past-3-month alcohol use assessed at each follow-up, the multiple logistic regression modeling with GEE controlled for baseline past-3-month alcohol use by including it as a predictor variable in the model. However, analysis of past-6-month substance use outcome measures included as a predictor variable baseline report of use of the substance in their lifetime, as no past-6-month variables were collected at baseline. This method of longitudinal data analysis corresponds to a Markov chain transition model where subsequent observations are conditional on previous observations (52). This type of analysis has the advantage of generating adjusted relative risk ratios (ARRR) and their 95% confidence intervals (95% CI) for the intervention compared to the waitlist control arm at each follow-up timepoint.

We also conducted a repeated measures analysis utilizing data across all three timepoints in generalized linear mixed effects modeling which modeled site- and subject-specific coefficients as random effects and tested a group-by-time interaction effect to examine the intervention effect. The findings from the mixed effects and multiple logistic regression modeling with GEE were no different, so we are presenting the ARRRs from the latter analyses only.

For all other outcome measures, which had continuous data distributions, we conducted generalized linear mixed effects modeling, adjusting for the cluster sampling design (nest variables urban/rural district, school, and participant) and the following covariates which were significantly different between study arms at baseline: age, sex, and usual number of meals eaten per day (<3 vs. 3 or more). We evaluated the intervention effect by testing statistical significance of a group-by-time interaction term in each model. To examine change over time within each study arm, we conducted adjusted pairwise comparisons of scores within each arm across the three data collection timepoints.

We examined potential non-response bias in our follow-up data by comparing baseline characteristics of participants who completed surveys at each follow-up timepoint and those who did not. Overall, compared to those retained at 6-month follow-up, those with missing data tended to be older in age (15.4 ± 1.3 yrs vs. 14.9 ± 1.2 yrs, p =.001), more often males (58.1% vs. 44.4%, p =.021), and more likely to report lifetime alcohol use at baseline (29.1% vs. 19.3%, p =.044). Only age continued to be significantly different between those lost to follow-up and retained at the 12-month timepoint (15.4 ± 1.2 yrs vs. 15.0 ± 1.2 yrs, p =.002). We performed missing data imputation in two ways to assess sensitivity of the findings to non-response bias. We utilized SPSS v. 24 to carry out multiple imputation to produce 5 imputed datasets, using multivariable regression modeling with baseline demographic and substance use variables as predictors. We then re-ran outcomes analyses on the imputed dataset. Intervention effect size estimates were somewhat attenuated compared to those from the original non-imputed data; however, the overall findings about intervention effect significance were not different from the non-imputed analyses, so we present the original results from the non-imputed models.

Sample characteristics

As shown in the study CONSORT flow diagram (Figure 1), a total of 469 students enrolled in the study (240 from urban setting, 229 from rural setting) and completed the baseline questionnaire (239 INT participants, 230 WLC). Nine individuals were subsequently withdrawn from the study after discovering they did not meet study age criteria (5 INT, 4 WLC; 8 urban, 1 rural). The final baseline analysis sample included 460 participants (234 INT and 226 WLC), with school cluster size medians of 58.5 for both INT and WLC schools. Follow-up survey completion rates were not significantly different between the INT and WLC groups at either follow-up timepoint (T2, 81.6% vs. 80.1%, p =.858; T3, 81.2% vs. 85.8%, p =.180), with no difference between urban/rural settings (see setting-specific flow diagrams in Supplementary Material).

The overall sample had a gender distribution of 53% female/47% male, with mean age 15.0 + 1.2 years (Table 1). Two-thirds of participants reported residing with their parents, and nearly all reported living with at least one household member who earned an income (95.9%). Most lived in homes that had three or more rooms (86.9%) and consumed three or more meals per day (84.1%). Of these variables, only the usual number of meals per day differed between study arms at baseline in the overall sample, with WLC participants being more likely to report consuming less than 3 meals per day than INT (22.1% vs. 9.8%, p =.018). Demographic comparisons of study arms stratified by urban/rural setting are presented in Supplementary Table 1.

Program feasibility/acceptability

Program staff retention

Among 18 program staff (two coaches and 16 group leaders) recruited and trained prior to starting the project, five individuals (27.2%)—one coach and 4 group leaders—withdrew during the initial 1–2 months of the program. All five who withdrew were full-time school employees (1 administrator, 4 urban teachers), who noted their heavy job responsibilities made it too difficult for them to continue with the project. All five were quickly replaced by other qualified candidates from among the initial pool of program trainees and the curriculum was able to be delivered in its entirety at all INT schools.

Training of schoolteachers to be a GL, as a step toward longer-term sustainability of the program in participating schools, was successful in the rural setting, where teacher participation was 100%, but not in the urban setting, where all teachers withdrew from the program.

Program enrollment and attendance

Parents of approximately 50% of age-eligible eighth-grade students in chosen classes in the participating schools (∼50–65 students per school) attended recruitment meetings. All of those in attendance agreed to participate, providing written parent/guardian consent and student assent (per school median, IQR: 58.5, 57.5–59.5) (Figure 1).

Ninety-eight percent (230/234) of students in the INT study arm attended at least one session. Among those attending any sessions, 70% (160/230) completed the program by attending 30 or more sessions. Median sessions attended was high at 34 of 36 [interquartile range (IQR) 26–36], with no difference in median sessions attended in the first half of the program (17, IQR: 14–18) vs. the second half (17, IQR: 13–18). There were 37 students (16%) who started the curriculum but then completed fewer than 18 sessions (half the program) due to relocation, failure to pass exams for promotion to the 9th grade, or some other reason.

There were differences in attendance levels between the urban/rural settings (Supplementary Table 2), with the rural INT group having a significantly higher median number of sessions attended (35, IQR: 30–36) compared to the urban INT group (32, IQR 26–35) (p <.001), and a higher program completion rate (75.4% vs. 61.5%, p =.022). However, rural INT participants had a higher rate of stopping early (i.e., completing less than half the program) compared to those in the urban setting (22.0% vs. 8.5%, p =.004).

Outcome measures

Character strengths, resilience, and perceived ability to refuse alcohol

As shown in Figures 2–5, students in the INT group demonstrated significant increases in their adjusted mean scores in pre-post comparisons for all five primary character strengths of interest (forgiveness, hope, prudence, self-regulation and spirituality), as well as for psychological resilience, hopeful future expectations, and perceived ability to refuse alcohol when others are drinking. The WLC group showed significant pre-post increases in hopeful future expectations and psychological resilience only. Group-by-time interaction effects, which compare trends over time between the INT and WLC groups, were significant for hope, self-regulation, and resilience, as well as perceived ability to refuse alcohol when others are drinking.

We explored whether intervention effects on these measures were moderated by urban/rural setting by conducing analyses within setting subgroups (see Supplementary Table 3). Stronger intervention effects were seen in the urban setting, with significant group-by-time interaction effects (p <.05) in the urban setting for hope, hopeful future expectations, psychological resilience, and self-regulation. There were no significant group-by-time interaction effects in the rural setting, likely due to the WLC group often showing increases in scores over time as well, albeit more modest than in the INT group.

Substance use and sexual risk behaviors

Alcohol use

In the overall sample, baseline rates of any lifetime alcohol use were 24.8% in the INT group and 17.3% in the WLC, which was not a statistically significant difference in adjusted analyses (ARRR, 95% CI: 1.50, 0.96–2.36) (Figure 6). These numbers were lower than the 42.6% previously reported by Swahn et al. (19). Numbers of participants reporting any past-6-month and past-3-month alcohol consumption during the follow-up period were low in both groups at both T2 (past-6-month use = 19 INT and 10 WLC; past-3-month use = 5 INT and 2 WLC) and T3 (past-6-month use = n = 7 INT and 11 WLC; past-3-month use = 4 INT and 8 WLC), reducing statistical power to detect an intervention effect. Adjusted relative risk ratios comparing INT and WLC groups at T3 (post-intervention) for past-6-month and past-3-month alcohol use were promising (ARRR 0.57 and 0.42, respectively); however, the 95% confidence intervals were wide and included 1.0, likely due to small cell sizes (Figures 6A,B).

The past-6-month alcohol use rate fell from 9.9% to 3.7% in the INT group from 6 to 12 months follow-up, while the rate showed little change (5.5% to 5.7%) during the same period in the WLC group (Figure 6A). The past-6-month use measure was not collected at baseline.

Results of alcohol outcomes analysis stratified by urban/rural setting are presented in Supplementary Table 4. These subgroup results should be viewed with caution due to very small cell sizes.

Other substance use

Baseline rates of any lifetime use of psychoactive substances other than alcohol were lower than for alcohol: 8.1% in the INT and 11.5% in the WLC groups (Figure 7). Only past-6-month use of other substances reported at follow-up are presented in Figure 7 due to near-zero values for past-3-month use. The numbers of participants reporting past-6-month use of other substances (<10 per group) were even lower than for alcohol. Similar to alcohol, the adjusted relative risk ratio comparing INT to WLC at the end of the intervention period (T3) was promising (0.45) but had a wide confidence interval (95% CI: 0.12–1.69). In addition, there was a drop in the past-6-month use rate seen in the INT group (from 4.7% to 2.1%) that was not observed in the WLC (2.7%–5.2%).

Risky sexual behaviors

The numbers of participants reporting any sex after substance use or any transactional sex in the past 2 months were too small for analysis (cell sizes were 0 or 1), thus are not reported further. For number of sex partners, because of very small numbers reporting more than one sex partner during follow-up, we analyzed this measure as a dichotomous outcome variable (any/none). At both follow-ups, INT participants tended to report lower rates of any past-6-month sexual partners compared to WLC, even after adjusting for baseline sexual experience (Figure 8). Consistent with the alcohol and other substance use outcomes, the ARRR at T3 was promising (0.43), but had a wide confidence interval (0.13–1.43). Interestingly, subgroup analysis by setting did, however, reveal a significant intervention effect in the rural setting at 12 months (5.9% INT vs. 23.5% WLC, ARRR: 0.36, CI: 0.13–0.97) (Supplementary Table 4).

Quality of life

We found significant increases in PedsQL subscale and Total scores over the 12-month study period in both the INT and WLC group (Table 2). In adjusted analyses, the INT group showed significantly higher scores at both 6- and 12-month follow-ups compared to baseline for all 4 subscales and the Total Score, while the WLC group had significant increases in all but the Social Functioning scale. With HRQOL scores increasing over time in both study arms, there were no significant group by time interaction effects.

In analysis stratified by urban/rural setting, we found significant group by time interaction effects in the urban sample only, with significantly greater increases in Physical Functioning, Social Functioning, and Total Scores by 12 months follow-up in the urban INT group compared to WLC (Supplementary Table 5).

School functioning

School attendance

We analyzed data for 416 participants who had school records data available (200 INT and 216 WLC). As shown in Figure 9A, there was a significant intervention effect on school attendance, with a pre-post increase from 94.8% to 97.0% in the INT group, while the rate declined from 94.3% to 90.8% in the WLC group (group-by-time interaction term p <.001).

School performance

There was also a borderline significant intervention effect (group-by-time interaction term p =.05) for average end-of-term exam scores (Figure 9B). INT participants increased their average score on final exams pre-post from 49.0% to 52.4%, while the WLC group had a smaller increase from 40.4% to 41.8%. Surprisingly, in urban/rural subgroup analyses, the significant effects were seen only in the rural setting (Supplementary Table 6).

This study's findings suggest that GROW-HFLS shows promise as a feasible and acceptable positive youth development program that could impact a broad range of youth character, educational and behavioral outcomes. Key findings of this study are the greater increases in psychological assets, along with relative improvement of pre-post school attendance rates and exam scores, among youth receiving GROW-HFLS compared to those in the waitlist control group. Participants in the intervention group showed significant pre-post increases for five key character strengths previously shown to correlate with lower alcohol use: forgiveness, hope, prudence, self-regulation, and spirituality. Interestingly, the waitlist control group also showed increases over time in hopeful future expectations and psychological resilience. When comparing trends over time between the two groups, intervention participants demonstrated significantly greater increases in hope, self-regulation, and psychological resilience. These internal changes were accompanied by changes in student behavior, with improvements in school attendance and gains in academic performance in the GROW-HFLS group compared to the control group. It is noteworthy that, while school attendance levels in the control group declined over the 12-month period, they increased in the intervention group. Such tangible academic benefits are compelling, as education is a key determinant of future health, earning potential, and overall well-being (53–56). We did not find a group difference in trends over time on the School Functioning measure of the Pediatric Quality of Life Inventory (PedsQL), however. This may be due to the PedQL items’ focus, in part, on health-related barriers to school such as missing school due to feeling unwell or needing to go to the doctor/hospital (51).

These results expand upon the findings of the previous GROW pilot study (16), which documented increased hope and resilience in a younger student population (ages 10–13). We hypothesize that the longer one-year program, which reinforced character strengths in both the resilience (GROW Strong) and substance use-focused (GROW Free) portions of the curriculum, may have assisted in cultivating a broader range of character strengths and skills than the previous shorter program. This is consistent with previous research suggesting greater impact of youth development and substance use prevention programs that are longer (57).

This study found preliminary support for program implementation feasibility and acceptability within a relatively low-resource setting, with strong staff retention and high student participation and completion rates. Early modest staff attrition due to competing time demands was easily managed by the staff training model, which recruited and trained more staff than was needed for the project, creating a “replacement pool” which was effectively utilized to complete curriculum implementation. However, training of schoolteachers as group leaders, as a strategy to support program sustainability in schools, was successful only in the rural setting, not the urban setting, highlighting the need for further research to identify implementation facilitators/barriers across various school settings. The program's positive effects on participants’ school performance and attendance, if confirmed in future studies, may accelerate the program's dissemination by school administrators and officials within ministries of education.

In terms of behavioral outcomes, this pilot study may have been underpowered to detect significant intervention effects on self-reported substance use and sexual behavior due to the low prevalence of these behaviors in both study arms. This may be due in part to the fact that youths who are more vulnerable and higher risk often drop out of school after Grade 6 in Zambia. Findings may nonetheless provide a preliminary signal of positive change in these behaviors. There were declining trends in alcohol and other substance use, as well as in sexual behavior, in the intervention group between the 6 and 12-month follow-ups, a trend not observed in the control group. This finding is particularly noteworthy as these risk behaviors typically increase during this developmental period (58). Critically, the intervention enhanced participants’ self-perceived ability to refuse alcohol when offered by others, a key target of other evidence-based substance use prevention programs (59).

It is noteworthy that, in the rural setting, the intervention effect on rates of any sex partner during follow-up was significant. This finding suggests a potential role for GROW-HFLS in helping to accelerate progress in addressing HIV and other sexually transmitted infections in Zambia, where an aggressive public health response has already reduced HIV prevalence to 8.0% among men and 13.9% among women (60). Reasons for the larger impact of this program on both sexual behavior and school performance among rural participants are unclear, given that student attendance and retention were similar in both settings. One possible explanation is that in rural settings, resources are often more limited, and an integrated community-based program such as GROW-HFLS may have a greater impact on student motivation and behavior than in a busy urban environment. On the other hand, we found significant intervention effects on self-reported Physical and Social health-related quality of life scale scores in the urban sample and not the rural. These setting-specific findings need confirmation in larger studies.

Of particular salience to lower-resource settings such as in sub-Saharan Africa is this study's demonstration of a feasible, scalable multi-sectorial implementation model of an evidence-based positive youth development program. Program implementation in Zambia combined the efforts of community leaders, schools, local churches, a local NGO, and one of Zambia's premiere universities. The GROW-HFLS program was delivered by trained community members rather than credentialed professionals, addressing a common barrier in many low- and middle-income countries where trained manpower is limited. This model aligns with the African Union's call for integrated, community-involved prevention activities and offers a replicable, sustainable approach to adolescent health promotion (9). The program's success also reinforces findings from earlier effective African prevention programs which included multiple components, were highly participatory and targeted students before initiation of alcohol use (11, 57).

Limitations

This study has several limitations. The lower-than-expected rates of alcohol use and other risk behaviors may have been due to underreporting due to social desirability bias or distrust that the information would be kept confidential. In addition, participants were informed during the informed consent process of the project's goal of enhancing resilience and character and decreasing or delaying alcohol use. This could have contributed to performance and reporting bias regarding self-reported outcomes of character strengths and of substance use. Since the study's power analysis did not account for clustering of participants within schools and districts, the sample size may have lacked sufficient power to assess some intervention effects. Self-reported substance use was not confirmed with biomarkers. Assessment of program feasibility/acceptability did not include standardized quantitative acceptability measures completed by intervention participants. The study did not assess potential unintended or iatrogenic effects such as increased curiosity, more favorable expectancies about alcohol or other drugs, or initiation among baseline non-users. The sample had limited religious diversity, therefore results of this study may not be generalizable to settings dominated by other religious traditions or populations with significant numbers of people who are not Christians. Those lost to follow-up in both the intervention and control groups were more likely males, older, with higher rates of lifetime alcohol use at baseline, which may have resulted in underestimates of actual substance use rates for this age group of Zambian youths attending secondary school. The study did not include follow-up data after the 12-month intervention period to assess the long-term sustainability of the observed effects. Future research should aim for larger sample sizes, include more intensive follow-up to minimize dropouts and longer-term follow-up, utilize biomarkers to confirm changes in substance use, utilize standardized quantitative assessments of program feasibility/acceptability, seek to recruit a student population with greater religious/spiritual diversity, and consider offering a secular or non-religious intervention option as a way of broadening the study's inclusivity and generalizability. Other effective program components such as inclusion of peer mentors and booster sessions, which were part of the original GROW intervention (16) and have proven to be efficacious in other African studies (11), might also be integrated into such future studies.

The GROW Hopes for Life-Schools (GROW-HFLS) intervention shows promise as an effective, multi-component program for promoting positive youth development among Zambian adolescents. Initial signals of intervention efficacy for this 12-month program were seen in strengthened internal psychological assets, including hope, self-regulation, and psychological resilience; improved self-perceived efficacy to refuse alcohol; and improved school attendance and academic performance, which are linked to success in later life.

Overall, GROW-HFLS presents a scalable, culturally adaptable program with potential utility among global populations with limited resources. By utilizing trained community members and delivering a curriculum that enhances character, well-being, and educational success, this intervention offers a practical and holistic approach to addressing the urgent need for adolescent health promotion in Sub-Saharan Africa and other similar settings.