Effect of Shamiri Layperson-Provided Intervention vs Study Skills Control Intervention for Depression and Anxiety Symptoms in Adolescents in Kenya: A Randomized Clinical Trial

Key Points

Question  Can scalable psychological interventions that invoke simple psychological principles, rather than explicit references to psychopathology, alleviate depression and anxiety symptoms in Kenyan adolescents?

Findings  In this randomized clinical trial of 413 high school students, Shamiri, a 4-week layperson-delivered group intervention that teaches growth mindset, gratitude, and value affirmation, appears to significantly reduce depression and anxiety symptoms in symptomatic adolescents compared with a control group, although both groups showed symptom reductions. Effects persisted at 7 months.

Meaning  Simple psychological interventions that focus on positive human attributes rather than psychopathology, are delivered by laypersons, and are developed through multicultural collaboration may reduce depression and anxiety symptoms and should be considered for use in low-resource settings.

Importance  Low-cost interventions for adolescent depression and anxiety are needed in low-resource countries such as those in Sub-Saharan Africa.

Objective  To assess whether Shamiri, a 4-week layperson-delivered group intervention that teaches growth mindset, gratitude, and value affirmation, can alleviate depression and anxiety symptoms in symptomatic Kenyan adolescents.

Design, Setting, and Participants  This school-based randomized clinical trial included outcomes assessed at baseline, posttreatment, and 2-week and 7-month follow-up from 4 secondary schools in Nairobi and Kiambu County, Kenya. Adolescents aged 13 to 18 years with elevated symptoms on standardized depression or anxiety measures were eligible. Intent-to-treat analyses were used to analyze effects. Recruitment took place in June 2019; follow-up data were collected in August 2019 and February 2020.

Intervention  Adolescents were randomized to the Shamiri intervention or to a study skills control. All adolescents in both conditions met in groups (mean group size, 9) for 60 minutes per week for 4 weeks.

Main Outcomes and Measures  Primary outcomes were depression (Patient Health Questionnaire–8 item) and anxiety (Generalized Anxiety Disorder–7 item) symptoms. Analyses of imputed data were hypothesized to reveal significant reductions in depression and anxiety symptoms for adolescents assigned to Shamiri compared with those in the study skills group.

Results  Of 413 adolescents, 205 (49.6%) were randomized to Shamiri and 208 (50.4%) to study skills. The mean (SD) age was 15.5 (1.2) years, and 268 (65.21%) were female. A total of 307 youths completed the 4-week intervention. Both Shamiri and study skills were rated highly useful (4.8/5.0) and reduced symptoms of depression and anxiety, but analyses with imputed data revealed that youths receiving Shamiri showed greater reductions in depressive symptoms at posttreatment (Cohen d = 0.35 [95% CI, 0.09-0.60]), 2-week follow-up (Cohen d = 0.28 [95% CI, 0.04-0.54]), and 7-month follow-up (Cohen d = 0.45 [95% CI, 0.19-0.71]) and greater reductions in anxiety symptoms at posttreatment (Cohen d = 0.37 [95% CI, 0.11-0.63]), 2-week follow-up (Cohen d = 0.26 [95% CI, −0.01 to 0.53]), and 7-month follow-up (Cohen d = 0.44 [95% CI, 0.18-0.71]).

Conclusions and Relevance  Both the Shamiri intervention and a study skills control group reduced depression and anxiety symptoms; the low-cost Shamiri intervention had a greater effect, with effects lasting at least 7 months. If attrition is reduced and the clinical significance of outcome differences is established, this kind of intervention may prove useful in other global settings where there are limited resources, mental illness stigma, or a shortage of professionals and limited access to mental health care.

Trial Registration  Pan-African Clinical Trials Registry Identifier: PACTR201906525818462

Adolescent depression and anxiety account for 45% of the global burden of disease in youths aged 10 to 24 years,¹ based on worldwide estimates per disability-adjusted life-years, and are linked to important medical and life outcomes.^2,3 Both are prevalent in Sub-Saharan Africa,⁴ but youths with these problems in Sub-Saharan Africa may not find help. Few mental health professionals exist,^5,6 and social stigma surrounding mental illness prevents help seeking.⁷ Attempts to improve the psychological well-being of youths in Sub-Saharan Africa have had mixed success.^8,9 As the population of Sub-Saharan Africa becomes increasingly youthful,^10,11 developing interventions to overcome these barriers is a global health priority.¹²

One approach to expanding access to care may involve using brief theory-driven treatments that are sometimes called wise interventions.^13,14 Wise interventions differ from most evidence-based treatments in that they (1) are often delivered by laypersons rather than trained clinicians, an approach the World Health Organization recommends for low-resource settings,^15,16 (2) focus on simple psychological concepts rather than behavioral and cognitive skills, (3) are often delivered in nonclinical settings, and (4) invoke positive human attributes and principles rather than psychopathology. One example is growth-mindset intervention, which conveys the concept that one’s traits and characteristics are malleable and can be improved via effort^17-19 and has been shown in US samples to reduce symptoms of depression and anxiety and to improve functioning.^18-21 Similar effects have been found in the US for gratitude interventions, which teach individuals to notice and appreciate good things in their lives and express their gratitude.²² A third form of wise intervention is value affirmation: encouraging individuals to identify and reflect on their self-defining values and plan actions consistent with those values.²³

Given the apparent clinical, conceptual, and practical advantages of these 3 simple interventions, our multicultural Kenya-US team brought elements of the 3 together to form a combined intervention for anxiety and depression symptoms in Kenyan youths. This intervention, Shamiri (Kiswahili for thrive), is designed for implementation with adolescents meeting in groups led by trained laypersons. A preliminary proof-of-concept trial tested Shamiri with 51 Kenyan youths (aged 14-17 years) who showed elevated anxiety and/or depression symptoms.¹⁹ Youths randomly assigned to Shamiri (n = 28) showed significantly greater reductions in depression and anxiety symptoms than youths randomized to a study skills control (n = 23).²⁴ That study was weakened by small sample size and limited follow-up.

Here, we built on that initial pilot by conducting a trial with a large sample and follow-up extended to 7 months. We predicted that symptomatic adolescents randomized to Shamiri would experience greater reductions in depression and anxiety symptoms than those assigned to the study skills control. We chose a study skills control intervention because such control conditions provide a more rigorous comparison than passive controls (eg, waiting list).^25,26

This single-blind, parallel-group randomized clinical trial took place in 4 Kenyan public secondary schools in Nairobi and Kiambu County that differed markedly on several characteristics, providing sample diversity. The study protocol is published elsewhere²⁷ and is available in Supplement 1. The characteristics included sex, location, and tribe (eAppendix 1 in Supplement 2). School A (all boys) and school B (all girls) had a student body drawn from all counties in Kenya. School C (all girls) and school D (boys and girls) were day schools serving primarily local students.

All students aged 13 to 18 years were eligible if they had elevated scores on baseline depression or anxiety symptom measures (below). The study design, which was approved by the Maseno University Ethics Review Committee and the National Commission for Science, Technology, and Innovation, was consistent with the requirements of the Declaration of Helsinki.²⁸ Parental consent was sought for adolescent minors (age <18 years) per research ethics procedures approved by the Maseno University Ethics Review Committee. All students provided written informed consent or assent before screening. This study was preregistered prior to participant enrollment in the Pan-African Clinical Trials Registry (PACTR201906525818462).

From June 20, 2019, through July 5, 2019, 2192 students completed baseline questionnaires in their classrooms. In all 4 schools, administrators required that all willing students participate in screening: 715 of 2192 youths assessed for eligibility met the inclusion criteria. In 2 schools, more students met criteria than could be accommodated and we did not know a priori how may would meet cutoffs (eAppendix 2 in Supplement 2). As a result, we randomly selected 180 of 211 students at school A and 192 of 316 students at school B. In total, 560 total students were invited across the 4 schools (schools differed in population). We used a random assignment code generated in R version 4.0.3 (R Foundation), stratified by sex and form (school grade). During the baseline screening, all students were invited to contact the study team if they had questions or needed support, and a tiered system of support was made available (eAppendix 2 in Supplement 2). Staff estimated that 40 students contacted the study team with questions and concerns.

Lay group leaders (aged 18-26 years; 8 of 13 female [61.54%]) for the Shamiri and study skills groups were Kenyan high school graduates fluent in English and Kiswahili who were hired and trained by the study team. They were selected via a semistructured interview that gauged past experiences and interpersonal skills; they underwent 10 hours of training (trial protocol in Supplement 1 and previously published²⁹), covering content of Shamiri and study skills interventions, and including counseling techniques, role play of intervention and control content, and safety protocols. Leaders were randomly assigned to groups; all leaders led both Shamiri and study skills groups. Because all leaders delivered Shamiri and study skills content, the study team led weekly supervision meetings for both conditions (eAppendix 3 in Supplement 2).

Groups included 7 to 15 youths (mean [SD] group size, 9 [2.3]). Each group had 4 weekly meetings lasting 1 hour, which were conducted in a combination of English and Kiswahili because all participants spoke both languages. Groups met in the schools during the afternoon time designated for extracurriculars. Group sessions were conducted between July 1, 2019, and August 2, 2019, with 7-month outcome assessment on July 28, 2020.

The Shamiri intervention included 3 modules: growth mindset (2 sessions), gratitude (1 session), and value affirmation (1 session). In session 1, group leaders explained growth and personal improvement, and participants learned about neuroplasticity through a video and article and heard testimonials about personal growth. They wrote stories about their experiences of personal growth and about how they could use growth to solve a specific problem. In session 2, participants discussed strategies for overcoming challenges, learned a framework for problem solving, wrote a letter to a friend explaining what they had learned, and wrote about applying problem solving to a specific and current life challenge. In session 3, students discussed the importance of gratitude, then wrote a gratitude letter to someone. Participants also listed 3 things for which they were grateful on each day of the following week. In session 4, participants learned about values (called virtues), selected (from a list) values that were important to them, wrote about a time when they had demonstrated one of their values, considered how they might live one of their values in the future, and planned a specific action to promote that value. eAppendix 4 in Supplement 2 contains intervention materials.

The study skills control condition was developed, in collaboration with Kenyan educational and research experts, to help students address the intense academic expectations of secondary school, thought to be stressful and potentially a source of depression and anxiety symptoms. The skills included note-taking, effective study strategies, time management, and the study cycle. To control for nonspecific aspects of the intervention, the study skills and Shamiri intervention sessions had the same structure (ie, didactics, exercises, discussions, and homework) and equal duration. See eAppendix 2 in Supplement 2 for study skills control development and eAppendix 4 in Supplement 2 for intervention materials.

Intervention fidelity was assessed using session audio recordings, via a rubric (eAppendix 5 in Supplement 2) to evaluate layperson adherence to and competence in protocol components. Two independent raters, who were not affiliated with the study team, independently rated a randomly selected 10% of sessions. Interrater agreement was assessed using Gwet AC2 for ordinal ratings using the rel package in R version 4.0.3 (R Foundation).³⁰ Mean ratings (maximum, 7.00) and AC2s (with 1.0 indicating perfect reliability) for each domain, all in the acceptable range, were 6.65 (AC2 = 0.84) for delivering protocol content, 6.76 (AC2 = 0.87) for completing specific tasks (eg, distributing worksheets), 6.18 (AC2 = 0.95) for thoroughness, 6.03 (AC2 = 0.95) for skillfulness, 6.26 (AC2 = 0.96) for clarity, and 6.11 (AC2 = 0.84) for including content exclusively from the relevant session protocol.

Outcome measures were administered at baseline, midpoint (after 2 weeks), end point (after 4 weeks), and at 2-week and 7-month follow-up. Prior to and during the study, it appeared that national education regulations might ban research activity beyond 2 weeks postintervention, so we did not preregister a later assessment on the Pan-African Clinical Trials Registry. The ban occurred but was eventually lifted, permitting a 7-month follow-up assessment with all the study participants whom we could locate (n = 118 for Shamiri; n = 105 for study skills).

The primary outcome was change in depression and anxiety symptoms from baseline to the 4-week end point. To qualify for the study, participants had to self-report elevated symptoms of anxiety or depression. We assessed depression symptoms with the Patient Health Questionnaire–8 item (PHQ-8); the 8-item version of the PHQ-9.³¹ Per the study protocol,²⁷ we used the PHQ-8 because school administrators believed the societal stigma associated with the suicidal ideation item might upset or alienate students. PHQ-8 scores are highly correlated with PHQ-9 scores and have the same cutoffs for depression severity.³² Scores range from 0 to 24 (most severe symptoms). Psychometrics of the PHQ have been reported for adolescent and adult community samples in Kenya,^33,34 and the measure has been used in recent trials with Kenyan youths.^24,35 Cronbach α was 0.78 in the present sample. We used cutoff norms from PHQ studies with adolescents in the US³¹ and in Kenya,^24,33 requiring a score of 15 or higher (moderately severe to severe depression symptoms).³¹

We assessed anxiety symptoms with the Generalized Anxiety Disorder–7 item (GAD-7).³⁶ GAD-7 scores range from 0 to 21 (most severe symptoms). The GAD-7 has been used and has met psychometric standards in prior studies with Kenyan adolescents.^33,35 Cronbach α was 0.82 in the present sample. We used cutoffs from GAD-7 studies with adolescents in the US³⁶ and in Kenya,^24,33 requiring a score 10 or higher (moderate to severe anxiety symptoms).³⁶

At the end of the last session, participants completed a program evaluation survey. On a scale of 1 to 5, participants rated “How helpful was the program as a whole to you?” and “How likely would you be to recommend the program to a friend?”

Per study protocol,²⁷ we conducted a power analysis using Optimal Design³⁷ to estimate sample size required for power of 0.80 to detect an effect of d = 0.30 with α = .05 and 5 measurement points: we arrived at 200 participants per condition, assuming 10% attrition. We invited 560 participants to participate, assuming that approximately 30% of those eligible would not be available for or wish to participate.²⁷

Our data had 4 levels of clustering: school (level 4), group leader (level 3), groups (level 2), and participant (level 1). Because a group leader could lead multiple groups (groups in each school met on different days), they were cross-classified in an unstructured manner. Groups were strictly nested within group leaders as were participants within groups (structured cross-classification). See eAppendix 6 in Supplement 2.

To address attrition, we conducted multiple imputations of the primary outcome measures (ie, depression and anxiety symptoms) using the mice package³⁸ in R version 4.0.3 (R Foundation) under the assumption that data were missing at random. We conducted sensitivity analyses, using the δ method,³⁹ imputing under the assumption that data were missing not at random. The results held up for situations with data missing not at random in which we overestimated the missing values (ie, assigned higher values to missing data). For situations with data missing not at random in which imputations underestimated missing values, our results held up qualitatively, although some of the treatment differences were no longer significant (eAppendix 6 in Supplement 2).

Given the hierarchical nature of our data, we ran random-intercept linear mixed-effects models with the lme4 package⁴⁰ in Rversion 4.0.3 (R Foundation) with crossed random effects for schools and group leaders and strict hierarchical nesting for observations in participants, participants in groups, and groups in group leaders. This addressed dependencies that may have arisen from the nested design and adjusted standard errors accordingly. We operationalized our primary question of interest: differences in symptoms between the 2 conditions (with Shamiri as reference) at different time points as a time × condition interaction. Time was coded as a categorical variable with the 4-week end point acting as the reference point; the main effect of condition was estimated as the difference between Shamiri vs study skills control at end point. Because the variables sex and form (school grade) were used for stratified randomization, we included them as covariates in our model, thus addressing any dependencies that may have arisen during randomization.⁴¹ We plotted the primary effects of condition × time (as marginal means) and conducted pairwise comparisons for the effects of condition at the end point and at 2-week and 7-month follow-up. We used inference provided by the package lmerTest.⁴² Analyses were done on 7 imputed data sets and we pooled results. Details of the statistical analyses, plus underlying R code and explanations, are in eAppendix 6 in Supplement 2.

Because of rather high attrition, we ran 2 models for each primary outcome. First, per our protocol, we used an intent-to-treat approach including all randomized participants in our analyses. Missing data were imputed 7 times as described earlier. Second, we ran models for actual observations (ie, unimputed). Because estimation of variance of the random effects hinges on the number of groups at each nesting level and because our highest nesting level (school) included only 4 groups, we conducted sensitivity analyses by building models with school introduced as a covariate. We further extended the random intercept models to random slope models as a sensitivity check. These sensitivity analyses and their results are described in eAppendix 6 in Supplement 2. Results from our random intercept models held up for these sensitivity models. We used mean gain scores (Cohen d) pooled from the imputations to calculate effect sizes for changes in depressive and anxiety symptoms. Finally, clinically reliable change thresholds were calculated using s√(1 − r) × 1.96,^43,44 using SDs at baseline and reliability (r) from psychometric publications for the PHQ-8 and GAD-7 administered to Kenyan adolescents.

In total, 205 adolescents (mean [SD] age, 15.4 [1.2] years) were randomized to Shamiri and 208 adolescents (mean [SD] age, 15.7 [1.2] years) were randomized to study skills (Figure 1). Baseline characteristics of the 413 participants (268 female [65.21%]) are in the Table (eAppendix 7 in Supplement 2 shows characteristics of all adolescents assessed at baseline). No significant group differences emerged on baseline demographic or symptom measures. Of 413 youths, 310 (75.06%) completed midpoint assessment, 307 (74.33%) completed end point assessment, 244 (59.07%) completed 2-week follow-up assessment, and 223 (54.00%) completed 7-month follow-up assessment (Figure 1).

The model using imputed values revealed significant differences in depression symptoms between Shamiri and study skills control at end point (B = 1.11 [95% CI, 0.15-2.07]; t = 2.27; df = 1282.54; P = .02). The model using only unimputed values revealed similar significant differences in depression symptoms (B = 1.24 [95% CI, 0.16-2.31]; t = 2.26; df = 761.36; P = .02). Similarly, youths in Shamiri experienced greater declines in depression symptoms from baseline to end point than control group youths (Cohen d = 0.35 [95% CI, 0.09-0.60]; Figure 2 and eAppendix 8 in Supplement 2).

The model using imputed values revealed significant differences in anxiety symptoms between Shamiri and study skills control at end point (B = 1.37 [95% CI, 0.35-2.39]; t = 2.63; df = 109.07; P = .009). The model using only unimputed values revealed similar significant differences in anxiety symptoms (B = 1.68 [95% CI, 0.65-2.70]; t = 3.21; df = 109.07; P = .002). Similarly, youths in Shamiri experienced greater declines in anxiety symptoms from baseline to end point than youths in the control group per the model using unimputed values (Cohen d = 0.37 [95% CI, 0.11-0.63]; Figure 2 and eAppendix 8 in Supplement 2).

Both models using imputed and unimputed data showed nonsignificant differences in depressive symptoms at 2-week follow-up between Shamiri and study skills groups (imputed values model: B = 0.82 [95% CI, −0.19 to 1.83]; t = 1.58; df = 273.78; P = .11; unimputed values model: B = 1.14 [95% CI, −0.07 to 2.33]; t = 1.85; df = 974.37; P = .06). Thus, although youths in the Shamiri group showed lower symptoms than control group youths at 2-week follow-up, this difference was not significant. For anxiety symptoms, the model using imputed data revealed a significant difference at 2-week follow-up between the Shamiri and study skills groups (imputed values model: B = 1.01 [95% CI, 0.01-2.00]; t = 1.99; df = 161.73; P = .048; unimputed values model: B = 1.07 [95% CI, −0.076 to 2.19]; t = 1.85; df = 329.48; P = .07). Similarly, anxiety symptoms were reduced significantly more for youths in Shamiri than for youths in the control group at 2-week follow-up (Figure 2 and eAppendix 8 in Supplement 2).

Both models using imputed and unimputed data revealed significant differences in depressive symptoms at 7-month follow-up between youths in the Shamiri and study skills groups (imputed values model: B = 1.68 [95% CI, 0.53-3.03]; t = 3.08; df = 103.59; P = .002; unimputed values model: B = 2.28 [95% CI, 1.04-3.51]; t = 3.6; df = 977.58; P < .001). Similarly, both models revealed significantly lower anxiety symptoms at 7-month follow-up for the Shamiri than study skills groups (imputed values model: B = 1.78 [95% CI, 0.53-3.03]; t = 2.8; df = 25.54; P = .01; unimputed values model: B = 2.25 [95% CI, 1.08-3.4]; t = 3.8; df = 345.25; P < .001).

In the intervention group, depression symptom outcomes met the threshold for clinically reliable change (3.40) from baseline to end point, 2-week follow-up, and 7-month follow-up. In the control group, depression symptom outcomes met the threshold for clinically reliable change from baseline to end point and 2-week follow-up. In the intervention and control groups, anxiety symptom outcomes met the threshold for clinically reliable change (2.24) from baseline to midpoint, end point, 2-week follow-up, and 7-month follow-up.

An independent-samples t test revealed high mean ratings and no significant difference in usefulness ratings between the Shamiri (mean [SD], 4.84 [0.52]) and study skills (mean [SD], 4.82 [0.69]) groups; t₂₈₉ = −0.7; P = .72). Participants’ ratings of whether they would recommend the program to their peers were also high for the Shamiri (mean [SD], 4.87 [0.44]) and study skills (mean [SD], 4.85 [0.53]) groups, with no significant group difference (t₂₉₁ = −0.23; P = .82).

Our multicultural team of Kenya-US researchers created and tested Shamiri, an intervention for youth depression and anxiety symptoms that focuses on positive human attributes and is designed for the Kenyan cultural context, in which resource limitations, stigma, and a shortage of health care professionals limit access to mental health care. Both the Shamiri and the study skills control led to symptom reductions among Kenyan adolescents, and these reductions generally met clinically reliable change criteria. The novel layperson-provider Shamiri intervention led to greater reductions in symptoms of depression and anxiety posttreatment and extending to 7-month follow-up. Both interventions received similar high ratings on helpfulness and likelihood of being recommended to a friend, suggesting that the significant differential effects of Shamiri on symptoms were not attributable to acceptability or appeal, but rather that Shamiri was more successful than the study skills control in specifically targeting mental health problems.

To our knowledge, this is one of the first adequately powered tests in this population of a scalable intervention grounded in simple positive psychological elements. The results align with findings on wise interventions in Western settings,^14,19,20 our small-sample pilot study,¹⁷ and a recent test in Kenya of a digital single-session adaptation of Shamiri.³⁵ Our effect sizes, within the range of effects reported for other global mental health interventions,^45,46 may warrant attention given Shamiri’s brevity, simplicity, low cost, and use of minimally trained laypersons. Finally, using Kenyan group leaders who were familiar with the local context may have enhanced Shamiri’s feasibility and acceptability.

Although the psychometric properties of our measures have been assessed and found adequate for Kenyan youths,^24,33,35 the measures were not designed or fully validated for Kenyan youths, as suggested for cross-cultural research.⁴⁷ Second, the absence of a suicide item on the PHQ-8 limited our ability to gauge intervention effects on suicidality and to compare findings with PHQ-9 studies. Third, the government ban described previously led to significant attrition at 7-month follow-up. Fourth, we lack definitive information on any treatments students might have received between the 2-week and 7-month follow-ups; however, no participant asked the study team for further treatment, and formal mental health treatment is largely inaccessible locally. Fifth, we were underpowered for, and thus did not preregister or conduct, tests of potential moderators (eg, sex, tribe). Sixth, supervisory meetings were held for both treatments at the same time, which differs from standard practice. Seventh, students in both Shamiri and study skills showed symptom reduction, and the Shamiri vs study skills effect sizes for anxiety symptoms (averaging 0.37 across assessment points) were below those reported in recent meta-analyses of more intensive treatment (0.61,²⁵ 0.66⁴⁸); however, effect sizes for depression symptoms (averaging 0.36) exceeded those in recent meta-analyses (0.29,²⁵ 0.30⁴⁸), and most outcome assessments met criteria for reliable clinical change.^43,44 The school setting of this study may be both a strength and a limitation. It provided a test in real-world conditions, with implementation supported by the school structure, and students generally liked the intervention; however, we learned that government educational policies can change suddenly, limiting opportunities for extended evaluation.

Low-burden interventions such as Shamiri and a study skills control can lead to clinically reliable change in symptoms of depression and anxiety in a low-resource setting where formal mental health care is generally not available. The Shamiri intervention led to greater symptom reductions compared with the control intervention. Given high attrition rates and some nonsignificant findings at 2-week follow-up, Shamiri, the study skills control, and other simple, brief interventions warrant further study. Shamiri was created, and the study designed and implemented, by a multicultural team with combined expertise in intervention science and the relevant cultural context—an approach that may have value for global mental health research.⁴⁹ The positive findings suggest the testable possibility that interventions that are simple in design, low in cost, focused on positive human attributes and character strengths, and delivered by laypersons may contribute usefully to global mental health.

Corresponding Author: Tom L. Osborn, AB, Shamiri Institute, 115 Applewood Adams, Ngong Road, Nairobi, Kenya (osborn@shamiri.institute).

Accepted for Publication: March 19, 2021.

Published Online: June 9, 2021. doi:10.1001/jamapsychiatry.2021.1129

Author Contributions: Mr Osborn and Dr Rusch had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Mr Osborn and Ms Venturo-Conerly contributed equally to the work and share first authorship.

Concept and design: Osborn, Venturo-Conerly, Arango G., Roe, Rodriguez, Alemu, Gan, Wasil, Ndetei, Wasanga, Schleider, Weisz.

Acquisition, analysis, or interpretation of data: Osborn, Venturo-Conerly, Arango G., Roe, Rodriguez, Alemu, Gan, Wasil, Otieno, Rusch, Schleider, Weisz.

Drafting of the manuscript: Osborn, Venturo-Conerly, Arango G., Roe, Alemu, Gan, Wasil, Rusch, Ndetei, Weisz.

Critical revision of the manuscript for important intellectual content: Osborn, Venturo-Conerly, Arango G., Roe, Rodriguez, Alemu, Gan, Wasil, Otieno, Rusch, Wasanga, Schleider, Weisz.

Statistical analysis: Osborn, Venturo-Conerly, Arango G., Roe, Wasil, Rusch.

Obtained funding: Osborn, Venturo-Conerly, Roe, Rodriguez, Alemu, Gan, Wasil.

Administrative, technical, or material support: Osborn, Venturo-Conerly, Arango G., Roe, Rodriguez, Gan, Wasil, Otieno, Wasanga, Weisz.

Supervision: Osborn, Venturo-Conerly, Arango G., Roe, Alemu, Wasanga, Schleider, Weisz.

Conflict of Interest Disclosures: Mr Osborn and Ms Venturo-Conerly report support from the Templeton World Charity Foundation. Dr Schleider reported other support from New Harbinger Publications and Oxford University Press and grants from National Institutes of Health Office of the Director, Klingenstein Third Generation Foundation, American Psychological Foundation, and Limbix Health outside the submitted work. No other disclosures were reported.

Funding/Support: This study was funded by Shamiri Institute, a nonprofit committed to improving the lives of at-risk youths in Sub-Saharan Africa through evidence-based research.

Role of the Funder/Sponsor: Although several individuals affiliated with the funder were involved in the study, the funding body had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: Gail O’Keefe, MA (Harvard University), purchased pens, pencils, and notebooks that were used by students in the study. Eliot House and Currier House (residential colleges at Harvard College) and the Harvard College Din & Tonics club contributed t-shirts for study participants. We thank the principals, school administrators, and teachers at the secondary schools in Nairobi and Kiambu County, Kenya. We extend our gratitude to Jarius Akweya, AB (Alliance High School), James Kinyanjui, AB (Alliance High School), Elizabeth Boro, AB (Kanjeru Girls Secondary School), Charles Kariuki, AB (Riara Day Secondary School), and Mwangi Muchira, AB (Alliance Girls High School). We also thank Sophie Ochola, PhD (Kenyatta University), and Rose Limbe, AB (Maseno University), for their guidance on local customs and regulations in Kenya. We are also grateful to all the group leaders for their hard work throughout the study. Compensation was not received.