Effectiveness of a digital, school-based, intervention in adolescent family planning and reproductive health in Rwanda: a cluster randomized implementation trial

The study addresses whether a digital, school-based intervention (CyberRwanda) improves adolescent family planning and reproductive health (FP/RH) outcomes in Rwanda. Digital health tools may increase access to accurate FP/RH information and services, counter misinformation, and suit adolescents’ preferences for private, technology-based channels. In sub-Saharan Africa, adolescents face low contraceptive knowledge and use, high unintended pregnancy rates, and high HIV incidence, intertwined with school dropout and harmful gender norms. Despite promise, few rigorous evaluations have tested adolescent-focused digital FP/RH interventions or compared implementation models. CyberRwanda was designed through human-centered design to deliver edutainment (STORIES), information (LEARN), and access to products/services (SHOP). The trial aimed to assess effects on modern contraceptive use, childbearing, and HIV testing, and to compare a self-service digital model versus a facilitated model with peer-led clubs.

The manuscript situates CyberRwanda within evidence that digital FP/RH interventions in sub-Saharan Africa can improve knowledge, increase contraceptive use, and reduce adolescent pregnancy. UN agencies advocate equitable digital health technologies to meet SDGs by addressing information disparities. Adolescents’ preferences for private, tech-enabled information make digital interventions potentially well-suited, particularly where youth-friendly services are scarce and harmful norms persist. The paper highlights bidirectional links among poor FP/RH outcomes, HIV incidence among youth (~30% of new infections), and school dropout. It also notes limited rigorous impact evaluations of adolescent-focused digital interventions and a gap in comparing delivery models designed via community-led approaches.

Design: Three-arm, parallel-group, cluster-randomized Type 2 Hybrid Effectiveness-Implementation trial in eight Rwandan districts (pre-registered; CONSORT-adherent). Clusters were secondary schools. Eligibility for schools: within 4.5 km of a pharmacy, ≥1.5 km from another secondary school, ≥150 students; boarding schools excluded. Sixty of 61 eligible schools were randomized 1:1:1 to: (1) CyberRwanda self-service; (2) CyberRwanda facilitated; (3) control. Randomization: stratified by district, participatory allocation events with opaque-ball draws; no masking of participants or data collectors due to intervention nature. Participants: Students aged 12–19 in S1–S2, with consent/assent and contact information; no participant exclusions. Target ~100 per school (sex-stratified random selection). Enrollment and baseline: Feb–May 2021. Midline: Feb–Aug 2022. Endline: Mar–Aug 2023. Interventions: All arms received the national Comprehensive Sexuality Education curriculum. CyberRwanda components included: STORIES (webcomics on relationships, consent, contraception, pregnancy, HIV/STIs), LEARN (guides, videos, audio, FAQs on puberty, menstruation, gender equality, relationships, education, careers, money, goal setting), and SHOP (online ordering of health products via youth-friendly pharmacies, health posts, and facilities; products free at health posts/facilities and market-priced at pharmacies with periodic discounts). A facility finder was included. Content was in Kinyarwanda and English with regular updates based on feedback and analytics. - Self-service model: Self-guided platform access; ~8 tablets per school, hotspots, marketing; trained student ambassadors promoted and supported use. - Facilitated model: All self-service components plus peer facilitators leading weekly voluntary club sessions (1–2 hours; 50–150 students) aligned to digital content using an activity booklet (skits, role-plays, discussions), with group tablet access. Training: 1.5-day training for ambassadors/facilitators/lead teachers; facilitators received an additional one-day club facilitation training. Data collection: Tablet-based surveys (Kinyarwanda; Qualtrics) measuring sociodemographics (baseline), exposure/engagement, FP/RH knowledge, attitudes/beliefs, self-efficacy, sexual behavior, contraceptive use, pregnancy history, HIV testing (endline). Transport reimbursement provided. Outcomes: Pre-specified primary outcomes at individual level: (1) current modern contraceptive use among females (self or partner); (2) initiation of childbearing among females (ever pregnant); (3) ever HIV testing among all participants. Secondary outcomes included knowledge (fertility window, emergency contraception, HIV knowledge; menstrual tracking tools among females), attitudes/beliefs (FP/RH service utilization; beliefs conducive to condom use), self-efficacy (consent; partner contraceptive communication; access/use of contraceptive services among females; HIV testing), and behaviors (ever sexually active; current modern method use among males; current condom use; current use of non-condom modern methods; ever condom use among sexually active; FP counseling in past 12 months; partner contraceptive discussion among sexually active; HIV testing in past 12 months). Operational definitions are detailed in the supplementary table. Sample size and power: Simulation-based power focused on primary outcome of modern contraceptive use among females, targeting 50 females per school (n≈3,000) and equal number of males (total ≈6,000). Assuming an odds ratio ≥2.5 for facilitated vs control (increase from 2% to 5%), 80% power with 10% attrition. Statistical analysis: Generalized linear mixed models estimated prevalence ratios (PRs) with 95% CIs using log-binomial or log-Poisson (robust SE) when non-convergent; models adjusted for district and included school-level random intercepts. Joint likelihood ratio tests assessed any difference across arms. Primary analyses used inverse probability of censoring weighting (IPCW) for attrition; sensitivity analyses included unweighted models, covariate-adjusted IPCW, and an instrumental variable approach based on a composite fidelity score. Intraclass correlation coefficients (ICCs) were reported for primary outcomes. Analyses were conducted in R 4.3.1 and Stata 17. Ethics approvals obtained in the U.S. and Rwanda.

Enrollment and retention: 6,078 adolescents enrolled (mean baseline age 15; 51.5% female); 5,552 completed 24-month endline (91.3% retention). At endline, 26.6% had ever had sexual intercourse (34.7% males; 19.0% females); 11.1% reported sex in past year. Implementation fidelity and engagement: Awareness was high (mean 98.4% per intervention school). Among participants in CyberRwanda schools, 75.9% reported ever using the platform; mean 1 use in the past month. Access predominantly occurred in school (99.0%) using tablets (99.2%); smartphone access was rare (1.9%). STORIES was most used (65.7% overall; 86.5% of users), followed by LEARN (39.7% overall) and SHOP (30.1% overall). Only 12.9% ever placed a SHOP order; 70.8% of non-users cited no need/want for products. Tablet/internet challenges were common (reported by 39.2% of participants per school on average). Primary outcomes (intention-to-treat, IPCW): - Current modern contraceptive use among females: 10.7% in CyberRwanda vs 10.6% control; PR 1.04 (95% CI 0.76–1.42); ICC 0.02. - Ever pregnant among females: 2.6% in CyberRwanda vs 2.0% control; PR 1.33 (95% CI 0.71–2.50); ICC 0.02. - Ever HIV testing (all participants): 51.6% in CyberRwanda vs 51.4% control; PR 1.00 (95% CI 0.91–1.11); ICC 0.12. Secondary outcomes: Knowledge: - Emergency contraception knowledge higher in CyberRwanda: 55.9% vs 49.2%; PR 1.14 (95% CI 1.04–1.24). Facilitated: PR 1.17 (1.06–1.28); self-service: PR 1.10 (1.00–1.21). - Fertility window, HIV knowledge, and menstrual tracking tool knowledge (females) showed no meaningful differences overall. Attitudes/beliefs: - Beliefs conducive to condom use improved: CyberRwanda combined PR 1.05 (95% CI 1.01–1.09); facilitated PR 1.06 (1.01–1.12). - Views on FP/RH service utilization similar across arms. Self-efficacy: - Confidence to get partner to use contraception higher: combined PR 1.02 (95% CI 1.00–1.03) with both arms significant vs control. - Among females, confidence to access/use contraceptive services higher in facilitated arm: PR 1.02 (95% CI 1.00–1.04). Behaviors: - Current modern contraceptive use among males higher in CyberRwanda combined: PR 1.25 (95% CI 1.03–1.53); facilitated PR 1.31 (1.04–1.65). - Among sexually active participants: ever condom use higher in facilitated vs control: 61.9% vs 53.3%; PR 1.19 (95% CI 1.05–1.35). Partner contraceptive discussion higher in facilitated vs control: 59.5% vs 52.9%; PR 1.15 (95% CI 1.03–1.30). Sensitivity analyses (sexually active subgroup, n=1,477): - Current modern contraceptive use higher in CyberRwanda combined and facilitated arms; facilitated vs control: 63.0% vs 52.7%; PR 1.24 (95% CI 1.12–1.38). - Effects partly driven by increased current condom use among sexually active males in facilitated arm: 62.6% vs 49.7%; PR 1.28 (95% CI 1.11–1.48). - Sexually active females showed higher current use of non-condom modern methods in both arms vs control; combined: 14.4% vs 7.8%; PR 2.06 (95% CI 1.18–3.60). Safety: Seven participant deaths occurred during follow-up, all unrelated to the study; no adverse events related to CyberRwanda or the evaluation.

CyberRwanda was widely known and used in intervention schools, with favorable perceptions of content. The intervention did not change the primary outcomes in the full cohort at 24 months, likely reflecting that most participants were not yet sexually active and thus had limited demand for contraception or HIV services during the study window. Among sexually active adolescents, CyberRwanda increased modern contraceptive use, driven by higher condom use among males and increased non-condom modern method use among females. Intermediate outcomes along the hypothesized pathway improved (emergency contraception knowledge, condom-supportive beliefs, partner-communication self-efficacy, and females’ confidence in accessing contraceptive services), supporting the mechanism of effect. The facilitated model generally showed stronger shifts in attitudes and behaviors than self-service, suggesting that peer-led, in-person components may enhance engagement with digital content and translation into behavior, though confidence intervals often overlapped and cost/complexity trade-offs remain to be evaluated. SHOP uptake was low, primarily due to self-reported lack of need for products and potential access/privacy constraints (tablet sharing, connectivity), consistent with the relatively low and infrequent sexual activity at endline. As the cohort ages, demand for FP/RH services may rise, potentially revealing larger effects over longer follow-up.

A large, multi-district, school-based cluster randomized trial found no overall effects of CyberRwanda on modern contraceptive use among all females, initiation of childbearing, or HIV testing at 24 months, but did find increased modern contraceptive use among sexually active adolescents and improvements in several intermediate outcomes aligned with the intervention’s theory of change. The facilitated delivery model showed indications of greater impact on FP/RH-related attitudes and behaviors than self-service. The intervention was implemented with high fidelity and strong acceptability. Future work should assess longer-term impacts as sexual activity increases with age, address access barriers (e.g., more devices, offline features, connectivity support, privacy for SHOP use), and compare the cost-effectiveness of facilitated versus self-service models.

Potential selection bias from loss to follow-up (participants lost were more likely older, male, lower parental education, lower wealth), which could relate to unobserved differences in FP/RH outcomes. Reliance on self-reported sensitive behaviors may introduce social desirability bias; if nondifferential, effects would bias toward the null. Limited quantitative data on engagement with in-person club sessions constrained assessment of the facilitated component’s dose-response. The majority of participants were not sexually active, limiting power to detect effects on primary outcomes in the full sample during the study window. Implementation barriers (limited tablets, connectivity issues, tablet sharing) may have reduced engagement, especially with SHOP. Lack of blinding is an inherent limitation. Generalizability may be limited for out-of-school youth who may have different risk profiles.