Reaching the Hardest to Reach: Lessons Learned from a Feasibility Trial to Assess Online Services for Low-Income, Intimate Partner Violence Survivors in Brazil

Digital platforms are increasingly proposed to reach survivors of intimate-partner violence (IPV) in low- and middle-income countries, yet few feasibility data exist. We assessed the viability of Mapa do Acolhimento, a volunteer-run Brazilian service that recently introduced video counselling by psychologists and remote navigation by social workers. Women (18 y) who self-registered between January and April 2024 and remained unmatched after 24 h (n 316) were randomized 1:1:1 to counselling (T1 127), navigation (T2 90) or care-as-usual referrals (CAU 99). Pre-specified feasibility metrics were recruitment yield, covariate balance, 90-day retention, and instrument performance. The online funnel enrolled 18.6 survivors per week initially, declining to 8.1 as social-media click-through rates fell; 91 % resided in high-connectivity municipalities and 41 % screened positive for severe anxiety/depression. Baseline balance was achieved on 19/20 variables; race/ethnicity was imbalanced (Black/Indigenous 24 % in T2 vs 6 % in CAU; 11.7, p .01). Overall retention was 14.2 % (45/316); dropout was higher for survivors with anxiety 8/10 (aOR 1.74) or shared devices (aOR 1.59). The PHQ-2/GAD-2 displayed acceptable reliability ( .78), but 28 % of respondents exited immediately after the mental-health block, indicating survey-length fatigue. An unanticipated platform upgrade reduced eligible intake by 60 %, rendering an adequately powered RCT infeasible. Findings show that a volunteer platform can recruit and randomize high-risk women at national scale, but rural reach, risk-responsive retention, and streamlined volunteer reporting must be strengthened. We propose evidence-based adjustments, plain-language messaging, hybrid recruitment, adaptive follow-up, and mobile-first reporting, and outline methodological safeguards for future digital IPV trials in resource-constrained settings.