Self-help mobile messaging intervention for depression among older adults in resource-limited settings: a randomized controlled trial

Depression in older adults is a major global health concern, especially in low- and middle-income countries where most older adults live and access to mental health care is limited. In Brazil, a substantial proportion of older adults experience depressive symptoms, often without diagnosis or treatment. Task-sharing and collaborative care programs have proven effective but typically require significant health professional involvement, which is challenging in resource-constrained health systems. Self-help digital mental health interventions may improve access with minimal professional support, yet evidence among older adults in LMICs is scarce. The Viva Vida program, a 6-week WhatsApp-delivered psychosocial intervention based on psychoeducation and behavioral activation and adapted from the PROACTIVE collaborative care program, was developed to address this gap. This trial evaluated whether Viva Vida improves depressive symptomatology at 3 months among older adults with significant depressive symptoms registered in primary care in socioeconomically deprived areas of Guarulhos, Brazil.

Prior research shows that task-shared psychological interventions and collaborative care can improve depression outcomes across age groups, including older adults, but usually require professional support. Digital behavioral activation interventions have produced small-to-medium effects on depression, anxiety, and quality of life in adults. Two CONEMO trials in Brazil and Peru demonstrated short-term (3-month) improvements in depressive symptoms with minimal nurse support, though benefits waned by 6 months. Systematic reviews indicate limited evidence for digital mental health effectiveness specifically in older adults, particularly in LMICs. WhatsApp and similar messaging platforms are widely used and may facilitate feasible delivery without requiring new app downloads or advanced digital skills. The Viva Vida program was designed to leverage these findings by providing fully self-help behavioral activation content to older adults via WhatsApp.

Design: Pragmatic, single-blind, two-arm, individually randomized controlled trial (1:1 allocation) conducted in 24 primary care clinics (UBSs) in socioeconomically deprived areas of Guarulhos, Brazil. Participants: Individuals aged ≥60 years, registered with participating UBSs, able to receive WhatsApp messages, screening positive for depressive symptomatology (PHQ-9 ≥10) with at least one core symptom (PHQ-2 ≥1). Exclusions: substantial visual/hearing impairment impacting message comprehension, acute suicidal risk at screening, co-resident already in trial, or prior participation in PROACTIVE. Verbal consent obtained and audio-recorded. Recruitment and procedures: A list of older adults from UBSs was prescreened for active WhatsApp numbers. Eligible individuals were phone-screened (PHQ-2 then full PHQ-9), completed baseline assessments and were invited to the trial. Follow-ups were conducted by phone at 3 months (weeks 12–16) and 5 months (weeks 20–24) after first/single message delivery. Randomization and masking: Stratified by age group (60–69, 70–79, 80+), sex, and baseline PHQ-9 severity (10–14, 15–19, 20+). Allocation via REDCap using randomly permuted blocks (sizes 6, 8, 10). Research assistants conducting recruitment and outcome assessments were blinded; participants could not be masked. Interventions: Both arms continued usual primary care. Technical support was available early to resolve message access issues. - Intervention (Viva Vida): 6-week WhatsApp program delivering 48 auto-sent audio/visual messages (two per day, 4 days/week). Content used psychoeducation and behavioral activation via storytelling (characters Ms Ana and Mr Leo), with eight visual summary messages and weekly quick-reply feedback prompts. One initial support contact message provided a technical support number. Participants were advised to seek UBS care if needed. - Control: Single brief (~6-min) audio message with basic education about depression, simple coping tips (e.g., sleep, diet), and advice to seek care if needed. Outcomes: Primary outcome was improvement from depressive symptomatology at 3 months (PHQ-9 <10). Secondary outcomes: improvement at 5 months; ≥50% reduction in PHQ-9 from baseline at 3 and 5 months. Exploratory continuous outcomes: PHQ-9, GAD-7, EQ-5D-5L, ICECAP-O, 3-item UCLA loneliness. Safety: Collected severe adverse events (acute suicidal risk, hospitalizations, death) at follow-ups using a standardized form; suicidal ideation triggered risk assessment and notification of UBS managers. Data management: Data captured in REDCap; WhatsApp Business API used for automated message delivery and engagement logs (timestamps, opens, quick replies). Sample size: Assuming 25% attrition, 440–500 participants provided 80–85% power (two-sided alpha 5%) to detect an absolute 15 percentage point difference in improvement (25% control vs 40% intervention) at 3 months. Statistical analysis: Intention-to-treat using regression models with multiple imputation by chained equations (MICE; 50 imputations) under missing at random, adjusted for stratification variables; continuous outcomes additionally adjusted for baseline values. Logistic regression for binary outcomes; linear regression for continuous outcomes. Prespecified subgroup interactions (sex, age, education, hypertension/diabetes, baseline PHQ-9 severity). Sensitivity analyses included timing adjustments and selection-model weighting to assess departures from MAR. Complier average causal effect (CACE) analyses used instrumental variables to estimate effects by message opening thresholds (≥24, ≥36, all 48). Diagnostics included Box-Tidwell and residual checks. Analyses conducted in Stata 17. Participants: 603 randomized (298 intervention; 305 control) between 8 Sep 2021 and 8 Apr 2022; 527 (87.4%) followed at 3 months; 510 (84.6%) at 5 months. Baseline characteristics were balanced across arms.

- Primary outcome (3 months): Improvement (PHQ-9 <10) occurred in 109/257 (42.4%) intervention vs 87/270 (32.2%) control; crude absolute difference 10.2 percentage points (95% CI 2.0–18.4). Adjusted OR (imputed) 1.57 (95% CI 1.07–2.29; P=0.019). Complete case results similar (adjusted OR 1.58; 95% CI 1.09–2.29; P=0.016). - Secondary outcomes: At 5 months, improvement was 127/250 (50.8%) intervention vs 130/260 (50.0%) control; adjusted OR 1.01 (95% CI 0.70–1.46; P=0.938). ≥50% PHQ-9 reduction favored intervention at 3 months (adjusted OR 1.58; 95% CI 1.08–2.29; P=0.016), but not at 5 months (adjusted OR 1.04; 95% CI 0.73–1.48; P=0.841). - Continuous secondary measures: No significant differences at 3 or 5 months for PHQ-9 means (3 mo: −1.14; 95% CI −2.33 to 0.05; P=0.060; 5 mo: −0.21; 95% CI −1.36 to 0.94; P=0.716), GAD-7, EQ-5D-5L, ICECAP-O, or 3-item UCLA loneliness (all P>0.05), with similar complete case results. - Engagement: 226/298 (75.8%) opened ≥36/48 messages (minimum therapeutic dose); 237 (79.5%) opened ≥24; 173 (58.1%) opened all 48. CACE analyses showed marginal evidence of greater PHQ-9 improvement at 3 months with higher message opening (e.g., ≥36 messages: adjusted mean difference −1.59; 95% CI −3.21 to 0.02; P=0.054), not sustained at 5 months. - Safety: Suicidal ideation reported by 6 participants (2.0%) in each arm; 18 (6.0%) intervention and 20 (6.6%) control hospitalizations (non-mental health); two deaths (0.7%) in control. No events related to intervention or study participation. - Sensitivity analyses: Results robust to adjustments for timing and to weighted analyses allowing data missing not at random; estimates at 3 months remained in favor of intervention (e.g., adjusted ORs ~1.52–1.56 across departures).

The trial demonstrates that a fully self-help, WhatsApp-delivered psychosocial intervention based on behavioral activation can accelerate short-term improvement in depressive symptomatology among older adults in under-resourced primary care settings. Compared with a single educational message, Viva Vida increased the odds of achieving PHQ-9 <10 and of ≥50% symptom reduction at 3 months, aligning with prior digital BA trials (e.g., CONEMO) that show early benefits that diminish over time. No between-group differences persisted at 5 months, suggesting natural improvement in controls and potential need for maintenance strategies (e.g., booster messages) to sustain gains. The intervention was feasible and acceptable, with high adherence via a widely used platform (WhatsApp), and minimal resource requirements, offering a scalable approach to reduce treatment gaps where specialist support is scarce. Although the absolute effect (≈10 percentage points) was smaller than the prespecified target (15 points), it remains meaningful in contexts with limited alternatives and could be integrated into stepped-care models in primary care. The absence of effects on anxiety, loneliness, and quality-of-life outcomes indicates specificity to depressive symptomatology and/or insufficient intensity or duration to influence broader domains without additional support.

A brief, self-help WhatsApp messaging program (Viva Vida) produced small but significant short-term improvements in depressive symptomatology among older adults in socioeconomically deprived settings without requiring health professional support, demonstrating feasibility and potential for scalable integration into primary care. Benefits were not sustained at 5 months, indicating a need for maintenance strategies. Future research should evaluate booster messaging to prolong effects, cost-effectiveness, process mechanisms (including engagement-dose relationships), personalization by participant characteristics (e.g., sex, comorbidities), and adaptations for individuals with low digital literacy or without WhatsApp access.

Assessments were conducted by phone and relied on the PHQ-9 threshold (≥10) as a pragmatic measure of clinical depression; comorbid conditions (e.g., diabetes, hypertension) were self-reported; cognition could not be formally assessed. Individuals without WhatsApp access or with low digital literacy were excluded, limiting generalizability and necessitating adaptations to reach these groups. Some attrition occurred (≈12–16% by 5 months) with potential for bias, though missingness was similar across arms and analyses (imputed and complete case) were consistent. Possible nonspecific effects from remote assessments or occasional technical support contacts could not be fully disentangled from intervention effects. The observed absolute effect was smaller than the target difference, and cost-effectiveness and process evaluations were not yet reported.