USSD-Based Digital Health in Rural Africa: A Machine Learning Research Direction for Low-Resource Health Signal Processing

Background: Sub-Saharan Africa (SSA) bears 25% of the global disease burden, yet accounts for only 3% of the world's health workforce [1]. Smartphone-dependent digital health platforms have failed to reach most rural populations in SSA, and the 2023 collapse of Babyl Rwanda demonstrated the structural fragility of externally owned digital health infrastructure [2].  Objective: To evaluate the feasibility, acceptability, and 90-day user retention of HealthDrive, a USSD-based telehealth platform with community health worker (CHW) integration, in a pilot study conducted in two rural SSA communities.  Methods: A mixed-methods pilot implementation study (n=50 enrolled patients, 12 CHWs) conducted August 2024 to March 2026, applying the Consolidated Framework for Implementation Research (CFIR) [3], the Technology Acceptance Model for Resource-Limited Settings (TAM-RLS) [4], and the RE-AIM evaluation framework [5]. USSD interaction logs (1,247 sessions across four short codes), CHW follow-up records, and structured satisfaction interviews were analysed.  Results: Three-month user retention was 78% (95% CI: 64–88%), exceeding SSA mHealth benchmarks (45–65%). Elderly user satisfaction reached 85%. Emergency triage sessions achieved 71% completion. Total platform expenditure was $2,580 over 19 months at $125/month.  Conclusions: USSD-based telehealth with CHW integration is feasible and acceptable in rural SSA. Five open machine learning and signal processing challenges are identified as critical barriers to scaling this model to population-scale voice-based health triage.