More Workers, Less Oversight

BRAC's Shasthya Shebika (community health worker) program is the world's largest CHW network, with over 100,000 workers across Bangladesh. The model's effectiveness was established at smaller scale: each CHW serves ~250 households, supervised by a Shasthya Kormi (health supervisor) covering 10–12 CHWs. As the program scaled beyond 50,000 workers, supervision ratios stretched — supervisors now cover 15–20+ CHWs across wider geographic areas, and the quality-maintaining feedback loops that made the model effective at pilot scale have weakened. The degradation is non-linear: a 50% increase in supervisor caseload produces far more than 50% loss in supervision quality because travel time, not meeting time, is the binding constraint.

CHW programs are the backbone of primary care delivery in low-income countries. WHO estimates that achieving universal health coverage requires 4.45 million additional health workers, and CHW programs are the primary strategy for closing this gap. BRAC's model is widely replicated — Bangladesh, Uganda, Tanzania, Sierra Leone, Liberia — but every replication inherits the supervision scalability problem. Dropout rates among BRAC CHWs in urban Dhaka slums reached 37% within two years, driven primarily by inadequate supervision and support rather than by low compensation. When supervision quality drops, CHW performance doesn't just decline — it becomes variable, eroding community trust in the entire system.

BRAC has experimented with group supervision (monthly cluster meetings replacing individual field visits), mobile phone-based reporting, and performance incentive structures. Group meetings reduce travel burden but lose the observational component — supervisors can't assess clinical technique from a meeting room. Mobile reporting captures activity metrics (visits completed, referrals made) but not quality metrics (correct assessment, appropriate counseling). Performance incentives tied to quantitative targets produce predictable gaming: CHWs log visits without delivering services, or concentrate on easy-to-count activities (product sales) over harder-to-measure ones (health education). The fundamental constraint is that quality supervision requires co-present observation, and co-present observation doesn't scale geometrically with workforce size.

Peer supervision models — where experienced CHWs supervise newer ones — could reduce the supervision ratio without proportionally increasing supervisory staff. Digital tools that capture process quality rather than activity counts (e.g., audio-recorded counseling sessions reviewed by AI or supervisors remotely) could decouple quality assessment from physical co-presence. Community accountability mechanisms — where households themselves provide structured feedback on CHW performance — could create a distributed quality signal that doesn't depend on supervisor visits. The key insight from BRAC's own research is that supervision's value is motivational and problem-solving, not primarily compliance-checking — redesigning for those functions may scale differently.

A design team could prototype a lightweight community feedback tool — perhaps SMS-based or paper card — that captures household satisfaction and CHW performance indicators, and evaluate whether this signal correlates with supervisor-observed quality. An HCI team could design and test a CHW self-assessment tool that guides reflection on practice quality without requiring supervisor presence. A systems team could model the non-linear relationship between supervision ratio and quality outcomes across different geographic densities (urban vs. rural) to identify optimal restructuring points.