A System Fix, Not a Therapist Fix: How Mozambique Closed Its MNS Care Gaps by 46 Points
- Cluster-RCT across 16 government facilities, 3,837 patients, 33,055 outpatient visits — the largest implementation trial of task-shared MNS care in sub-Saharan Africa to date
- SAIA-MH intervention improved functional outcomes by 46.0 percentage points (95% CI 34–58, p<0.0001) vs. attentional placebo control over 2 years
- Medication adherence rose by 18.1 pp and appointment attendance by 18.4 pp; WHODAS 2.0 scores fell an additional 5.9 points in the intervention arm
- The strategy requires no additional clinical staff — it works through provider team meetings, system-engineering tools, and external facilitation applied to existing psychiatric technicians and psychologists
Most global mental health research asks whether a specific therapy works. This trial asks something harder: can you fix the system that delivers therapy — and if so, how much does it matter? In Mozambique, where psychiatric technicians and psychologists are the primary workforce for all mental, neurological, and substance use (MNS) disorders, the answer is unambiguous. The SAIA-MH approach more than doubled functional recovery rates compared to a matched control condition, without adding a single clinician.
What SAIA-MH actually does
The Systems Analysis and Improvement Approach (SAIA) was originally developed for HIV and TB care cascades. The mental health adaptation — SAIA-MH — maps each facility's patient flow from diagnosis through treatment retention, then identifies the bottleneck. External facilitators run monthly "cascade workshops" with facility teams: staff review their own data, apply lean process-improvement tools (flow diagrams, plan-do-study-act cycles), and test micro-changes between sessions.
Critically, SAIA-MH is not a training programme. It does not teach new clinical techniques. It takes existing staff, existing protocols, and existing data — and makes the process of care more reliable. The trial ran from February 2022 to October 2024 across 16 government facilities in Sofala and Manica provinces, comparing SAIA-MH to an attentional placebo control (matched time with supervisors, without the improvement framework).
The patient population was heterogeneous: 67.3% had epilepsy, the remainder had psychiatric and substance use disorders. Mean age was 26 years. This is not a clean research sample — it reflects the actual outpatient MNS caseload in a Mozambican provincial hospital.
What this means for global implementation
The effect sizes are large by any standard. A 46 percentage-point gain in functional improvement is not a statistical artifact — it represents the difference between a patient who leaves outpatient care with residual disability and one who does not. The adherence and attendance gains (18 pp each) address the most common failure mode in LMIC mental health services: patients who are diagnosed and prescribed but never return.
For clinicians and system designers working in or with LMIC contexts, SAIA-MH offers a replicable model: it does not depend on importing Western psychotherapy, training lay counsellors in structured protocols, or building new facilities. The lever is the care cascade itself — making existing contacts count.
This matters for the Global North too. The UK, US, and European mental health systems are full of patients who are triaged, diagnosed, and then lost before treatment begins. SAIA-type cascade mapping is already used in chronic disease services; this trial makes the case for adapting it to outpatient psychiatry everywhere.
The largest functional gain came not from better therapy, but from making the path to therapy — and through it — reliably passable.
67% of the sample had epilepsy, limiting direct extrapolation to psychiatric-only services. Single-country trial; implementation fidelity may vary with different health system infrastructure. No cost data are reported in this publication, which constrains economic case-building.