Relieving depression through talk therapy

Why this opportunity

Psychological relief belongs in the picture

A suffering measure should make room for persistent emotional distress as well as physical pain. Talk therapy offers a useful test: can a modest cost help people experience less distress over weeks or months?

Friendship Bench is a promising candidate because treatment has been studied in randomized trials, its costs have been examined publicly, and the program remains active. The calculation below makes the remaining assumptions visible.

One USU is a stipulated six-hour kidney-stone pain reference of 47.25 intensity-hours. Using the same unit for emotional and physical experiences is a proposed reporting convention, not a clinically validated equivalence. How the unit works

Read the evidence before the number

A promising result, with important limits

The original trial provides a starting point

A 2016 trial randomized 24 Harare clinics and enrolled 573 adults screening positive for common mental disorders. It offered six therapy sessions plus optional peer support. At six months, 13.7% in the intervention group and 49.9% in enhanced usual care met the depression-screening threshold, PHQ-9 ≥ 11. [B1]

The reported proportions differ by 36.2 percentage points. That is the reference for this model, not a claim that every participant was depressed initially or that 36.2% were permanently cured. It is a rounded, unadjusted absolute difference in a secondary outcome.

Later evidence gives us reason to discount that result

A later rural trial in 516 adults living with HIV found benefits on a broader symptom measure at some follow-ups, but no difference in depression-score changes. [B3] The populations and delivery settings differ, so the original result should not be treated as universal.

Happier Lives Institute’s 2025 implementation review also highlighted historical routine attendance averaging only 1.12 of six possible sessions. [B7] Friendship Bench’s 2025 monitoring report had complete clinical follow-up for 3,009 of 13,864 enrolled clients, without a concurrent control group. [B4] Those data cannot by themselves establish the causal effect of current delivery.

Our central scenario therefore retains half of the original trial difference. This exact discount is a WSI choice, not an estimated correction. We also show a no-added-effect scenario. A mixed evidence base warrants that possibility, while leaving open the potential for substantial benefit.

Explore the estimate

Try an amount and a set of assumptions

The no-added-effect case changes only effectiveness from the central assumptions. The higher case combines cheaper delivery, more additional eligible care, a larger retained effect and greater relief. The endpoints do not bound actual benefit or harm.

Follow the calculation

From support to an assumed reduction in distress

1. Estimate additional eligible care

   Allow $50 per client, then assume half the support becomes additional adult care resembling the reference population. This factor covers allocation, age and setting eligibility, and care that otherwise would have happened anyway.

   $100 ÷ $50 × 50%
   = 1 additional eligible client

2. Estimate the incremental outcome

   Apply the 36.2-point trial difference, retaining half for future delivery. The comparison already accounts for improvement in the control group, so we do not count all treated people as additional successes.

   1 × (49.9% − 13.7%) × 50%
   = 0.181 additional improved outcomes

3. Make the USU assumption explicit

   For each additional improved outcome, assume a 1-point average reduction in experienced distress over 90 days. This creates a relief scenario from the screening result. It is not a conversion of PHQ-9 or SSQ-14 points into suffering units.

   0.181 × 90 days × 24 hours × 1 intensity point ÷ 47.25
   = 8.3 USU relieved per $100

Linear weighting, p = 1. Use one overall psychological reduction so overlapping symptoms are not added twice. No USU is assigned to death itself or to a diagnosis.

Why these cost, duration and intensity choices?

A broad cost allowance

A published economic model used $170,419 in annual fixed costs, $3.37 in variable costs per client and 12,364 clients annually. This gives about $17.15 per client in 2019 dollars, including program and city health resources. Some inputs were modeled or anticipated, rather than audited spending. [B2]

We use $50 as a provisional future allowance, with $25 in the higher scenario. The uplift leaves room for changed prices, delivery costs and fuller support. It is not a measured inflation adjustment or a quote for one more client. We assume donation dollars can support this broad cost envelope and claim no extra multiplier from government co-funding.

A short, assumed relief window

We credit days 91–180 after care begins: 90 days within the first six months. The trial’s six-month endpoint does not establish continuous relief across those days. That duration is an explicit assumption, chosen to avoid a multiyear projection. Earlier and later benefits are unestimated.

An average change, not a clinical conversion

The central 1-point reduction allows modest improvement averaged across every hour, including sleep and periods of relief. The higher scenario uses 2 points. These are WSI judgments about experienced suffering, not established averages. Crossing a screening threshold does not tell us how much a person’s distress changed.

The 50% additional-care factor and the 50% retained trial effect answer different questions: did the donation bring additional eligible care, and how effective was that care? Neither is directly estimated from current program data.

Funding snapshot · 12 October 2026 (UTC)

Ongoing work, with the next allocation unconfirmed

Friendship Bench Zimbabwe’s Q1 2026 report records 88,436 clients seen and invites further funding partnerships. [B5] The organization describes a shift toward supporting government delivery, alongside broader populations and delivery channels. [B6] That makes additional reach and outcome quality especially important to verify.

An official donation page is available, but we have not verified the current remaining gap, a specific unfunded project or the share of a new gift that will support comparable adult care. Historical funding-gap figures are not presented as current need.

What the number covers

We do not infer suicides prevented from changes in suicidal thoughts or depression scores. Unestimated deaths remain unknown, not zero. Gross relief also does not subtract potential treatment harms.

The three opportunity reports use different populations, benefit windows and cost assumptions. This number is a starting point for examining a possible use of funding, not evidence that it outranks malaria prevention or fistula repair.

Inspect and improve

Sources and reproducible inputs

1. Chibanda et al. (2016): randomized trial in HarareHistorical trial reference; reported absolute proportions, not an adjusted absolute effect. Abstract: six-month PHQ-9 threshold proportions, 13.7% versus 49.9%
2. Healey et al. (2022): economic analysis of Friendship Bench scale-upHistorical cost context, not a current marginal donation price. Table 2, printed page 51; payer perspective and 2019 USD in methodsRead the public PDF copy
3. Haas et al. (2023): randomized trial in rural ZimbabweEvidence that results differ across populations and outcomes. Abstract: SSQ-14 and PHQ-9 results at 3, 6, 9 and 12 months
4. Friendship Bench Zimbabwe: 2025 annual reportRecent monitoring, with incomplete follow-up and no concurrent comparison group. Printed pages 12-13: follow-up cohort and limitations; not the current USU effect input
5. Friendship Bench Zimbabwe: Q1 2026 impact reportCurrent operations and request for support, not verified remaining need. Printed page 9: clients seen; page 17: funding invitation
6. Friendship Bench Zimbabwe: evolution of the programTransition toward government delivery and technical assistance; broader populations.
7. Happier Lives Institute (2025): mid-year charity updateEvaluator's original implementation analysis; not assumed to measure 2026 attendance. Friendship Bench section: historical routine attendance averaging 1.12 of 6 possible sessions

The Zimbabwe reports are linked from the organization’s official report archive. The source audit records the retained facts and why we did not use before-and-after recovery rates as causal effects.

The most useful refinements are current cost and funding data, outcomes from routine government-supported care with a comparison group, and direct measurements of experienced distress over time.

Prepared 12 October 2026 · Friendship Bench model v0.1 · Funding recheck due 11 November 2026 or after a material change · Suggest a correction