n.V1-7.14 | LAY-DELIVERED THERAPY

01| Zimbabwe has roughly ten psychiatrists for a population of about fifteen million. Any solution to common mental disorders that requires a psychiatrist is, in that arithmetic, not a solution. The Friendship Bench started from that constraint rather than from a theory.

02| Lay community health workers — many of them older women, grandmothers, with no clinical credential — deliver structured problem-solving therapy, activity scheduling and peer group support, from benches placed outside primary care clinics. At six-month follow-up, 14 percent of the intervention group had symptoms of common mental disorders, against 50 percent of controls.

03| In Goa, the Healthy Activity Program tested the same principle against a harder target: brief behavioural activation for moderately severe to severe depression, delivered by lay counsellors in primary care. It outperformed enhanced usual care, was accepted by a population that had previously received no treatment at all, and was cost-effective in that setting.

04| The temptation is to read these as evidence that lay people are as good as specialists. That is not what they show, and the weaker claim is more useful than the stronger one. Both trials transferred a manualised, time-limited, structured treatment — a protocol with steps, a defined ending, and supervision behind it. What transferred was not clinical judgement. It was the protocol.

05| That distinction is the practical yield of this chapter and it generalises. A task that can be written down as a sequence, taught in weeks, supervised against a checklist and bounded in time can move to non-specialists with results that survive randomisation. A task that requires open-ended judgement about which of many possible things to do next — navigating a household across housing, benefits, schooling, health and debt, for instance — has not been shown to move the same way, and nothing in these trials suggests it does.

06| So the honest reading is a rule about task design rather than about people. Protocolise the task and the credential stops mattering. Leave the task open-ended and the trials say nothing about what happens.

07| The unstated cost is supervision. Both programmes ran with clinical oversight standing behind the lay deliverer — someone to escalate to, someone checking fidelity. That layer is invisible in the headline result and is the first thing removed when a ministry scales the model, because it is the part that requires the specialists the country does not have. Whether the effect survives its removal has not been tested, and the community health worker literature in the previous chapter suggests the answer.

08| One further point, which the sector generally avoids. These programmes work partly because they are cheap in a domain where the alternative is nothing at all. Fourteen percent against fifty percent is a large effect measured against no treatment, in a country with ten psychiatrists. It is not a demonstration that this is the best available care; it is a demonstration that it is far better than the absence of care, which is what was actually on offer. That is a real and defensible thing to build, and it should not be dressed up as more.

09| Open: which categories of task are protocolisable, tested rather than assumed; whether effects hold when clinical supervision thins; and whether a lay deliverer carrying a general caseload rather than a single protocol produces anything at all.