Task-shifting worked
For most of the last century, the answer to too few Psychiatrists was to train more Psychiatrists. In a region with roughly one for every half a million People, that answer was arithmetic that never closed. Task-shifting proposed something else: train non-specialists to deliver structured, evidence-based support, and reserve scarce clinical time for the People who need it most.
It is not a compromise, and the evidence is not thin. Zimbabwe's Friendship Bench, where trained grandmothers deliver problem-solving therapy on wooden benches outside clinics, produced better symptom outcomes than enhanced usual care in a randomised trial published in JAMA in 2016. The World Health Organization built its mhGAP guidance on the same principle. In Kenya, Programs delivering brief structured interventions to adolescents through trained young Facilitators have run randomised trials, preregistered protocols and published results, including the ones that did not favour them.
The People running this work are among the most rigorous in global health. This piece is not an argument against any of it. It is about the one problem the model cannot solve from inside itself.
The escalation problem
Every well-run Program screens. Screening is the point: find the People who need support, and find early the smaller group who need more than the Program is designed to give. A validated Instrument does its job, a Facilitator sees a score that crosses a threshold, and a young Person is sitting in front of them who needs licensed clinical Care.
Then what?
In practice, then a phone call. Then a name someone remembers from a workshop, or a public facility with a six-week queue, or a private Therapist whose fee is four times what the family can carry. The referral is written into a report as a number, and nobody finds out what happened to the Person behind it. The Program did everything right and still could not finish the sentence.
This is not a clinical failure. It is missing infrastructure, and it has four parts:
- Nobody to refer to. There is no verified, searchable Network of licensed Practitioners with confirmed standing, current availability, the right language and a known rate.
- No way to pay. The Program's budget funds the Program. The Session on the other side of the referral is funded by nobody, so it happens or it does not depending on the family.
- Nothing follows the Person. Screening scores, consent and context stay inside the Program's own tools. The Practitioner starts from zero, and so does the young Person, retelling the hardest part to a stranger.
- No loop back. The Program cannot report whether escalated People got Care, because after the handoff it has no visibility at all. The most important number in the Funder's report is the one nobody can produce.
What we are building underneath
Heyrafiki is not a Program and will never compete with one. We do not deliver the intervention, train the Facilitators or own the relationship with the school, the county or the community. We build the layer that starts where a Program's mandate ends.
- A verified Network. Every Practitioner checked against the Official Records of the Board they practise under, re-checked on a standing schedule, filterable by language, focus and availability.
- Care that is funded before it is needed. A Program, Funder, Employer or Insurer funds a Care Budget, and escalated Sessions draw from it. The Practitioner is paid the moment the Session ends. Nobody negotiates a fee at the worst possible moment.
- Consent-led continuity. Screening results and context travel with the Person, when the Person agrees, so the first Session starts at minute one instead of minute thirty.
- Aggregate reporting that closes the loop. How many were escalated, how many were seen, how long it took, how scores moved. Aggregate only, never a name, never a diagnosis, in a shape a Funder or a County can read.
The bar this has to clear
Anyone building alongside this field is held to its standard, and the standard is high: preregistered protocols, published null results, open model weights, validated Instruments used as published. We hold ourselves to the same terms and say plainly where we are. Heyrafiki has not run its outcome study yet. The feasibility protocol is written, the instruments are chosen and the results will be published either way, including if they embarrass us.
What we can already say precisely is how the system works: what Iris weighs when she matches, what she never decides, which components are Open Source and which stay closed because publishing them would help someone evade a safety check. Precision now is the only honest substitute for evidence we have not yet earned.
An open invitation
If you run a lay-provider Program, a school Programme or a community Programme anywhere in the region, the escalation layer is ours to build and yours to use. Referral pathways into licensed Care, funding attached to the referral, Consent-led records and reporting your Funders can audit. Keep your model, your Facilitators and your evidence base. Hand over the plumbing.
Building a Program that needs somewhere to escalate to? Come and tell us what the handoff should do. We would rather design it with you than for you.
