Safeguarding When the Referral Is to a Coffee Morning, Not a Clinic

A link worker refers a 78-year-old widower with mild cognitive decline to a weekly walking group run by volunteers out of a community hall. No one from the referring service will see him again for three months. Nobody at the walking group has met him before. This is the normal shape of social prescribing, and it is also the point at which almost all of the safety thinking in the field quietly stops.

Social prescribing was built to move people out of clinical settings and into ordinary community life — that is the entire premise, and the qualitative evidence on why it works backs that up. People describe the value of social prescribing in terms of restored purpose and participation, not clinical contact. But the same feature that makes it effective — informal, volunteer-run, low on institutional scaffolding — is exactly what removes the structures that would normally catch harm: supervision, incident reporting, vetting, a record of who showed up.

The gap the evidence doesn’t cover

The safeguarding literature for social prescribing barely exists. The two systematic reviews most commonly cited in this space report positive outcomes — gains in self-esteem, confidence, and in some cases reduced use of GP, emergency, and inpatient services — but neither was built to detect harm, and neither claims to have looked for it. A 2025 protocol reviewing social prescribing for older adults notes flatly that only one peer-reviewed randomised controlled trial exists in this area at all. Absence of adverse-event reporting in this literature is not evidence of absence of adverse events. It is evidence that nobody was set up to record them.

The National Academies’ 2020 consensus report on isolation in older adults is instructive here for what it does and doesn’t ask of the health system. It calls for routine assessment of isolation and loneliness in clinical settings, and the clinician-facing commentary that followed pushes further on what that assessment obligation implies in practice. Neither document extends that obligation past the point of referral. The system is asked to notice loneliness. It is not asked to track what happens to the person after they are handed off to a walking group.

What this means in practice: treat every referral into an unsupervised or lightly-supervised community activity as carrying a safeguarding tail, not just a signposting outcome. If your programme cannot answer “did this person actually attend, and is anyone checking on them if they didn’t,” you do not have a safeguarding structure — you have a good intention with a phone number attached.

The failure mode: the warm handoff that isn’t

Programmes describe referrals as “warm handoffs” to distinguish them from a leaflet. In practice, most handoffs are warm only at the moment of referral. The link worker makes the call, the person is told when and where to turn up, and the file closes. Nobody confirms the person arrived. Nobody knows if the group turned out to be the wrong fit, if a volunteer behaved inappropriately, or if the person stopped attending after week two because someone made them uncomfortable. Call this the unwitnessed handoff: the file says “referred,” the outcome says nothing, and the only way anyone finds out something went wrong is if the person or a family member happens to complain, weeks later, to someone unconnected to the original referral.

The unwitnessed handoff is not a training problem. It is a design problem — nobody owns the interval between referral and outcome.

What a minimum safeguarding structure covers

1. Vetting proportional to contact, not to formality. A volunteer who will be alone with an isolated older adult on a regular basis needs the same background check as paid staff in equivalent roles, regardless of whether the group is run by a charity, a church, or an informal collective.
2. A named person who checks attendance, not just referral. Someone — the link worker, a coordinator at the host organisation, doesn’t matter who — needs to know within two weeks whether the referred person showed up, and to follow up if they didn’t. Non-attendance is itself a signal worth acting on, not a closed loop.
3. An escalation route that the referred person knows about. If something goes wrong at the activity — discomfort, exploitation, a volunteer overstepping — the person needs a way to report it that doesn’t require them to go back through the original clinical channel, which they may be reluctant to do.
4. Informed consent that names the limits. People referred into informal groups should be told, plainly, that the setting is not clinically supervised. Consent to a walking group is not consent to unsupervised risk; it should be an informed choice, not a default.

The two most relevant randomised trials in this space point the same direction. The HEAL-HOA trial comparing volunteering-based engagement against a control, and the follow-up HEAL-HOA trial testing telephone-delivered behavioural activation against a befriending control, both found that structured, trained delivery outperformed unstructured contact. A separate aged-care trial found that befriending alone did reduce loneliness scores — but was still outperformed head-to-head by the more structured approach. Read together, these three trials say something safeguarding teams should take literally: structure is not just what makes an intervention more effective. It is also what makes it more supervisable. An eight-session protocol delivered by trained laypeople leaves an audit trail that a drop-in coffee morning does not.

Evidence status

Claim	Status
Social prescribing improves self-esteem and confidence	Reasonably supported across qualitative reviews
Social prescribing reduces downstream service use	Weakly supported; only a handful of studies measured it
Adverse events in informal social prescribing are rare	Unmeasured, not established — the literature does not look for them
Structured, trained delivery outperforms unstructured befriending	Supported by the strongest trials available (HEAL-HOA, aged-care RCT)
Vetting and follow-up reduce harm	Not tested; inferred from general safeguarding practice, not from this literature

What this does not solve

None of this fixes the underlying shortage the UK’s 2018 loneliness strategy tried to address at a policy level: there are not enough well-run, adequately staffed community activities to refer people into, safeguarded or otherwise. A safeguarding checklist applied to a threadbare, under-resourced voluntary sector does not manufacture capacity. It also does not solve for the people who never reach a link worker in the first place — safeguarding structure, like the rest of social prescribing, only governs what happens to people who were already inside the system when the referral was made.