Safeguarding in Social Prescribing: How Programmes Have Approached the Risk

A GP refers a patient to a walking group run by a local charity. The patient has moderate cognitive impairment and no next of kin listed. The walking group has one volunteer leader, no clinical training, and no formal channel back to the GP practice if something goes wrong. This is not a hypothetical edge case. It is the standard operating model of social prescribing, and it is worth being precise about what that model does and does not cover.

The referral pathway was not built for risk

Social prescribing works by moving a person out of a clinical or clinical-adjacent setting — a GP practice, a link worker’s caseload — into a community or voluntary sector activity: a gardening group, a men’s shed, a choir, a walking club. The systematic reviews of the model report real benefits from this move. Participants describe gains in self-esteem and confidence, and one qualitative synthesis found the benefit runs deeper than social contact alone — people describe recovering a sense of purpose and meaningful participation, which structured group activity appears to deliver more reliably than unstructured contact.

None of that literature was designed to answer a different question: who is responsible if something goes wrong once the person is in the room. The link worker who made the referral typically has no ongoing line of sight into the activity. The voluntary organisation running it typically has no clinical information about the person and no formal duty to report back. The GP who wrote the referral has, in most cases, discharged their responsibility the moment the referral was made.

The National Academies’ 2020 consensus report on isolation in older adults calls for the health care system to routinely assess isolation and loneliness — treating it as a risk factor deserving the same systematic attention as blood pressure. A clinician-facing commentary on that report goes further, asking what routine assessment would actually require in practice: training, referral infrastructure, and follow-up. Neither document addresses what happens to that assessment once a referral crosses into a setting with no clinical governance at all. The gap is structural, not incidental.

Where the risk concentrates

Not all social prescribing referrals carry equal risk. Three factors, considered together, tell you how much safeguarding weight a given pathway is carrying.

Factor	Lower risk	Higher risk
Population	Adults with a strong existing network	Isolated older adults, people with cognitive decline, people living alone
Setting	Peer-led group, public venue, daytime	One-to-one contact, private premises, home visits
Oversight	Trained facilitator, incident reporting line to referrer	Volunteer-led, no formal feedback loop

The AARP Foundation’s 2018 national survey of adults 45 and older is a useful reminder of who tends to be on the referral list in the first place: network size and diversity, and physical isolation, are the two strongest predictors of loneliness in that population — which means the people social prescribing is most likely to reach are disproportionately the people with the fewest others checking in on them. That is precisely the group for whom an unmonitored community activity carries the most risk if something goes wrong, and precisely the group least likely to have someone else notice.

What “safeguarding” means here, concretely

Safeguarding in this context is not primarily about preventing abuse by staff, though that duty still applies to any organisation working with vulnerable adults. The more common and more neglected risk is absence of response: a participant deteriorates, stops attending, or discloses something concerning, and no one in the chain is positioned to act on it.

A programme approach to this risk needs to answer four questions before a single referral is made, not after an incident forces the issue.

1. Who receives a disclosure or a concern, and within what timeframe? A named individual, not a generic inbox.
2. What happens when someone stops attending? A missed session is either logged and followed up, or it isn’t — there is no middle setting.
3. What information travels with the referral? Link workers typically know more about a person’s vulnerability than the activity provider does. Deciding what crosses that boundary, and with what consent, is a design decision, not an afterthought.
4. What is the escalation route back to a clinician or statutory safeguarding lead? Voluntary organisations are not equipped to make clinical judgments, and should not be expected to.

The UK’s 2018 loneliness strategy, the first national strategy of its kind, funded social prescribing and embedded loneliness measurement into national statistics — a substantial commitment. It did not, in the material available, specify safeguarding infrastructure for the community organisations receiving referrals as a condition of funding. That absence has largely persisted in how the model has been implemented since: funding attaches to activity and reach, not to the governance structure sitting underneath it.

What this means in practice: before commissioning or accepting social prescribing referrals, get a written answer to the four questions above from every organisation in the pathway — including your own. If the answer to any of them is “we haven’t formalised that,” treat it as an open safeguarding gap, not a minor administrative gap. It is the same gap.

The failure mode: diffusion of responsibility

Call it the handoff illusion. Each party in the chain — GP, link worker, voluntary organisation, volunteer facilitator — reasonably assumes that responsibility for the participant’s wellbeing sits with whoever holds the relationship at that moment. The GP assumes the link worker is monitoring. The link worker assumes the activity provider will flag concerns. The volunteer facilitator, who has had no safeguarding training and no formal reporting relationship with anyone, assumes someone upstream is watching. Everyone is partly right and no one is accountable, which in practice is the same as no one being accountable at all.

This is not a criticism of any individual actor in the chain. It is a design defect. Diffusion of responsibility is a predictable property of any pathway that moves a vulnerable person across an organisational boundary without an explicit, written transfer of duty at each handoff. The systematic review of social prescribing’s impact on loneliness found that three of nine included studies reported reductions in service use — GP visits, emergency attendance, social work contact. That is the outcome funders want to see. It is also the exact signal that could mask a deteriorating participant who has simply stopped generating contacts with the health system, rather than one who has genuinely improved.

The evidence base does not cover this

Claim	Evidence status
Social prescribing improves self-esteem and confidence	Reasonably supported, multiple reviews
Social prescribing reduces use of GP, emergency, or social work services	Weakly supported — three of nine studies in one review, heterogeneous methods
Safeguarding failures in unmonitored community referrals are rare	Not measured — no source in the current literature tracks this
Named-responsibility handoffs reduce missed-disclosure risk	Plausible, not tested — an operational inference, not a finding

That last row matters most. The systematic reviews of social prescribing measure wellbeing and service-use outcomes. They do not measure safeguarding incidents, near-misses, or the adequacy of referral-pathway governance, because that is not what anyone has set out to study. A programme manager cannot currently point to a body of evidence that a given handoff protocol reduces harm. What can be pointed to is the absence of any structure at all in a large share of existing pathways, which is a lower bar to clear than proving an intervention works — and one that should be cleared regardless.

What this does not solve

None of this addresses the deeper reach problem in social prescribing: referral pathways depend on someone first presenting to a GP or link worker, which systematically excludes people who never make it into a clinical setting at all. A safeguarding protocol makes the pathway safer for the people already in it. It does nothing for the isolated adult who never gets referred in the first place, and no source reviewed here offers a fix for that.