Safeguarding in Social Prescribing: A Procedure, Not an Afterthought

A link worker refers an isolated 78-year-old to a walking group run by volunteers. Three months later, the group notices she has stopped eating properly and is confused about dates. No one on the walk is trained to assess this, no one is sure whether it is their job to raise it, and the referring service has no idea anything has changed. This is not a hypothetical edge case. It is the ordinary consequence of routing vulnerable people into informal, volunteer-led activity without building a safeguarding pathway around it first.

Social prescribing works by deliberately putting people who are isolated, unwell, or otherwise vulnerable into rooms with strangers, run by community groups that were never designed as care settings. That is the point of it — informal, non-clinical, low-barrier. It is also exactly why safeguarding needs to be designed in from the start rather than bolted on after something goes wrong.

Why this gets skipped

Safeguarding language belongs, in most commissioners’ minds, to formal care settings: a care home, a children’s service, a hospital ward. Community groups running a choir or a men’s shed do not think of themselves as safeguarding-relevant, and referrers often do not think to ask them to. The systematic reviews of social prescribing note limited trial evidence and heterogeneity across programmes generally — the same heterogeneity shows up in how, or whether, safeguarding is handled, because there is no standard structure imposed on the activities being prescribed. A qualitative synthesis of how people experience social prescribing found that participants value benefit going well beyond contact itself, toward restored participation and purpose — which is precisely why these activities can involve real trust, real dependency, and real risk once someone is inside them.

The procedure

Step 1: Classify the activity’s risk profile before you refer anyone into it. Not all social prescribing destinations carry the same exposure. A drop-in coffee morning in a public library has different risks than a home-visiting befriending scheme, which has different risks again than an activity involving physical contact, transport, or one-to-one time behind closed doors. Score each activity on three axes: level of physical proximity and privacy, degree of contact with people who have cognitive impairment or acute vulnerability, and whether volunteers have unsupervised access to someone’s home. High scores on any axis mean the activity needs a written safeguarding protocol before you send anyone there, not after an incident.

Step 2: Establish what the receiving organisation can actually see and report. Before referring, ask the group directly: who would notice if a participant’s presentation changed — confusion, injury, distress, disclosure of harm — and what would that person do next? If the honest answer is “we’re not sure,” that is not a reason to exclude the activity, but it is a reason to put a named contact and an escalation route in place before the first referral, not to discover the gap when it matters.

Step 3: Fix the handback loop. Referral pathways are usually built to send people one way — from clinician or link worker into the community. They are rarely built to send information back. Decide, in advance, what triggers a report from the community group to the referrer: a missed session pattern, a disclosure, a visible decline, a safeguarding concern under statutory thresholds. Put this in writing with the group, not as a compliance form nobody reads, but as a two-line agreement: what to watch for, who to call, how fast.

Step 4: Match supervision level to vulnerability, not to activity type. The temptation is to safeguard by category — “befriending needs a check, choir doesn’t.” Match instead to the person. Someone recently bereaved and cognitively sharp attending a well-populated group class carries different risk than someone with early dementia attending the same class. Build a light vulnerability flag into the referral itself — not a diagnosis, just a prompt for the receiving group to pay a bit more attention — without turning the whole system clinical.

Step 5: Train volunteers to notice, not to assess. Volunteers running community activities should not be expected to make clinical judgments about capacity, cognitive decline, or abuse. They should be expected to notice change and know exactly who to tell. That distinction matters and should be explicit in whatever training or induction they receive: your job is not to decide if this is serious, your job is to flag it to someone whose job is to decide.

Step 6: Set a review point, and mean it. Most safeguarding failures in these settings are not sudden. They are slow drift that nobody names because no one owns the moment of naming it. Build a fixed review point into every referral — six weeks, twelve weeks — where someone actively asks the receiving group and, where possible, the person themselves: has anything changed? This is a small administrative act that catches the case that would otherwise be caught only by luck.

What this means in practice: before you refer anyone into an informal activity, you need three things written down — who notices a change, who they call, and when the next deliberate check happens. If you cannot answer those three questions for a given activity, it is not ready to receive vulnerable referrals, whatever else it offers.

The named failure mode: silent handback

Call it silent handback. A referral goes from clinician or link worker to community group. The person attends, or stops attending, or something changes about their presentation — and nothing comes back the other way, because no one ever specified that anything should. The referrer assumes the activity is going fine because they hear nothing. The community group assumes any problem is someone else’s responsibility to escalate, because no one told them otherwise. Both sides are quietly relying on the other. This is the single most common structural gap in informal social prescribing pathways, and it is fixable with a one-page agreement at referral, not a policy department.

Evidence status

Claim	Evidence status
Social prescribing produces self-reported wellbeing and confidence gains	Reasonably supported across reviews, though trial evidence is thin and heterogeneous
Structured, purposeful activity outperforms unstructured contact	Supported by qualitative synthesis, not by controlled trials
Safeguarding protocols reduce harm in community-based referral	Not directly evidenced in the social prescribing literature; inferred from general safeguarding practice in health and social care
Older adults referred into community activity are at elevated risk of undetected isolation-related decline	Supported by consensus reporting on isolation in older adults generally, not specific to social prescribing pathways

The gap in that table is the honest one: nobody has published a trial measuring whether a safeguarding protocol attached to a social prescribing pathway actually reduces harm. The case for building one rests on general safeguarding practice and plain operational logic, not on social-prescribing-specific evidence. Say that plainly to any funder who asks, because they will ask.

What this does not solve

None of this addresses reach. Everyone who shows up in this pathway has already been referred by someone — a GP, a link worker, a self-referral through a known service. People who are isolated and never enter any system, who have no link worker and no GP visit that surfaces the conversation, remain entirely outside a safeguarding structure built around referral. A protocol this good only protects the people who make it through the door. It says nothing about the much larger number who never reach it.