How Informal Connection Programmes Handle Safeguarding Without Killing the Informality

A volunteer running a weekly walking group for isolated older adults notices that one participant has stopped washing, has lost weight since August, and mentioned in passing that she has not eaten a hot meal in four days. What is that volunteer supposed to do, and who told them?

Most connection programmes cannot answer that question in writing. They have a safeguarding policy, usually adapted from a template, that says concerns should be escalated to a designated lead. What they do not have is a definition of a concern that a non-clinical volunteer can apply in the moment, and a clear statement of what the volunteer is not responsible for. The result is predictable: volunteers either escalate nothing, because they are not sure it counts and do not want to breach someone’s confidence, or they escalate everything, which floods the lead and eventually gets ignored.

This is a training problem before it is a policy problem. And it is a specific kind of training problem, because the whole value of these programmes lies in them not feeling like services.

Why the informality is the active ingredient, not a shortcoming

The temptation, when risk surfaces, is to professionalise. Screen participants, require disclosure forms, put a paid worker in every room, log every interaction. That instinct is understandable and it will damage the thing you are running.

The qualitative meta-synthesis published in BMC Health Services Research in 2022 found that participants in social prescribing described benefit extending well beyond social contact itself — restored meaningful participation and a sense of purpose — and that structured, purposeful group activity outperformed contact alone. The register that produces that is not clinical. Ray Oldenburg’s account of third places, published in 1989, made the point that informal gathering spots work precisely because status and role are suspended in them. Eric Klinenberg’s 2018 work on social infrastructure made the parallel case at the level of buildings: libraries and recreation centres shape rates of social contact because nobody has to qualify to enter.

A group that requires an intake assessment is a service. A group that a lonely person can walk into on a Tuesday is social infrastructure. Both are legitimate, but you cannot get the benefits of the second by running it like the first.

So the design question is not “how do we make this safe” in the abstract. It is: what is the minimum risk apparatus that lets an untrained volunteer act correctly in the three or four situations that actually come up, without turning the group into an assessment?

What published programme descriptions suggest about the threshold

The intervention evidence base here is thin and you should be told that plainly. The 2021 systematic review in Perspectives in Public Health found all nine included studies reported positive individual impacts and three reported reductions in GP, emergency, social worker or inpatient service use — but nine studies is nine studies. The 2025 systematic review protocol on social prescribing for older adults notes flatly that effectiveness remains unclear despite growing adoption, with only one peer-reviewed randomised controlled trial in the area. The HEAL-HOA dual randomised trial published in The Lancet Healthy Longevity in late 2024, testing volunteering and prosocial engagement among lonely older adults in Hong Kong, is one of the few properly controlled tests that exists.

None of this literature reports on safeguarding incidents, because programme evaluations almost never do. What we have instead is the direction of travel in the health-system guidance. The National Academies concluded in 2020 that roughly a quarter of adults aged 65 and older are socially isolated and called on the health care system to routinely assess isolation and loneliness. The clinician-facing commentary in the American Journal of Geriatric Psychiatry that followed took up what routine assessment would actually require in practice — the point being that assessment is a clinical act with clinical consequences.

That gives you the boundary. If the health system is being asked to assess, your volunteers are not being asked to assess. Their job is observation and referral, and the training should say so in those words.

What this means in practice Write your volunteer role description with an explicit “not your job” section. It should say: you do not assess risk, you do not decide whether a concern is serious, you do not follow up on a concern once you have passed it on, and you do not keep a confidence that involves someone’s safety. Volunteers who know the ceiling of their responsibility escalate more, not less, because escalating is no longer an accusation of failure.

The four situations that actually come up

Rather than a generic safeguarding module, train against the small number of scenarios that recur in informal group settings. From how these programmes are commonly described, they cluster as follows.

Situation	What the volunteer does	Evidence status
Visible decline in a regular attender — weight, hygiene, cognition, mood	Note the observation, pass it to the designated lead the same day, keep attending as normal	No trial evidence; standard safeguarding practice across sectors
Disclosure of self-harm or suicidal intent	Stay with the person, do not leave them to travel home alone, contact the lead or emergency services immediately	No trial evidence specific to connection programmes
A participant who dominates, frightens, or isolates others in the group	Facilitator intervenes on group norms, not on the individual’s character; lead decides on exclusion	Weak — mostly practice consensus
Attachment forming between volunteer and participant outside the group	Disclose it, do not conceal it; lead decides whether the pairing continues	No evidence base; a governance judgement

The third row is the one most policies ignore and most facilitators lose sleep over. A single participant who makes the room uncomfortable will empty a group faster than any external risk, and the volunteer usually lacks the standing to act. Give them a script and give the exclusion decision to someone else.

The Tuesday-only problem

Here is the failure mode to name and train against. Volunteers see participants in a single, favourable, ninety-minute slice of the week. The person who arrives clean, chats, and leaves is invisible to you for the other 166 hours. Programmes routinely conclude that nobody in their group is at risk, when what they mean is that nobody presents as at risk during the one window in which they are observed.

Call it the Tuesday-only problem, and build two things around it. First, train facilitators to treat absence as information. Three consecutive non-attendances from a regular is the single most actionable signal you will get, and most groups treat it as churn. Second, ask the referral route — a link worker, a GP practice, a family member — to tell you what they are worried about at the point of referral, so the facilitator is not starting from a blank sheet. This does not require an assessment. It requires one sentence.

Volunteer risk runs the other way too

The people running these groups are often themselves drawn from the population the programme serves, and that is deliberate. It is also a duty of care you have taken on. AARP’s 2018 national survey of 3,020 adults aged 45 and older, using the 20-item UCLA scale, found one in three lonely; the 2025 follow-up, comparable because the instrument stayed consistent, reports the figure rising. Harvard’s Making Caring Common survey in 2021 found 36% of Americans reporting serious loneliness, and about half of lonely young adults said nobody had spent more than a few minutes recently asking how they were in a way that felt genuine.

Your volunteers are in those numbers. A supervision structure that only reviews participant risk and never asks the facilitator how they are is incomplete. Thirty minutes, monthly, with someone who is not their line manager.

Screening in the matching-app category, and why it is not your model

Programme managers increasingly get asked why they cannot just point people at an app. Meetup, running since 2002 and reporting around 60 million members, hosts recurring interest groups with organiser-level moderation. Timeleft seats strangers at weekly dinners matched by personality, with no messaging before the meal — a deliberate design choice with a safeguarding logic behind it. Buffet, covered by TechCrunch in April 2024, matches a person with one other person and a venue based on five interests and a short personality questionnaire, launched in Los Angeles.

These are reasonable options for confident adults choosing their own risk. They are not a referral destination for someone a link worker has flagged as vulnerable, because none of them carries a duty of care toward a named individual and none has a designated lead you can call. Knowing the difference is part of what you should be training link workers to hold.

What this does not solve

Everything above assumes the person is already in the room. That is the harder limit. The WHO Commission on Social Connection reported in June 2025 that one in six people worldwide is affected by loneliness, with an estimated 871,000 deaths a year linked to it; the Surgeon General’s 2023 advisory put roughly half of U.S. adults as experiencing loneliness, and CDC surveillance published in 2024 links disconnection to heart disease, stroke, dementia, diabetes, depression and premature mortality. Good safeguarding practice makes your group safer for the forty people who come. It does nothing for the person who has not left the house since March, and no amount of facilitator training will reach them. That is a referral-route problem and an outreach problem, and it should be funded as a separate line.