When to Stop Running a Connection Programme

Most guidance on connection programmes is about starting one. Almost none of it tells you when to close one down. That asymmetry costs money and, more importantly, it costs participants’ time when they’d be better off somewhere else.

This is a framework for the harder decision: when a programme that is running fine, has willing participants, and produces warm anecdotes should still be wound down, redesigned, or handed to someone else.

Why “it’s working” is not a reason to continue

Ask a programme manager whether their walking group or peer-support circle is working, and you will almost always get yes. Attendance is decent. People say nice things. Staff like running it. None of that answers the question a funder should be asking, which is whether this is the best use of this money now, compared with the alternatives.

The social prescribing literature illustrates the trap. A 2021 systematic review found that social prescribing initiatives consistently produced self-reported gains in self-esteem and confidence — but noted the underlying trial evidence is thin and inconsistent in design, making it hard to know how much of that improvement is the intervention versus the attention. A separate 2021 review of loneliness-specific social prescribing found all nine included studies reported positive impacts, and three found reductions in service use. That is a genuinely encouraging pattern. It is also nine studies, self-selected participants, and no control groups. Programmes citing this evidence to justify indefinite continuation are citing more certainty than the research supports.

The 2022 qualitative synthesis adds a sharper point: participants describe benefit coming from restored purpose and meaningful participation, not contact for its own contact. Structured, purposeful activity outperforms unstructured mixing. That distinction matters for the stop decision, because it means a programme can keep people showing up while delivering less and less of the thing that actually helped them.

The four questions that should trigger a review

Run these on a fixed schedule — annually at minimum, and at any funding renewal point. Do not wait for a crisis.

1. Has attendance plateaued below capacity for two consecutive reporting periods? A programme that isn’t filling its own seats is not undersubscribed by chance. Something about design, referral pathway, or fit has stopped working, and more marketing rarely fixes it.
2. Is the comparison group missing, and if so, can you get one? The one randomised trial available in this space — a 2024 dual RCT of volunteering against a control among lonely older adults in Hong Kong — exists precisely because uncontrolled evaluation is the norm and it systematically overstates effect. If you have run three years without ever testing against a comparison, you don’t know if you have a programme or a placebo.
3. Are you measuring loneliness, isolation, or attendance — and do you know which one you’re claiming to move? The American Heart Association’s 2022 scientific statement is explicit that the evidence connecting social isolation to cardiovascular and brain health is strong, but the evidence for which interventions actually reduce isolation is not there yet. A 2023 review of the loneliness and isolation research base makes the same point at a field level: measurement is inconsistent enough that outcome claims often don’t survive scrutiny. If your monitoring data can’t distinguish “people came” from “people felt less lonely,” you cannot justify continuation on outcome grounds — only on output grounds, which is a different and weaker argument to a funder.
4. Would the same budget, spent on a different design for the same population, plausibly do more? This is the question programmes almost never ask themselves, because it requires comparing your own effort unfavourably to something you’re not running.

The evidence-status check before you renew

Claim commonly used to justify renewal	Evidence status
“Social prescribing reduces loneliness”	Positive signal across small studies; no strong causal evidence; self-selected samples
“Participants report higher self-esteem”	Reasonably well supported at the individual, self-report level
“The programme reduces GP or service use”	Reported in a minority of studies; not a safe default claim
“Isolation is a serious health risk worth addressing”	Strongly supported — this is not the contested part
“Our specific intervention reduces isolation”	Rarely tested with a control group; treat as unproven until you’ve checked

What this means in practice: separate “loneliness and isolation matter” — which is settled — from “this specific programme reduces them” — which usually isn’t tested. Funders should stop accepting the first as evidence for the second.

The renewal-by-default failure mode

Call this the evergreen grant problem: a programme gets funded because a need is real, runs for several cycles because no one wants to be the funder who cut a well-liked service, and survives review after review because the review criteria never actually ask whether it’s the best option available — only whether it’s doing harm. The WHO Commission on Social Connection’s 2025 report frames loneliness as costing an estimated 871,000 deaths a year globally, which is exactly the kind of number that makes cutting anything in this space feel reckless. That is precisely why programmes need a stop test that isn’t about the scale of the problem, which is not in dispute, but about the marginal value of this specific use of funds.

What to do instead of extending

If a programme fails the review, the options are not just “continue” or “kill.” Consider handing the population to a lower-cost universal channel — the National Academies’ 2020 report recommends routine isolation screening within existing health care contact points rather than standalone programmes, which is often cheaper per person reached. Consider a fixed-term redesign with a built-in control group, using the HEAL-HOA trial design as a template rather than another uncontrolled pilot.

What this does not solve

This framework helps you decide whether to keep funding something that already exists and already has participants. It says nothing about the much larger group who never got referred, never walked in, or don’t know the programme exists at all. Stopping a weak programme frees money; it does not, on its own, reach anyone new.