What Social Prescribing Reports Leave Out When They Report Attrition

A 2021 systematic review of social prescribing and loneliness looked at nine studies. All nine reported positive impacts. If you run programmes for a living, that number should worry you more than it reassures you. No category of intervention in social science works for 100% of the people who try it. When a literature says it does, the likely explanation is not that social prescribing is uniquely effective — it is that the people for whom it didn’t work have already left the data before anyone measured the outcome.

This is the attrition problem, and it is not specific to social prescribing. It shows up in any programme that connects isolated people to groups, activities, or community contact and then reports on “participants.” The question a funder should always ask is: participants at what point?

Why the dropouts are not a random subset

Attrition in connection-focused programmes is not evenly distributed. The people most likely to disengage are frequently the people the programme most needed to reach: those with smaller, less diverse social networks, and those who are physically isolated rather than just socially uncomfortable. The AARP Foundation’s 2018 survey of adults 45 and older found that network size and diversity, along with physical isolation, were the strongest predictors of loneliness — and those same characteristics predict who struggles to sustain attendance at a weekly group, a walking club, or a befriending scheme. The National Academies’ 2020 consensus report on isolation in older adults makes a related point in the other direction: isolation is often invisible to the systems meant to catch it, precisely because isolated people are less likely to be in contact with anyone who would notice they’d stopped showing up.

Put those together and the mechanism is obvious. A programme measures wellbeing at intake and again at week twelve, among people who are still attending at week twelve. The people who dropped out at week four — plausibly the more isolated, harder-to-reach group — are gone from the denominator. The remaining group looks like a success story because it is now made up disproportionately of people who were doing better to begin with.

Call this the completers’ picture. It is a photograph of the people who stayed, presented as a photograph of the programme.

What the published evidence actually supports

The two systematic reviews of social prescribing and loneliness available by early 2024 are honest about this in their limitations sections, even where the framing of results is not. The 2021 review in Perspectives in Public Health reports that all nine included studies found positive individual impacts, and that three reported reductions in service use — GP visits, emergency contact, social worker or inpatient use. Read the finding at face value and it looks like a strong case for commissioning. Read the same review’s caveats and it is closer to: small samples, heterogeneous designs, and a literature dominated by studies that report on people who completed rather than people who were referred.

The 2021 review in the International Journal of Environmental Research and Public Health covering wellbeing outcomes more broadly makes the heterogeneity problem explicit — it flags limited trial evidence and inconsistent outcome measures across the studies it covers, which makes any single effect size close to meaningless as a benchmark. A 2023 review in BMC Public Health of the wider loneliness and isolation research base makes the same point at scale: inconsistent measurement across studies is a structural barrier to comparing findings, not a footnote.

None of this means social prescribing doesn’t work. A 2022 qualitative meta-synthesis in BMC Health Services Research found that participants who stayed with structured, purposeful group activity described benefit that went beyond contact itself — restored participation, restored purpose. That is a real and useful finding about what works for people who engage. It is not a finding about the programme’s reach or its overall effect, and the two get conflated constantly in commissioning documents.

Evidence status, stated plainly

Claim	Evidence status
Social prescribing improves wellbeing for people who complete the referral pathway	Reasonably supported, though effect sizes vary widely across small, heterogeneous studies
Social prescribing reduces GP/emergency/inpatient service use	Suggestive — reported in a minority of included studies, not yet a settled finding
Social prescribing works equally well regardless of how isolated the referred person is at intake	Not supported — no published social prescribing evidence directly tests this, and the mechanism of attrition argues against it
Structured, purposeful activity produces more benefit than unstructured contact	Supported by qualitative synthesis, though based on people who stayed engaged

What an honest attrition report actually contains

If you commission or run a connection programme, ask for four numbers before you accept a headline outcome figure, and expect a shrug or a vague answer if the provider hasn’t tracked them:

1. Number referred or enrolled at the start.
2. Number who attended at least once.
3. Number still engaged at the point outcomes were measured.
4. What is known, even anecdotally, about why the gap between (1) and (3) exists — cost, transport, mismatch with the activity, or the loneliness itself making sustained attendance hard.

That fourth number rarely exists in a form you can cite, and that absence is itself the finding worth reporting to a funder. Say so, rather than quietly reporting only on completers and letting the report imply the programme reached everyone it touched.

What this means in practice: if a programme report gives you a single outcome figure without a completion rate next to it, treat the figure as describing the people who stayed, not the people who were offered help. Ask for the denominator before you ask for the effect size, and build reporting on attrition into the contract from the start rather than requesting it retroactively once someone notices the gap.

What this does not solve

None of this fixes the underlying reach problem. Even a scrupulously honest attrition report only tells you what happened to the people who were referred in the first place — and referral itself is selective, running through GPs, link workers, and community organisations that tend to already be in contact with people who are somewhat connected to services. The people social prescribing evaluations most struggle to say anything about are the ones who were never referred at all: isolated enough to be invisible to the systems that would have sent them somewhere. Honest attrition reporting narrows the gap between what a programme claims and what it did. It does not close the much larger gap between what the programme did and who needed it.