When a Funder Asks for Evidence Your Programme Cannot Produce

A commissioner asks your social prescribing service to demonstrate a reduction in GP attendance, A&E visits, or mortality risk. You have a waiting list, a satisfaction survey, and maybe a pre/post loneliness score. That gap is not a reporting problem you can fix with a better slide. It is a mismatch between what the literature actually supports and what you have been asked to prove.

The mismatch, precisely

The observational evidence linking social connection to health is genuinely strong. Julianne Holt-Lunstad’s 2010 meta-analysis, covering more than 300,000 participants, found stronger social relationships associated with a 50% increased likelihood of survival. Her 2015 follow-up put the mortality odds ratio for social isolation at 1.29 and for loneliness at 1.26. The American Heart Association’s 2022 scientific statement put the increased risk of heart attack, stroke, or death from either at roughly 30%.

None of that is intervention evidence. It tells you that disconnected people fare worse. It does not tell you that a 12-week befriending scheme, a walking group, or a referral pathway changes that trajectory for the people you enrol. The AHA statement says so directly: it names the absence of intervention evidence as the central gap in the field, not a footnote to it. A 2025 systematic review protocol on social prescribing for older adults found exactly one peer-reviewed randomised controlled trial in the area, despite the model’s near-universal adoption across health systems.

That is the ask you are actually being handed: prove a distal, causally-loaded outcome (hospital use, mortality) using an evidence base that has barely established causal effects on the proximal outcome (loneliness itself).

The borrowed effect size

Programme reports that cite Holt-Lunstad’s 50% figure, or the AHA’s 30%, as though it describes what their own service achieved are committing a specific and common error: borrowing someone else’s effect size. The number is real. It describes an association between existing social connection and mortality in the general population. It says nothing about what a twelve-week intervention run by a link worker in one borough will do. Funders sometimes make this mistake themselves, and pass the borrowed number back to you as a target. Do not accept the target.

What the intervention literature actually supports

Claim	Evidence status
Social disconnection predicts worse health and higher mortality risk	Strong — large meta-analyses, consistent across decades
Social prescribing improves self-esteem, confidence, and sense of purpose	Moderate — repeatedly reported in qualitative and mixed-methods reviews
Social prescribing reduces loneliness scores on validated scales	Weak but improving — a small number of RCTs, mostly in older adults and residential care
Social prescribing reduces GP attendance, A&E use, or hospital admission	Weak — a handful of studies report this, evidence is heterogeneous and mostly observational
Social prescribing reduces mortality	Absent — no trial evidence exists at this outcome level

Two randomised trials are worth naming because they are rare. The HEAL-HOA trial tested volunteering against a control among lonely older adults in Hong Kong. A 2025 trial of befriending in aged care found reductions of 2.39 points on the UCLA Loneliness Scale at eight weeks and 2.71 points at sixteen, against control — and noted that structured psychological approaches outperformed simple befriending head-to-head. That is the shape of evidence you can defend: a validated instrument, a comparison group, a modest and specific effect. It is nothing like a mortality claim.

Qualitative work adds useful texture without adding causal weight. A 2022 meta-synthesis found participants describe benefit extending beyond contact itself to restored participation and purpose, and that structured, purposeful activity seems to work better than unstructured contact. A 2021 systematic review of social prescribing and loneliness found all nine included studies reported positive impacts, with three also reporting reduced service use — but the review’s own caveat is that heterogeneity across programmes makes generalising from that difficult.

What to do with the ask

1. Separate the outcome levels explicitly, in writing, before you agree to anything. Distal (hospital use, mortality) versus proximal (loneliness score, self-reported confidence, network size).
2. Decline the distal target, or reframe it as a hypothesis, not a KPI. Cite the AHA statement’s own admission of the intervention gap if you need external authority to do this.
3. Offer the proximal outcome instead, measured properly. Use a validated instrument — the UCLA Loneliness Scale is the one most trials use, which makes your numbers comparable to the wider literature rather than a bespoke score nobody else can benchmark against.
4. Report the качественный layer as texture, not proof. Participant accounts of restored purpose are worth including in a funder report. They are not evidence of population-level health effect.
5. Name the gap rather than papering over it. Funders spending real money respond better to “the evidence for this exact outcome doesn’t exist yet, here is what does exist” than to a number that later gets audited and doesn’t hold up.

What this means in practice: if a commissioner asks for hospital use or mortality reduction, negotiate down to a validated loneliness measure with a pre-post or comparison design, and say plainly that the intervention literature does not yet support the bigger claim. That is a stronger position than promising a number you cannot produce.

What this does not solve

This approach gets you an honest scope, not a bigger evidence base. The RCT literature for social prescribing remains thin — one trial for an entire field of practice, according to the 2025 protocol review — and nothing in this note changes that. Nor does it address who the number describes: trial and programme evaluation samples, like referral pathways generally, mostly capture people who were already in contact with a GP, a link worker, or a community service. People who never reach that door are absent from both the evidence and the target you eventually agree on.