Building the Cost Case for Social Prescribing Without Overselling It

Every social prescribing service that has survived past its first funding cycle has been asked, at some point, to show it reduces GP visits or A&E attendances. It’s the natural question for a commissioner facing a budget line. It is also a question the published evidence cannot fully answer, and services that pretend otherwise are building a cost case that will collapse under the first serious audit.

Here is what a class of programme — link worker and social prescribing services operating largely within or alongside the UK’s NHS, the system that has invested most heavily in this model — has actually had to work with, and how the more careful ones have approached it.

The claim commissioners want and the claim the evidence supports

The UK’s 2018 loneliness strategy folded social prescribing into a wider national approach and gave it institutional backing, which is why so much of the published evidence on this model comes from UK settings. That backing created pressure to show return on investment quickly, and the shorthand version of the pitch became: connect people to community activity, reduce their loneliness, reduce their demand on primary and emergency care, save money.

Each link in that chain is separately plausible. None of them is proven end to end.

A 2021 systematic review of social prescribing’s effect on loneliness looked at nine studies and found every one reported some positive individual impact. That is the good news, and it is real. But only three of those nine reported any reduction in downstream service use — GP contacts, emergency attendances, social worker time, or inpatient stays — and the review is explicit that the evidence base is thin and inconsistent across programmes. A companion review of well-being outcomes from the same year reports the most consistently observed benefits are increases in self-esteem and self-confidence, not utilisation. That is a genuinely different outcome from “fewer GP appointments,” and a genuinely different one to put in front of a funder.

A 2022 qualitative synthesis adds a useful qualification: participants describe the benefit as going beyond social contact itself, toward restored participation and purpose. Structured, purposeful activity appears to do more than contact alone. That is useful for programme design. It is not evidence of cost.

The failure mode: borrowing someone else’s mortality statistic

Call this the smoking-equivalence trick. It works like this: a business case cites the U.S. Surgeon General’s 2023 advisory, which states that the mortality risk of chronic social disconnection is comparable to smoking up to 15 cigarettes a day, or cites Julianne Holt-Lunstad’s 2010 meta-analysis finding a 50% increased likelihood of survival associated with stronger social relationships, or her 2015 review putting the mortality odds ratio for isolation at 1.29. The business case then implies that a twelve-week befriending programme will move a cohort’s health outcomes by something like that magnitude, and prices the projected NHS savings accordingly.

This is a category error, not a rounding error. Those figures come from population-level observational epidemiology comparing people with strong versus weak social ties over years or decades. They say nothing about what a bounded, time-limited intervention delivers, because — as the American Heart Association’s 2022 scientific statement says plainly — the intervention evidence for social isolation and cardiovascular outcomes essentially does not exist yet. The AHA statement identifies that absence as the central research gap in the field, from a body reviewing cardiovascular and brain health specifically. If the people producing the underlying epidemiology say the intervention trials haven’t been done, a local social prescribing service should not be the one claiming they have.

The National Academies’ 2020 consensus report on older adults, and the clinician-facing commentary that followed it, make a related point from the health systems side: they call for routine assessment of isolation and loneliness in care settings, but they frame this as a starting point for building better evidence, not as validation that any given referral pathway already works. Citing that report to say “the National Academies recommends this” is fair. Citing it to say “the National Academies proved this saves money” is not what it says.

An evidence-status table for the claims commissioners actually make

Claim	Evidence status
Social prescribing improves self-reported well-being, self-esteem, and confidence	Reasonably well supported across multiple reviews
Social prescribing reduces loneliness for participants	Supported in individual studies, but with small samples and no consistent comparator
Social prescribing reduces GP or A&E attendance	Reported in a minority of included studies (3 of 9 in one review); not a reliable general finding
Social connection interventions reduce mortality or cardiovascular risk	No adequately controlled intervention trial exists at this scale; population epidemiology only
Volunteering-style prosocial engagement reduces loneliness in a controlled trial	Supported by one dual RCT in older adults in Hong Kong — genuinely rare evidence in this field
Social prescribing pays for itself in avoided health system cost	Not established; treat any such figure in a funding bid as an estimate, not a finding

That last row is the one worth sitting with. The honest position, as of this evidence base, is that social prescribing has a real and reasonably supported case on subjective well-being, and an unproven case on cost avoidance. Those are not the same pitch, and funders who have seen enough business cases can usually tell which one they’re being given.

What an honest cost case looks like instead

1. Lead with the outcome the evidence actually supports. Report change in loneliness scores, self-esteem, or confidence measures, using validated instruments where possible, rather than projected NHS savings.
2. Cite service-use reduction only where you have your own comparator data, and describe it as a local finding, not a general property of social prescribing. The published reviews show it happens in some programmes, not in all of them.
3. Use the mortality and cardiovascular literature to justify why the problem matters, never to size the financial benefit of your specific programme. It is legitimate context. It is not a savings calculation.
4. Flag the trial gap to the funder directly. The 2024 Lancet Healthy Longevity trial of volunteering among lonely older adults in Hong Kong is one of the only randomised designs in this space; naming that scarcity signals that you understand the evidence rather than that you’re hoping the funder doesn’t check.
5. Separate isolation reduction from loneliness reduction in your own reporting. They are different constructs with different drivers, and conflating them in an evaluation makes the resulting numbers harder to defend later.

What this means in practice: if a business case for social prescribing includes a specific projected saving to the health system, ask what study that figure came from. If the answer traces back to a mortality meta-analysis or a national loneliness prevalence survey rather than a controlled evaluation of a comparable programme, the number is a guess wearing a citation.

What this does not solve

None of this tells you whether social prescribing reaches the people who need it most. The published studies are drawn overwhelmingly from people who accepted a referral and stayed engaged — a self-selected group by definition. The commissioning conversation about cost is a separate problem from the reach conversation, and building a more honest financial case does nothing to fix a programme that only ever serves people already connected enough to show up to the GP practice in the first place.