Building a Cost Case for Social Prescribing Without Overselling It

A commissioner will ask you, at some point, what the programme saves. Not what it achieves — what it saves. You need an answer that survives scrutiny, and the honest answer is that the evidence base cannot yet give you a hard number to defend. That does not mean you have no case. It means you need a different kind of case than the one you are being asked for.

The number commissioners want, and why you cannot give it to them

The instinct is to reach for something like “social prescribing reduces GP visits by X%” and multiply by the average cost of a GP appointment. Resist it. A 2021 systematic review in Perspectives in Public Health found that of nine studies examining social prescribing and loneliness, three reported reductions in service use — GP contacts, emergency attendance, social worker time, inpatient stays. Three out of nine, with small samples and inconsistent measures across them. That is evidence of a plausible mechanism, not a costable effect.

The broader systematic review on social prescribing and wellbeing from the same year is blunter about the state of the field: it reports gains in self-esteem and confidence as the more consistently observed outcomes, and explicitly flags limited trial evidence and heterogeneity across programmes as the reason nothing firmer is available. If you present a savings figure derived from this literature as if it were settled, you are not making a stronger case — you are making a fragile one, because the first analyst who checks your citation will find the caveat you left out.

What you can defend instead

The defensible cost case has three tiers, and you should present them as tiers, not blend them into one number.

Tier one: the risk case. Isolation and loneliness are established, independent risk factors for poor health outcomes. The National Academies’ 2020 consensus report puts roughly a quarter of adults 65 and older in the socially isolated category and calls for the health system to treat this as a routine clinical variable, not a social nicety. Holt-Lunstad’s 2010 meta-analysis, covering over 308,000 participants, found weak social relationships carried a mortality risk comparable to established risk factors already built into clinical guidelines. The American Heart Association’s 2022 scientific statement puts the increased risk of heart attack, stroke, or death from either at roughly 30% for people who are socially isolated or lonely — but it also states plainly that intervention evidence is the field’s central gap. You can use this tier to justify why the population you are targeting is expensive to leave unaddressed. You cannot use it to justify a specific savings figure for your specific programme, because none of these sources measured an intervention.

Tier two: the mechanism case. A 2022 qualitative meta-synthesis in BMC Health Services Research is useful here for a specific reason: it found that participants describe benefit extending past social contact itself, into restored purpose and participation, and that structured, purposeful activity appears to work better than contact alone. That is a design finding, not a savings figure, but it tells a funder something real: this money buys structure, not just company. That distinction matters because “structure” is a thing a funder can picture staff delivering and evaluate against.

Tier three: your own data. This is the tier that actually earns you the number the commissioner wants, and it is the one most programmes skip because it is the hardest.

What this means in practice: Do not import a savings figure from a review of other people’s programmes. Build a small, honest before/after comparison from your own referrals — self-reported service contact in the three months before referral against the three months after, for people who completed the programme, reported alongside the number who dropped out and when. It will be a weaker study than anything published, but it will be about the population you actually serve, and it will be yours to stand behind.

The named failure mode: borrowed-savings syndrome

The recurring error is what might be called borrowed-savings syndrome — taking a savings percentage estimated for a different cohort, in a different health system, using a different measure of “use,” and presenting it as though it describes your programme. It is seductive because it is fast: a slide with “£X saved per participant” closes a funding conversation that a paragraph of caveats does not. It also collapses the first time a commissioner’s own analyst asks for the underlying study, finds it does not match your population, and starts discounting everything else you say.

The fix is not to drop the cost conversation. It is to separate the risk case (why this population is expensive) from the outcome case (what changed for the people you served) and be explicit that the second is thinner evidence than the first.

A table for the conversation you’ll actually have

Claim	Evidence status	How to use it
Social isolation raises health risk independent of underlying illness	Strong — multiple large meta-analyses and a national consensus report	Justify targeting and eligibility criteria
Social prescribing reduces GP or emergency service use	Weak — a minority of small studies, inconsistent measures	Flag as a hypothesis being tested, not a stated saving
Structured, purposeful activity outperforms unstructured contact	Moderate — consistent qualitative finding across studies	Justify programme design choices to funders
Your programme reduces cost for your population	Unknown until you measure it	Build your own before/after comparison; report dropout honestly

What to put in the funding proposal

State the risk case with citations. State the mechanism case with citations, framed as design rationale rather than promised outcome. Then commit to a specific, modest measurement plan for your own cohort — self-reported service contact, retention by session number, and a simple wellbeing measure — and tell the funder you will report it whether or not it flatters the programme. Funders who have sat through enough of these conversations recognise the difference between a partner offering you a number and a partner offering you a method. The method is more fundable, not less, because it is the only version of the claim that will still be true in eighteen months.

What this does not solve

None of this produces a national benchmark figure you can quote with confidence, because one does not exist yet — the field’s own reviews say so. It also does not solve the reach problem underneath any cost case: the savings you can measure come from people who completed a referral pathway, which systematically excludes the people hardest to reach and often most isolated. A cost case built honestly on your completers will always look better than the true population effect, and a good proposal says so rather than letting the funder assume otherwise.