The Cost Case for Social Connection Programmes, Built Honestly

A commissioner asking whether to renew a social prescribing contract will ask, at some point, for a number: what does this save. The temptation is to reach for a mortality statistic — Julianne Holt-Lunstad’s finding that weak social ties carry a mortality risk comparable to smoking, or the World Health Organization’s 2025 estimate that loneliness contributes to roughly 871,000 deaths a year — and attach it to a programme that runs a weekly walking group for forty people. That is the wrong number for the wrong question, and funders who have seen enough of these submissions know it.

The cost case for a connection programme has two separate legs, and they rest on evidence of very different quality. Confusing them is the single most common way these business cases fail scrutiny.

The two legs, and why they don’t support each other

Leg one: the cost of loneliness and isolation as a population phenomenon. This is well established. Holt-Lunstad’s 2010 meta-analysis of 308,849 people found stronger social relationships associated with a 50% greater likelihood of survival, an effect size comparable to established risk factors like smoking and obesity. Her 2015 review put isolation’s mortality odds ratio at 1.29 and loneliness at 1.26. The American Heart Association’s 2022 scientific statement found isolation and loneliness associated with roughly 30% higher risk of heart attack, stroke, or death from either. Cigna’s 2020 workplace survey found lonely workers miss work twice as often for illness and five times as often for stress. These are population-level associations from large samples, and they are about as solid as this literature gets.

Leg two: what a specific intervention changes in a specific population. This is where the evidence thins out fast. A 2021 systematic review in Perspectives in Public Health looked at nine studies of social prescribing and loneliness: all nine reported positive individual impacts, but only three reported any reduction in GP, emergency, social worker, or inpatient service use — and the review does not let readers separate a rigorous trial from a satisfaction survey when reading that count. A 2025 review protocol on social prescribing for older adults notes bluntly that only one peer-reviewed randomised controlled trial exists in this entire area, despite social prescribing’s rapid adoption across health systems. The National Academies’ 2020 report calling on health systems to screen for isolation is explicit that screening tools are ahead of intervention evidence, not behind it.

Leg one tells a funder that disconnection is expensive at population scale. It says nothing about whether your six-week programme moves any of those numbers. Treating leg one as proof of leg two is the failure mode worth naming.

The borrowed-statistic problem

Call it that: taking a large, well-supported population statistic and implicitly claiming it as the return on a small, specific intervention. It shows up as sentences like “loneliness costs the NHS £X per year, and our programme addresses loneliness, therefore our programme saves £X” — with no bridge between the two. A funder who has read even one systematic review in this space will notice the missing bridge immediately, and the whole submission loses credibility, including the parts that were genuinely defensible.

What an honest cost case actually contains

1. State leg one as context, not as your result. Use it to establish why the problem matters — the AHA’s cardiovascular findings, or the Surgeon General’s 2023 advisory framing loneliness’s mortality risk as comparable to up to 15 cigarettes a day — and stop there. Do not carry these numbers into your savings line.

2. Report leg two at the evidence quality it actually has. A 2022 qualitative meta-synthesis in BMC Health Services Research found that participants describe benefit extending beyond social contact to restored purpose and meaningful participation — a real finding, but qualitative and about lived experience, not service use. A 2021 systematic review of well-being outcomes reports increases in self-esteem and confidence, again with the same caveat: limited trial evidence, high heterogeneity across programmes. Cite these as what they are — indications, not proof — and say so explicitly rather than letting the funder assume otherwise.

3. If you claim service-use reduction, cite the specific fraction of studies that found it, not the review’s overall verdict. Three of nine studies in the 2021 systematic review, not “social prescribing reduces GP visits.”

4. Separate the isolation problem from the loneliness problem in your own outcome measurement, since they respond to different things and are tracked separately in the recent literature. A programme that gets someone into more social contact but does not touch their subjective loneliness will show up as a win on one measure and a null result on the other. Report both.

Claim	Evidence status
Loneliness and isolation carry population-level mortality and morbidity risk	Strong — multiple large meta-analyses and a formal AHA scientific statement
Lonely workers cost employers more in absenteeism	Reasonably strong — single large industry survey, not independently replicated
Social prescribing improves participants’ self-reported wellbeing and confidence	Moderate — consistent across qualitative reviews, but few controlled comparisons
Social prescribing reduces GP, A&E, or inpatient use	Weak — a minority of studies in existing reviews report this, and none isolate causation
Social prescribing for older adults is supported by randomised evidence	Very weak — one peer-reviewed RCT exists as of mid-2025

What this means in practice: Present the mortality and morbidity data to establish why the problem is worth funding at all. Present your own programme’s outcomes — participation, retention, self-reported wellbeing — as a separate line, using the language of association rather than saving. If you want to claim service-use reduction, commission or point to a study designed to measure it; do not infer it from a mortality statistic that was never about your programme.

What this does not solve

None of this produces the single number a funder actually wants. The honest version of a cost case is slower to build and less impressive on a single slide, because the intervention literature genuinely is thin — the American Heart Association’s own statement names the absence of intervention evidence as the field’s central research gap, not a detail to be smoothed over. It also does not solve the reach problem that sits underneath every social prescribing evaluation: the studies being cited, weak or strong, are almost always about people who were referred into a service and showed up for it. What the evidence says about that population tells you very little about the people never referred in the first place, and no cost case built from these sources should claim otherwise.