Building a Cost Case for Connection Training That Survives Scrutiny

A commissioner asks for the return on training your link workers in connection-focused conversation skills. Someone on your team reaches for the statistic that social disconnection carries mortality risk comparable to smoking 15 cigarettes a day, cited in the Surgeon General’s 2023 advisory, and puts it on slide two as the justification for a £40,000 training budget. This is the moment the cost case falls apart, and it falls apart in front of the one audience that will notice: a funder who has seen a hundred slide decks and knows the difference between a population statistic and an intervention effect.

The mortality and cardiovascular numbers are real. The Surgeon General’s advisory cites a mortality risk comparable to heavy smoking. The American Heart Association’s 2022 scientific statement puts the increased risk of heart attack, stroke, or death from either at roughly 30%. Julianne Holt-Lunstad’s 2010 meta-analysis, covering more than 300,000 participants, found stronger relationships associated with a 50% greater likelihood of survival. None of these numbers tell you what happens when you train a link worker to run a better first conversation. They describe the health consequences of chronic isolation and loneliness at a population level, measured over years, in people who were not participants in a training programme. Borrowing them to justify a training spend is a category error, and it is the single most common one in this sector’s funding applications.

The borrowed statistic

Call this the borrowed statistic problem: reaching for the largest, most alarming number in the loneliness literature and attaching it to a much smaller, much more specific intervention, on the assumption that the size of the problem justifies the size of the ask. It doesn’t. A funder can hold two facts at once — that loneliness is a serious population health issue, and that your training programme has not been shown to move the needle on it — and the second fact is the one they’re actually deciding on.

The honest fix is not to drop the population statistics. It’s to stop using them as the headline number and instead use them as context, then build the actual cost case from evidence that sits closer to what the training does.

What the training actually changes

Before you can cost anything, you need to separate three layers that get collapsed into one slide far too often:

1. What the training changes directly — staff confidence, conversation quality, referral accuracy, follow-through on connection to activity.
2. What that produces for the person referred — engagement with the activity, a sense of purpose or restored participation, which a 2022 qualitative meta-synthesis on social prescribing found participants describe as more valuable than social contact alone.
3. What that might eventually produce for the system — reduced GP contacts, fewer emergency attendances, lower social care demand.

Layer 1 is measurable directly and quickly. Layer 2 has moderate evidence behind it. Layer 3 is where almost all the money-saving claims live, and it’s also where the evidence is thinnest.

A systematic review of social prescribing and loneliness found that three of nine included studies reported reductions in GP, emergency, social worker, or inpatient service use. That is worth citing — but three of nine is not “social prescribing reduces NHS costs,” and a funder who checks the source will find that out. A separate systematic review of well-being outcomes reports gains in self-esteem and self-confidence as the more consistently observed effect, alongside a direct note that the trial evidence is limited and the programmes studied are heterogeneous. Use that limitation in your own document before the funder finds it. It reads as competence, not weakness.

An evidence-status table for the claims you’ll want to make

Claim	Evidence status
Loneliness and isolation carry serious mortality and cardiovascular risk	Strong — multiple meta-analyses and a national scientific statement
Structured, purposeful activity produces more benefit than social contact alone	Moderate — consistent qualitative finding across reviews
Social prescribing improves self-esteem and confidence	Moderate — repeated finding, but few controlled trials
Social prescribing reduces GP or ED utilisation	Weak — reported in a minority of included studies, mechanism unclear
Training frontline staff in connection skills reduces downstream healthcare costs	Unestablished — no direct evidence chain exists; this is an inference stacked on two other inferences
Routine isolation screening in health settings is feasible and useful	Emerging — recommended by the National Academies, but implementation guidance is still being worked out

The bottom row matters because it’s the one commissioners increasingly ask about. The National Academies’ 2020 consensus report calls for the health care system to routinely assess isolation and loneliness, and a follow-up clinical commentary discusses what that would actually require in practice — but “the health system should screen for this” is a recommendation about future practice, not evidence that it currently saves money.

What this means in practice: build your cost case around the layer you can actually measure — referral quality, activity engagement, self-reported confidence and purpose — and present any utilisation or cost-saving claim as a hypothesis the programme is testing, not a result it has already delivered. Funders fund credible hypotheses. They eventually stop funding organisations whose promised savings never show up in year two.

How to sequence the argument

1. State the population problem in one sentence, with one citation. The AHA’s newsroom summary of its 2022 statement is a serviceable, plain-language version if you need something citable and non-technical.
2. Name what the training changes, and commit to measuring it directly — pre/post confidence, observed conversation quality, referral follow-through rates. This is your near-term evidence, and it’s the only part of the case you fully control.
3. Attach the intermediate outcome literature, clearly labelled as moderate evidence, to argue why staff behaviour change should matter for the person referred.
4. Present the cost-saving possibility as a testable proposition, with a stated evaluation plan, rather than a projected saving. If you have baseline utilisation data, say you will track it. Do not forecast a reduction you cannot yet support.
5. State the AHA’s own limitation explicitly — the scientific statement identifies the absence of intervention evidence as the central gap in this entire field. Citing that gap yourself, rather than hoping the funder doesn’t ask about it, is what makes the rest of the document credible.

What this does not solve

A cost case built this way will be more honest, but it will also be smaller and slower than the one built on borrowed mortality statistics — and that’s a real trade-off if you’re competing for a grant against an application that overpromises. There’s no clean answer to that; overselling wins some funding rounds. What an honest cost case buys you is a second and third round. Funders and commissioners increasingly notice when promised savings don’t materialise, and organisations that were candid about the limits of their evidence the first time keep more credibility on the second ask than organisations that weren’t.

This note also doesn’t solve the deeper measurement problem in the field. A 2023 review of loneliness and isolation research names inconsistent measurement across studies as a structural barrier to comparing findings at all — which means even a well-built cost case is working with a literature that doesn’t fully agree on how to count the thing it’s trying to change. And none of this addresses reach: a training programme that improves how link workers talk to the people already referred to them says nothing about the people who never get referred in the first place, who are disproportionately the most isolated.