Reach and Outcome Are Different Numbers: A Framework for Measuring Both

A social prescribing service reports 500 referrals in a year, a 40% improvement in self-reported wellbeing among people who completed the programme, and asks for renewed funding. Both numbers are true. Neither tells the commissioner what they need to know, because the report never says who the 500 were relative to who was eligible, and it never says what happened to the people who didn’t complete the programme.

This is the standard shape of a social connection evaluation, and it is missing a dimension. Reach and outcome are not two ways of describing the same success. They are answers to different questions, and a programme can be excellent on one and mediocre on the other without anyone noticing, because most reporting only asks one of them.

The two questions, and why they don’t substitute for each other

Reach asks: of the population this programme is meant to serve, who actually got in the door, and who did not? It is a denominator problem. You cannot answer it from the people who showed up; you need to know the shape of the population you started with.

Outcome asks: for the people who got in the door, what changed? It is a before-and-after problem, ideally with some counterfactual sense of what would have happened anyway.

A programme with strong outcomes and weak reach is a boutique service that works well for people who were already inclined to seek help. A programme with strong reach and weak or unmeasured outcomes is activity without evidence that it did anything. Funders have historically accepted both as sufficient on their own. Neither is.

The systematic review evidence on social prescribing illustrates the outcome side reasonably well: studies collected in a 2021 review reported increases in self-esteem and self-confidence, and a separate review of nine studies found all reported positive individual impacts, with three showing reductions in GP, emergency, social worker, or inpatient service use. That is a real signal. But almost none of that literature reports who was excluded from referral in the first place, how many referred people never attended, or how the completers differ demographically from the eligible population. The 2022 qualitative meta-synthesis on perceived benefits goes further into mechanism — participants describe restored purpose and meaningful participation, not just contact — but again this is a study of people already inside the programme.

The failure mode: the completer’s illusion

Call it the completer’s illusion. It happens when a programme reports outcomes only for people who finished, and reports reach only as a raw referral count, with no bridge between the two. The two figures never meet, so no one can ask the question that matters: did this programme work well for a narrow slice of an already-motivated population, or did it work for something close to the people it was designed to serve?

The National Academies’ 2020 consensus report on isolation in older adults is instructive here for a different reason. It calls on health systems to routinely assess isolation and loneliness as part of standard care, which implicitly makes reach a measurement obligation from the start, not something bolted on afterward. If you screen everyone, you know your denominator. If you only measure people who chose to engage, you never learn what fraction of the target population that even is. The accompanying 2020 clinical commentary makes the same point from the practitioner side: routine assessment in clinical settings requires infrastructure most services don’t have, which is precisely why reach is so rarely reported honestly — it requires counting people you never spoke to.

Building a reach denominator when you don’t have a clean population list

Most programmes object that they have no clean list of “everyone eligible” to measure against. That is usually true, but it is not a reason to skip the exercise — it is a reason to build a proxy denominator deliberately, using one of these:

1. Referral source population. If GPs, employers, or link workers are the referral pathway, get the total caseload or eligible patient list from that source, even approximately, and report referrals as a share of it.
2. Local prevalence estimate. Use a published prevalence figure for the population you serve — the AARP Foundation’s 2018 survey found one in three US adults aged 45 and older report loneliness, using the same UCLA Loneliness Scale used in the academic literature, which makes it usable as a rough local benchmark — and compare it against your enrolled cohort’s demographic profile.
3. Waitlist or declined-referral tracking. Record who was offered the programme and declined or dropped before their first session, and why, if you can get it. This is the single most informative number most programmes fail to collect.
4. Demographic gap analysis. Compare age, sex, ethnicity, and deprivation profile of enrollees against the catchment area’s census profile, not against a target population you’ve defined to flatter yourself.

None of these are exact. All of them are better than reporting a referral count with no denominator at all.

An evidence-status table for what you can currently claim

Claim	Evidence status
Social prescribing produces short-term wellbeing gains for people who complete it	Reasonably supported by multiple systematic reviews, though studies are heterogeneous and mostly small
Social connection interventions reduce downstream healthcare utilisation	Suggestive, not established — only a minority of included studies measured this, and none with strong causal design
Current programmes reach a representative slice of the eligible isolated population	Largely unmeasured — most published evaluations do not report a denominator at all
Structured, purposeful activity produces more benefit than unstructured social contact	Supported by qualitative synthesis, not yet by controlled comparison
Routine isolation screening in clinical settings is feasible at scale	Recommended by the National Academies, but implementation evidence is thin

What this means in practice: report reach and outcome side by side, on the same page, using the same cohort definitions, every time you report to a funder. If your reach number and your outcome number describe different populations — referrals for one, completers for the other — say so explicitly rather than letting the juxtaposition imply they’re the same group.

What good reporting looks like

A funder report that takes this seriously states four things, not two: the eligible or estimated target population, the number reached (with the denominator made explicit), the number who received a full dose of the intervention, and the outcome measured against that last group specifically — with a clear flag if outcome data exists only for completers. This is more work than most services currently do, and it will sometimes produce a report that looks worse than the current version, because the completer’s illusion currently flatters almost everyone who uses it.

The Surgeon General’s 2023 advisory frames loneliness as a population-level condition requiring population-level response, not an individual deficit to be treated one referral at a time. That framing only holds up if programmes can show they are reaching a meaningful share of the population, not just serving well the fraction that walks in. The mortality evidence behind the urgency — Holt-Lunstad’s 2015 meta-analysis found isolation and loneliness carry mortality odds ratios above 1.25, comparable to other major clinical risk factors — is population evidence. Programme evaluation that only measures the people already through the door cannot speak to whether that population-level risk is actually being addressed.

What this does not solve

This framework does not tell you whether your programme works. It tells you whether you can currently answer that question honestly, which most programmes cannot, because reach data is expensive and unglamorous to collect and nobody’s funding renewal has ever depended on admitting a denominator problem. It also does nothing about the deeper reach limit that no measurement fix solves: referral-based and self-selecting programmes, however well they measure their own denominator, will always reach people who are already connected to a GP, a link worker, or an employer scheme. The people hardest to reach — those with no such connection point — remain invisible to a framework that starts counting the moment someone enters a system.