Social Prescribing: What the Evidence Actually Supports

Social prescribing is the practice of a health worker referring a patient not to a drug or a specialist but to a non-clinical activity: a walking group, a gardening scheme, a befriending service, an art class. A link worker usually manages the referral and follows up. The idea is that some of what shows up in a GP’s office as a health complaint is, at root, a problem of disconnection, and that the correct response is social rather than pharmaceutical.

The United Kingdom built social prescribing into national policy with its 2018 loneliness strategy, which funded link workers as part of a wider effort to embed loneliness measurement into the health system. It is now a recognisable feature of primary care in England, and versions of it appear in other countries’ connection strategies, including the framework the U.S. Surgeon General laid out in the 2023 advisory on loneliness and isolation. The National Academies’ 2020 consensus report on isolation in older adults endorsed the general principle that health systems should be doing more to identify and respond to social deficits, and social prescribing is the mechanism most often proposed for the “respond” half of that sentence.

The question this piece asks is narrow: does the trial evidence support social prescribing as a policy, or has policy run ahead of it?

What the reviews actually find

Two systematic reviews published in 2021 looked directly at social prescribing’s effect on loneliness and wellbeing. Both found something, but not what a strong policy claim requires.

The review published in the International Journal of Environmental Research and Public Health reported increases in self-esteem and self-confidence among participants as a key outcome, but it also flagged limited trial evidence and substantial heterogeneity across programmes — different referral criteria, different activities, different follow-up periods, pooled together because there was not enough of any single design to analyse separately. A second review, in Perspectives in Public Health, looked specifically at loneliness outcomes across nine included studies and found that all nine reported positive individual impacts, with three reporting reductions in use of GP, emergency, social worker, or inpatient services. Nine studies, uniformly positive, is the kind of result that should invite scrutiny of publication bias and study design before it is treated as settled. Programmes that fail tend not to generate papers.

A 2022 qualitative meta-synthesis in BMC Health Services Research adds a different kind of finding, one that matters more than it might look. Participants described benefit that went beyond social contact itself — restored meaningful participation, a sense of purpose, feeling of use again. The synthesis suggests structured, purposeful group activity performs better than contact alone. That is a real result, but it is a result about what kind of intervention works, not evidence that social prescribing as a general referral pathway reliably produces it. A link worker referring someone to a gardening group that folds within six weeks is not delivering the same intervention as one referring them to a gardening group that has run for a decade with a stable membership. The reviews mostly cannot distinguish between these outcomes because the underlying studies rarely report programme durability.

None of this is nothing. It is evidence of promise, gathered mostly from small, heterogeneous, often uncontrolled studies, with self-reported outcomes and short follow-up windows. It is not the kind of evidence — randomised, adequately powered, with outcomes tracked well past the intervention period — that would justify treating social prescribing as an established, cost-effective health intervention rather than a plausible one.

The gap the cardiology literature names explicitly

The American Heart Association’s 2022 scientific statement on social isolation and cardiovascular and brain health is useful here precisely because it is not about social prescribing at all. It reviews the observational evidence linking isolation and loneliness to a roughly 30% increased risk of heart attack, stroke, or death from either, and to worse prognosis among people who already have cardiovascular disease. Then it says, plainly, that the central research gap is the absence of intervention evidence. The statement does not claim that any specific programme reduces cardiovascular risk by treating loneliness. It says the field does not yet know whether reducing loneliness reduces that risk, because almost nobody has run a trial designed to test it.

This matters for social prescribing specifically. The observational case for social connection as a health-relevant exposure is strong; the American Journal of Lifestyle Medicine review by Julianne Holt-Lunstad in 2021 argues connection belongs in preventive frameworks alongside diet, exercise, and smoking cessation, precisely on the strength of that observational base. But an exposure being real and consequential does not mean any given intervention that targets it will work, at the scale and cost a health system can sustain. The AHA statement is a reminder that the size of the problem and the strength of a proposed fix are two separate questions, and only one of them is well answered.

What the commentary literature already flags

The clinician-facing commentary on the National Academies report, published in the American Journal of Geriatric Psychiatry in 2020, argues for routine assessment of isolation in clinical settings and discusses what that would actually require in practice — screening tools, staff time, referral pathways that lead somewhere real. That “somewhere real” is the quiet condition on which the whole social prescribing model depends. Assessment without a functioning, adequately resourced referral network is a diagnosis with no treatment. A link worker who can offer only a handful of overstretched voluntary programmes in a given area is not equivalent, in evidence terms, to one operating in a well-resourced pilot site where most of the favourable published studies were run.

This is the strongest reason for caution about generalising the current evidence into broad policy. Nearly all of the positive social prescribing studies come from settings with above-average programme investment and, often, evaluation built in from the start. Scaling the same referral concept into ordinary primary care, with ordinary funding and no dedicated evaluation, is a different intervention in practice even if it carries the same name.

What this does and does not settle

The evidence supports a modest conclusion: social prescribing is a reasonable thing for a health system to try, particularly where it connects patients to activities that are structured and durable rather than one-off. It does not support the stronger claim, sometimes implicit in how governments describe these programmes, that social prescribing is a proven cost-effective response to loneliness or isolation with measurable downstream effects on mortality or chronic disease. The trials that would establish that — randomised, adequately powered, tracking hard outcomes like hospital utilisation or cardiovascular events over years rather than self-reported wellbeing over months — mostly do not exist yet.

The National Academies report’s core recommendation is more cautious than the policy rhetoric that has followed it: it calls for health systems to routinely assess isolation and loneliness, not for any particular treatment pathway to be assumed effective in advance. That distinction has gotten lost as social prescribing moved from pilot programme to embedded policy in the UK and prompted framework, and it is worth restating precisely because the underlying case for taking loneliness seriously as a health exposure is genuinely strong. Overselling the fix is not necessary to make that case, and it makes the eventual reckoning with weak trial results more damaging than it needs to be.

A stronger evidence base would look like this: randomised assignment to social prescribing versus usual care, in ordinary (not enhanced) primary care settings, with pre-registered hard outcomes — hospital admissions, mortality, validated loneliness scales — followed for at least two years, and reported regardless of whether the result is positive. Almost none of the current literature meets that bar. Until more of it does, social prescribing should be described as promising and under-tested, not as a solved problem in the loneliness policy toolkit.