The Loneliness Interventions Keep Getting Tested Without a Control Group

Two systematic reviews of social prescribing, published seven weeks apart in 2021, reach the same conclusion by different routes. A review in Perspectives in Public Health found that all nine included studies reported positive individual impacts on loneliness, and that three reported reductions in GP visits, emergency attendances, social worker contact, or inpatient use. A separate review in the International Journal of Environmental Research and Public Health, looking at a wider set of well-being outcomes, reported gains in self-esteem and self-confidence as the recurring finding, and then flagged limited trial evidence and heavy heterogeneity across programmes as the reason those gains are hard to interpret.

Read together, the two reviews describe an intervention that appears to work and a literature that cannot yet say why, how much, or compared to what.

What social prescribing is, and why it is hard to evaluate

Social prescribing is the practice of a clinician referring a patient to a non-clinical activity — a walking group, an arts programme, a befriending scheme, a community gardening project — on the theory that social and structural deficits contribute to poor health and can be addressed the way a medical condition is: by referral. The United Kingdom’s 2018 loneliness strategy, the first national strategy of its kind, made social prescribing a central plank of government policy and committed funding to link workers who would make these referrals routine within primary care.

That policy commitment arrived ahead of the evidence base that would justify it at scale. Both 2021 reviews describe a field of small, non-randomized studies. Programmes vary by design, population, dose, and outcome measure, which is what “heterogeneity” means in a review’s methods section: no two studies asked quite the same question in quite the same way, so pooling their results into a single effect size is not really possible. The Perspectives in Public Health review’s finding that all nine studies reported positive impacts sounds unambiguous until one asks what a null or negative finding would have looked like, and whether such a study would have been conducted, completed, and published at all.

The National Academies named this in 2020, and named it again

The National Academies of Sciences, Engineering, and Medicine’s 2020 consensus report on isolation and loneliness in older adults is not primarily about social prescribing, but it arrives at the identical structural complaint from the clinical side. The report estimates that roughly one quarter of adults aged 65 and older are socially isolated, and it calls on the health care system to routinely assess isolation and loneliness the way it assesses blood pressure or smoking status. But a call to assess presumes something to prescribe once the assessment is positive, and the report’s own review of interventions is notably more cautious than its review of prevalence. A 2020 commentary on the report in the American Journal of Geriatric Psychiatry, written for a clinical audience, pushes on exactly this: it argues for routine assessment in clinical settings while discussing, at some length, what that would actually require — meaning it treats the assessment infrastructure as more ready for deployment than the treatment infrastructure that would need to follow it.

This is the gap. It is not that nobody has studied whether connecting isolated people to activities helps. It is that fifteen years of an increasingly confident public health argument — that loneliness is a mortality risk factor on par with smoking, that social isolation is a modifiable and preventable condition — has outpaced the trial evidence for the specific interventions being funded to fix it.

The prevention argument is stronger than the treatment argument

Julianne Holt-Lunstad’s 2021 review in the American Journal of Lifestyle Medicine makes the case explicitly: social connection belongs in preventive health frameworks alongside diet, exercise, and smoking cessation. The argument rests on genuinely strong evidence at the population level. Her earlier meta-analyses established that social isolation carries an odds ratio of 1.29 for early mortality, loneliness 1.26, and living alone 1.32, drawing on datasets covering hundreds of thousands of participants. That is a real, replicated, dose-response relationship between social deficit and death.

But a population-level risk factor is not the same claim as an intervention-level treatment effect, and the 2021 review is candid that connection is being positioned as modifiable and preventable rather than shown, trial by trial, to respond predictably to a given clinical intervention. The distance between those two claims is precisely where social prescribing sits. It is plausible — even likely, given what is known about the health correlates of isolation — that connecting an isolated person to a recurring group activity improves outcomes. Plausibility is not evidence of effect size, and the reviews available in March 2022 do not close that distance.

Instrument mismatch compounds the problem

Part of why these reviews struggle to pool results is that the underlying prevalence data they are responding to is not measured consistently either. The AARP Foundation’s 2018 survey of adults 45 and older used the 20-item UCLA Loneliness Scale and found one in three respondents lonely — a figure that is comparable to the academic literature precisely because it used a validated instrument rather than a bespoke question. Many programme evaluations reviewed in the social prescribing literature do not have that luxury; they use whatever locally constructed measure a given clinic or charity had on hand, often a single self-report item asked before and after a programme with no comparison group. A study that measures loneliness with a validated scale before and after an eight-week gardening referral, with no control arm, cannot distinguish a genuine effect from regression to the mean, seasonal variation, or the general tendency of people who volunteer for a study to report improvement regardless of the intervention.

What would actually resolve this

The reviews themselves are unusually clear about what is missing, which is worth taking seriously rather than treating as boilerplate limitations language. What the field needs is not another single-arm pre/post study of a well-being programme. It needs randomized or at minimum well-matched controlled trials, using a consistent validated instrument such as the UCLA scale across sites, with a follow-up period long enough to distinguish a durable change from a novelty effect, and with health service utilization tracked as a secondary outcome rather than inferred anecdotally. The Perspectives in Public Health review’s finding on reduced GP and emergency contact in three of nine studies is the most policy-relevant data point in either review, because it is the kind of outcome a health system actually budgets against — and it is also the outcome resting on the thinnest evidentiary base.

Until that trial infrastructure exists, government strategies that fund social prescribing at scale are making a reasonable bet informed by strong adjacent evidence on the harms of isolation, not a bet informed by strong direct evidence on the benefits of the specific programmes being funded. Those are different claims, and the difference is exactly what keeps showing up, review after review, as a footnote rather than a headline.