Running a Loneliness Group by Phone, Video, or in Person: What the Trial Evidence Actually Supports

A social prescribing team choosing between a phone-based programme and an in-person group is usually making that choice on grounds of cost, staffing, and reach into rural or mobility-limited patients — not evidence. That is defensible, because until recently there was almost nothing to weigh it against. The 2022 American Heart Association scientific statement on social isolation named the absence of intervention evidence as the central gap in the field, not a footnote to it. Most of what commissioners had to go on was uncontrolled programme evaluation: people who joined a group reported feeling better, with no comparison group to say what would have happened anyway.

That gap has started to close, and the result is genuinely surprising. The best-controlled trial available found that a fully telephone-delivered intervention, run by laypeople who were themselves older adults with lived experience of loneliness, outperformed in-person befriending — the modality almost every health system defaults to.

The trial that should change the default

The 2026 HEAL-HOA randomised clinical trial enrolled 1,151 older adults who were living in poverty, living alone, and digitally excluded — a population that, on paper, looks like the hardest possible case for a remote intervention. Participants received eight 30-minute telephone sessions of behavioural activation and mindfulness over one month, delivered by trained older-adult peers rather than clinicians. The comparison arm was befriending: the standard-of-care approach most social prescribing services already run, in person or by phone, built around regular friendly contact.

The behavioural activation arm produced significantly greater reductions in loneliness at 12 months. Befriending is not nothing — a separate 2025 randomised trial of befriending in residential aged care found it reduced UCLA Loneliness Scale scores by 2.39 points at 8 weeks and 2.71 points at 16 weeks against a no-treatment control. Befriending works. It was simply beaten, head-to-head, by something more structured and delivered over the phone.

The lesson is not “digital beats in-person.” It is that structure beat contact, and the structured version happened to be delivered remotely. An earlier HEAL-HOA trial testing prosocial engagement and volunteering against a control, also among lonely older adults, is one of the only other randomised tests of a loneliness intervention in the literature — which says something about how thin this evidence base still is. Two or three well-controlled trials is not a settled science. It is a start.

What the two modalities are actually good for

Claim	Evidence status
Structured, skills-based remote delivery (behavioural activation, mindfulness) can reduce loneliness more than unstructured befriending	Supported by one randomised trial (HEAL-HOA 2026), replication needed
In-person befriending reduces loneliness relative to no intervention	Supported by randomised trial evidence (2025 aged care RCT)
Video-based group meetings replicate the benefit of in-person groups	Not established either way — no randomised comparison exists
Remote work and unstructured screen-based contact are associated with more isolation, not less	Supported by observational and experimental data (Science, 2026; Gallup, 2024)
Social prescribing overall improves self-esteem, confidence, and service use	Supported by systematic review, but trial evidence is sparse and heterogeneous (2021, 2022 reviews)
Telephone delivery is a viable default for reaching digitally excluded populations	Supported — the HEAL-HOA sample was deliberately digitally excluded and phone delivery still worked

The pattern across this table is consistent: the modality that wins in the strongest trial is not “digital” in the sense most services mean when they talk about running groups on Zoom. It is low-bandwidth, structured, peer-delivered, and scheduled — a phone call with content, not a video call standing in for a room.

That distinction matters because the same year’s evidence on unstructured remote contact points the other way entirely. A 2026 study in Science found that remote work increases time spent alone and worsens mental well-being, explaining roughly a third of the rise in isolation and mental distress observed between 2011–2019 and 2022–2024. Remote workers living alone saw their time spent alone increase roughly tenfold compared with workers living with others. Gallup’s 2024 workplace data shows the same gradient: fully remote employees report loneliness at 25%, hybrid at 21%, on-site at 16%. None of this is about loneliness groups specifically, but it is a warning against assuming that moving a programme onto a screen carries the same social value as moving it into a room, just more conveniently.

The failure mode: mistaking attendance for structure

Call this the video call substitution problem. A service replaces an in-person peer support group with a video call, on the assumption that seeing faces on a screen is functionally the same activity, just cheaper to run and easier to attend. Attendance may even hold up — video removes transport and mobility barriers that keep people from in-person groups. But nothing in the current trial evidence supports the idea that unstructured video contact carries the same effect as either in-person befriending or telephone-delivered behavioural activation. The two things that produced measurable reductions in loneliness — a structured programme with defined content and sessions, or sustained one-to-one human contact — both had a deliberate shape. A video call that reproduces a coffee-shop chat format, without that shape, has not been tested and should not be assumed to work.

What this means in practice: if you are choosing between in-person and remote for a loneliness-focused group, the question that predicts outcome is not “screen or room” but “does this have content, a trained facilitator, and a defined number of sessions, or is it open-ended contact.” If it’s the latter, in-person befriending has trial support and unstructured video does not. If you can build the former, telephone delivery by trained peers has now cleared a randomised trial in a hard population, and it is considerably cheaper to scale than transporting people to a room.

For a service with a fixed budget, this reframes the modality decision. Phone-delivered behavioural activation, run by trained laypeople rather than clinicians, is a staffing model that a social prescribing hub could plausibly build without new clinical capacity. That is a genuinely different resourcing conversation than “should we do Zoom or in-person,” and it is the one the 2026 trial evidence actually supports.

Where the evidence runs out

A single trial in Hong Kong, in a population defined by poverty and digital exclusion, does not tell a service in a different country with a different population that the same eight-session phone protocol will replicate. The 2025 systematic review protocol on social prescribing for older adults notes that only one peer-reviewed randomised controlled trial existed in that area at the time of writing — the field is still counting its RCTs in single digits. A 2022 qualitative synthesis of social prescribing found that participants describe benefit extending beyond contact itself to restored purpose and participation, which is a real finding but not one that distinguishes cleanly between delivery modes.

Remote workplace evidence on isolation is not the same evidence as remote group-delivery evidence, and this piece has drawn a connection between them that the source material does not test directly — it establishes that unstructured remote work correlates with more isolation, not that unstructured remote groups fail. Healthcare workers specifically show that perceived social support moderates how remote work affects well-being, which suggests the modality matters less than what surrounds it.

What this does not solve

None of this addresses reach. Every trial discussed here recruited people who were already identified, referred, or enrolled — the population that shows up in a randomised trial is, by construction, the population a system already found. A service deciding between phone and in-person delivery is optimising for people already inside its net. The much larger group who never get referred, never answer a recruitment call, or fall outside whatever criteria define “digitally excluded but reachable” does not appear in this evidence at all, and no amount of modality tuning changes that.