Matching the Intervention to the Driver, Not the Label

A GP referral that says “socially isolated, please advise” tells you almost nothing. It does not tell you whether the person has no one around them, or has people around them and still feels alone. Those are different problems with different fixes, and treating them as the same thing is the single most common way social prescribing referrals fail to change anything.

The National Academies made the distinction formal in 2020: social isolation is an objective shortage of contact, loneliness is the subjective distress of feeling disconnected, and a person can have either without the other. Someone can live with three generations of family and still be lonely. Someone can live alone by choice, see few people, and not be lonely at all. Recent work on isolation and age has gone further, showing the relationship between the two shifts depending on how old someone is — meaning a fix that works for a 40-year-old is not automatically the fix for an 80-year-old with the same referral note. If your intake process does not capture which one you’re looking at, you are prescribing blind.

Step 1: Separate the objective from the subjective before you do anything else

Ask two different kinds of question, not one blended one.

For isolation, ask about contact: how many people did you speak to this week, do you have someone you could call in an emergency, how often do you leave the house. These are countable.

For loneliness, ask about feeling: the UCLA Loneliness Scale, used in AARP’s 2018 survey of adults 45 and older, is the standard instrument and it asks about feeling left out, feeling in tune with people, feeling part of a group of friends — not about contact volume. AARP’s data show why the split matters practically: people who had spoken to a neighbour were lonely 33% of the time, against 61% of those who never had. Contact reduces loneliness on average, but the gap is not zero, and a third of people with a neighbourly relationship were still lonely. Volume of contact is not a proxy for feeling connected.

Cacioppo’s framing of loneliness as an aversive signal — closer to hunger than to a mood — is useful here because it points at mechanism: loneliness is the brain flagging a mismatch between the connection someone wants and the connection they have. That mismatch can exist at any level of actual contact. Isolation is a supply problem. Loneliness is a mismatch problem. They need different interventions.

Step 2: Route by driver, not by demographic

Presentation	Likely driver	Intervention family with the better fit
Low contact, functional capacity to socialise, wants more people around	Isolation	Group activity referral, transport support, third-place connection (library programmes, activity groups)
Adequate contact, reports feeling unseen or unheard within it	Loneliness	Structured psychological approaches — behavioural activation, purposeful group participation
Recent loss of role or routine (retirement, bereavement, job loss)	Situational, likely to resolve with re-engagement	Volunteering or purpose-based activity
Long-standing, treatment-resistant loneliness despite social contact	Cognitive/chronic loneliness	Referral toward a psychological intervention rather than a social one

The recent work on risk factors in socially isolated older adults reinforces that isolation itself is not evenly distributed or evenly caused — some isolated older adults are at much higher risk of becoming lonely than others, which argues against a single generic group-activity referral for everyone flagged as isolated.

Step 3: Know what the evidence actually supports for each route

Practitioners are told, correctly, to be sceptical of the social prescribing evidence base. Here is where it currently stands, claim by claim.

Claim	Evidence status
Structured, purposeful group activity works better than unstructured social contact alone	Reasonable — supported by qualitative synthesis (2022) describing restored purpose as the mechanism, not contact volume
Volunteering reduces loneliness in isolated older adults	Solid for this group — one of the few RCTs in the field (HEAL-HOA, 2024) found this
Befriending reduces loneliness	Real but modest — a 2025 RCT in aged care found a measurable UCLA scale reduction at 8 and 16 weeks
Behavioural activation outperforms befriending for chronic loneliness	Emerging but notable — a 2026 RCT with 1,151 older adults found telephone-delivered behavioural activation beat a befriending control at 12 months
Social prescribing reduces GP and emergency service use	Weak — only three of nine studies in a 2021 systematic review found this
Effectiveness of social prescribing for loneliness in older adults generally	Unclear — a 2025 protocol notes only one peer-reviewed RCT exists in this specific area

The pattern across the strongest evidence is consistent: interventions with structure and a defined activity beat interventions that are simply “someone to talk to.” Befriending is not worthless — it produced a real effect in a controlled trial — but where a head-to-head test exists, it lost to a more structured approach. That is a meaningful signal for anyone deciding where to put a limited number of referral slots.

What this means in practice: if your programme has one referral pathway called “loneliness support” and it defaults everyone into a befriending scheme, you are giving your highest-need, most chronically lonely referrals the intervention that performs worst against a structured alternative, and giving your isolated-but-not-lonely referrals more one-to-one contact than they actually need. Split the pathway before you split the budget.

The befriending default

Name the failure mode so staff can catch it in themselves: the befriending default is the tendency to route every loneliness referral to a one-to-one companion match because it is the easiest thing to commission and the easiest thing to explain to a funder. It feels responsive. It is also, on the current trial evidence, the intervention most likely to be outperformed by something else for the people whose loneliness is chronic rather than situational. Use it deliberately for the cases it suits — recent bereavement, someone rebuilding contact after a health crisis — not as the standing answer to every referral marked “lonely.”

Building the intake question set

A workable triage needs three things, asked in this order:

1. A contact count (isolation check) — separate from a feeling scale (loneliness check).
2. A timeframe — is this new (past six months) or long-standing (years)? Situational loneliness responds to re-engagement; chronic loneliness usually needs more than a group activity.
3. A capacity check — does the person have the physical and cognitive capacity to attend a group activity, or does the intervention need to come to them (telephone-delivered, home visit)?

Three answers, and you already know roughly which row of the routing table applies.

What this does not solve

None of this fixes the supply side. You can triage perfectly and still have nowhere structured to send someone, because the volunteering programme has a waiting list or the behavioural activation service does not exist in your area. Triage tells you what to look for; it does not create the intervention. It also assumes the person engaged with intake in the first place — this entire framework applies only to people who already made contact with a GP, link worker, or service. It says nothing about the larger number who never reach that door.