Measuring the Handoff, Not Just the Outcome

Most social prescribing evaluations measure the wrong stage. They track whether the patient who arrived at the community group felt better afterwards, which is worth knowing, but it tells you nothing about the much larger group who were referred and never arrived. If you are commissioning or running a pathway, the handoff between the clinician who identifies isolation and the link worker or activity that is supposed to address it is where most of the value leaks out. Evaluate that stage separately, or you will keep funding programmes that look effective on paper and reach almost nobody.

Why the handoff is the weak point, not the activity

The National Academies’ 2020 consensus report called for routine assessment of social isolation in health care settings, and the clinical commentary that followed argued for building that assessment into everyday practice. Both assumed the hard part was identification. In practice, identification is the easy part — a validated screening question takes thirty seconds. The hard part is what happens in the next ten minutes: does the clinician actually make the referral, does the link worker actually reach the patient, and does the patient actually show up.

Systematic reviews of social prescribing report real gains once someone actually reaches the intervention — a 2021 review linked to increases in self-esteem and confidence, and another found all nine included studies reported positive individual impacts, with three showing reduced use of GP, emergency, or inpatient services. But both reviews were built on people who completed the pathway. Neither review — nor most published evaluations — reports the size of the group who dropped out between referral and first contact, because most programmes do not measure it. A 2025 systematic review protocol notes that despite growing adoption, the effectiveness of social prescribing for older adults remains unclear, and that only one peer-reviewed randomised controlled trial exists in the field. That gap is partly a design problem: evaluations are built around the people who arrived, not the pathway that was supposed to deliver them.

The three-drop-off model

Treat the pathway as three separate transitions, each with its own denominator and its own failure rate.

1. Identification to referral. A clinician screens positive for isolation. Do they refer? Screening without a clear, low-friction referral route in the same consultation produces identification without action — a documented risk flagged in clinical commentary on the isolation literature.
2. Referral to first contact. The link worker receives the referral. Do they reach the patient, and does the patient respond? This is where warm handoffs matter — a named person calling within days outperforms a letter arriving weeks later, though the comparative evidence here is thin and mostly descriptive.
3. First contact to sustained engagement. The patient attends once. Do they attend again? A qualitative synthesis of social prescribing found that participants describe benefit extending beyond social contact to restored purpose and participation — but that finding applies to people who stayed engaged long enough to feel it. One visit rarely produces it.

Most dashboards report only step three’s endpoint — outcome scores among people still attending at follow-up — and silently drop everyone lost at steps one and two. That is not dishonest, usually. It is just measuring the part that was easy to measure.

What to actually track

Claim	Evidence status
Structured, purposeful group activity produces more benefit than contact alone	Reasonably supported — qualitative synthesis of social prescribing found benefit tied to restored participation and purpose, not contact alone
Social prescribing reduces loneliness in older adults	Weak — only one peer-reviewed RCT exists; most evidence is uncontrolled programme evaluation
Warm, named-person handoffs improve completion versus passive referral	Plausible, under-evidenced — widely assumed in practice, rarely isolated as a variable in published trials
Structured psychological interventions can outperform simple befriending	Supported by trial evidence — a 2025 aged-care RCT found befriending reduced UCLA Loneliness Scale scores at 8 and 16 weeks, but the 2026 HEAL-HOA trial found telephone-delivered behavioural activation and mindfulness beat a befriending control at 12 months among older adults living in poverty and digitally excluded

That last comparison matters for pathway design, not just intervention choice. Befriending is the default referral for most link workers because it is cheap and easy to staff. The 2024 HEAL-HOA trial tested prosocial engagement and volunteering against a control in Hong Kong and found real effects. But the follow-up 2026 trial, run with 1,151 older adults in poverty and digitally excluded, found that eight short telephone sessions of behavioural activation, delivered by trained laypeople who were themselves older and had experienced loneliness, outperformed befriending on loneliness at 12 months. If your pathway’s only “step three” option is generic befriending, you may be routing people into the weaker arm of a comparison that has already been run.

What this means in practice: before you evaluate outcomes, evaluate completion. Track the denominator at each of the three transitions — screened, referred, contacted, attended once, attended three times — for at least one full quarter before you report a single wellbeing score. If you cannot produce those five numbers, you do not yet know what your programme does; you know what happens to the people it did not lose.

The naming problem

Call this failure mode what it is: the silent referral. It is the referral that gets logged in the clinical record, counted in commissioning returns as “an intervention delivered,” and never converts into a single conversation with the patient. It is invisible in most dashboards precisely because it looks, from the clinician’s side, like the job is done. Auditing silent referral rates — comparing referrals logged against contacts made by the link worker — is one of the few pathway metrics that is cheap to collect and reliably exposes where the design is failing.

What this does not solve

None of this fixes the underlying evidence problem: the intervention literature this pathway feeds into is still thin. Most social prescribing evidence comes from uncontrolled evaluations, and the two rigorous trials available — HEAL-HOA’s two iterations — cover a narrow population of older adults, mostly outside primary care referral pathways as usually built. Tightening the handoff will get more of the right people to whatever is on offer at the end of it. It will not tell you whether what is on offer is the right thing. And pathway metrics, however carefully built, still only describe people who were screened in the first place — anyone missed at that first step never appears in any denominator at all.