social · health
the last step into the room
On the UK's largest social prescribing platform, nine in ten cases reach a link worker, but seven in ten referred for social-relationship reasons such as loneliness get no intervention — and the records cannot show who ever arrived
Source: Bu et al., “Equal, equitable or exacerbating inequalities: patterns and…” Open ↗source: primaryLast updated: 2026-11-12
Problem statement
Social prescribing — in England part of the NHS Long Term Plan's commitment to embed universal personalised care — gives people who are lonely or isolated a health-system route: a GP or other professional refers the person to a link worker, who is meant to connect them to community groups and activities. The first hand-off works. In an analysis of 201,037 cases from 169,818 people recorded on Access Elemental, which the authors describe as the largest digital platform for social prescribing across the UK (January 2017 to November 2022), about 90% of cases had at least one contact with a link worker. The second hand-off does not: only 38.3% of cases had any intervention prescribed (35.4% in England), and among cases referred for social-relationship reasons (e.g. loneliness and social isolation), 71.0% had no intervention at all. Even where an activity is prescribed, the record "concerns if any prescription was issued by the link worker and what activities it entailed, without addressing service attendance or engagement" — so nobody can tell from the data how many lonely people actually walked into a group. The problem is the step from link worker to community activity, and the fact that the system cannot see it.
Why this matters
Social relationships were the third most common referral reason (22.5% of cases) after mental health and practical support. A referral that stops at the link worker leaves the person with contact from the link worker alone — across all cases, 81.2% of recorded link-worker contacts were by phone, video, email or letter, only 6.3% face to face, and most contacts lasted 0–15 minutes (59.4%) — rather than the new relationships the pathway is meant to open. The authors say it is unclear how much of the gap is voluntary — some people may find "sessions of talking and a supportive relationship sufficient for their needs" — but argue that "several factors suggest that it is not entirely voluntary": only around 20% of people referred for mental health, practical support or social-relationship reasons had an intervention matched to that reason, against 47.9% for employment, where free provision has long been a UK priority (employment referrals also had a much lower no-intervention rate, 37.1%), which they read as suggesting that "when free interventions are readily available, intervention rate could be higher." Because 91.7% of prescribed interventions were free to the user, the pathway depends on community activities existing locally; the authors conclude that "a shortage of provision of community activities – especially ones relevant to mental health, practical support and social relationships – was evident." The WHO Commission on Social Connection (2025) lists social prescribing among the community strategies for loneliness but notes the evidence "is not yet robust enough for conclusions to be drawn."
What’s been tried and why it hasn’t worked
Link workers are now employed in primary care networks, secondary care and the voluntary sector across England, and national guidance mandates outcome reporting via the Social Prescribing Information Standard, but three things break the chain. (1) Supply of activities. In England urban residents were more likely to have an intervention prescribed than rural residents (in Wales and Scotland the reverse held), and the authors read the urban–rural unevenness as "suggesting that a lack of provision may be an issue"; the authors call for either building community infrastructure outside employment support or ringfenced funding that link workers can spend in under-resourced domains. The 2026 national link-worker survey (385 respondents in England) again found lack of provision or waiting lists "the most cited barrier across almost all types of provision." (2) The person's own steps. In a qualitative study of a Luton programme (four practices; 10 service users, 3 GPs, 2 navigators), uptake was helped by trust in the GP and by the navigator's first phone call ("Oh yes, I forgot about it and then she [the navigator] called"), but blocked by fear of stigma, expectations of a medical solution, and the short-term nature of the programme — free gym sessions limited to "a couple of weeks" lost people's interest. Two service users engaged with the navigator and with other services but one "did not engage with community group" and the other "did not engage with community services after first visit." The 2026 survey likewise lists patients choosing not to attend as a notable barrier across all categories. The WHO report adds that a person's willingness to join a prescribed group "may depend on their interest in the activity prescribed, the similarity of their interests with others in a group, relationships with link workers and group dynamics." (3) Data that cannot see the gap. A national standard (the Social Prescribing Information Standard) mandates outcome reporting, yet in the 2025 national survey of 409 caseload-holding link workers in England only 57.5% recorded patient outcomes "often" or "very often" and only 48.7% used quantitative outcome measures such as ONS4. Primary-care coding is also unreliable: across 3,086 records from 11 GP practices in Northwest England, wider determinants of health were recorded inconsistently, and in the three practices that added free text, ten patients coded as "offered" had in fact declined. Among age groups, recorded decline rates were highest for people aged 60–74 (32.3%) and 75–89 (30.8%). On the activity side, the 2026 survey finds "very limited feedback is collected from activity providers, despite their central role in delivery."
What would unlock progress
Two linked pieces are missing. First, a reliable way to close the loop between link worker and activity: a confirmation that the person arrived (and came back), returned from the activity side to the link worker's record, so that the 38% figure becomes a measured attendance rate rather than a prescription count. The software side of this has been offered: one commercial platform's brochure (Social Rx Connect) says providers can "receive referral notifications so you can manage capacity effectively and confirm attendance" and link workers can "access attendance reports", and a 2019 NHS England capability specification, since retired, required that a person's "attendance or non-attendance can be recorded", including by "telephoning the service provider". No source read here reports how often activity hosts send that confirmation, so the open question is whether and why that confirmation comes back in practice, not whether a field for it can be built. Second, a local, current map of activity supply against referral reasons, so that link workers and commissioners can see where "no intervention" means "nothing suitable exists here" rather than "the person declined." The Luton evidence that a single proactive phone call recovered people who had forgotten their referral suggests that light-touch, well-timed prompts at the hand-off may matter as much as new provision. Adjacent fields offer possible models, not tested in social prescribing by any source read here: closed-loop referral tracking in clinical referral systems, and "warm handover" protocols in mental-health services.
How we’d know it’s being solved
The measurement basis is attendance, not prescription: the share of people referred for social-relationship reasons who attend a first community activity, and the share still attending at a fixed later point (for example 8–12 weeks), recorded in a way that can be linked to the link-worker case. Progress would show as (a) a falling proportion of social-relationship referrals ending with no intervention, broken down by urban/rural and age; (b) attendance data existing at all for prescribed activities; and (c) a loneliness measure (the ONS direct question or a validated scale) collected at referral and follow-up for the same people. Beware the proxy: counting prescriptions or link-worker contacts measures the pathway's activity, not whether anyone met anyone.
Where to start
Facility-free door: pick one English town or Primary Care Network area and build an activity-supply map from public sources (council and voluntary-sector community directories, library and leisure listings, open data on community assets), coded against the seven referral domains Bu et al. use; then compare it with the published national referral-reason mix to show where social-relationship provision is thin, and prototype the map as a tool a link worker could search in a 15-minute call. Design door: prototype a closed-loop "arrival confirmation" — the smallest thing an activity host (a choir leader, a men's shed volunteer) could do to tell the link worker that a referred person came — and test its usability with volunteer group organisers, who need no NHS data access; start by finding out whether hosts use attendance features like the one a referral platform already advertises, and if not, why. Partner door (design the trial, don't run it): with a Primary Care Network or local social-prescribing service, write the protocol for a pilot comparing usual practice with an accompanied or prompted first visit, using attendance at a first activity as the primary outcome, a validated loneliness instrument, and a pre-registered analysis plan, and hand it to the partner. Access note: pathway data like the Access Elemental extract used by Bu et al. and the national link-worker survey microdata are not public; the student versions above use public directories, synthetic records, or data a partner chooses to share under its own governance. Relevant skills: service design, information systems, survey and evaluation methods, community engagement.
Genes
Related problems
Stuck the same way, in other fields.
field: the built worldcalibrated, but is it right?After a repair, shops have no settled way to prove the car's crash-avoidance sensors work as designed again — about half of owners who had these features repaired report problems afterwardbuilt without the people who'd live with itsource: secondaryfield: water & energydilute, dispersed, and too potent to ignoreHundreds of thousands of engines, flares and coal mine shafts emit dilute methane, where conventional solutions often fall shortbuilt without the people who'd live with itsource: primaryfield: food & the living worldknowledge that will not cross the fenceFarmer field schools cut participants' pesticide use by nearly a quarter — and their next-door neighbours learn nothing, because season-long experiential knowledge cannot be passed on by word of mouthbuilt without the people who'd live with itsource: secondaryfield: water & energypower for shacks the grid can't reachRegulations keep the grid out of most of South Africa's informal settlements, so solar has to work where the wires can't gobuilt without the people who'd live with itsource: primarySources
- Bu F, Hayes D, Burton A, Fancourt D. "Equal, equitable or exacerbating inequalities: patterns and predictors of social prescribing referrals in 160 128 UK patients." The British Journal of Psychiatry 226(2):91–99, 2025, doi:10.1192/bjp.2024.141, accessed 2026-10-22. Link ↗
- Pescheny J, Randhawa G, Pappas Y. "Patient uptake and adherence to social prescribing: a qualitative study." BJGP Open 2(3):bjgpopen18X101598, 2018, doi:10.3399/bjgpopen18x101598, accessed 2026-10-22. Link ↗
- Fancourt D, Bu F. "Patterns and predictors of outcome monitoring amongst link workers: Learnings from the National Social Prescribing Link Worker Survey 2025." PLoS One 21(4):e0346234, 2026, doi:10.1371/journal.pone.0346234, accessed 2026-10-22. Link ↗
- Khan K, Al-Izzi R, Montasem A, Gordon C, Brown H, Goldthorpe J. "The feasibility of identifying health inequalities in social prescribing referrals and declines using primary care patient records." NIHR Open Research 3:1, 2023, doi:10.3310/nihropenres.13325.2, accessed 2026-10-22. Link ↗
- WHO Commission on Social Connection, "From loneliness to social connection - charting a path to healthier societies: report of the WHO Commission on Social Connection," World Health Organization, 2025, ISBN 978-92-4-011236-0, accessed 2026-10-22. Link ↗
- National Academy for Social Prescribing, "Social Prescribing Link Worker Survey 2026 – Full Report," prepared for NASP by Kate Sewel, May 2026, accessed 2026-10-22. Link ↗
- Social Rx Connect, social prescribing software brochure (vendor PDF, undated), accessed 2026-11-12. Link ↗
- NHS England, "Social Prescribing v1.0.0" (capability C37, status Retired, effective date 01 Apr 2019), accessed 2026-11-12. Link ↗
Revision notes
This brief gets checked against its sources and contains the following revisions. Future revisions get added here.
Worth knowing
- the seven-in-ten figure counts referral cases, not people, and "no intervention" means none was prescribed
- the survey's top barrier, lack of provision and waiting lists, applies mostly to mental health, housing and social care services, not community groups
Full check
Sources read
- Bu et al., social prescribing pathways on a national platform (British Journal of Psychiatry, 2025): full text
- Pescheny et al., social prescribing in Luton (BJGP Open, 2018): full text
- Fancourt & Bu, survey of link workers (PLoS One, 2026): full text
- Khan et al., social prescribing records in general practice (NIHR Open Research, 2023): full text
- WHO Commission on Social Connection, report (2025): full text
- National Academy for Social Prescribing, link worker survey (2026): full text
- Social Rx Connect, software brochure: full text
- NHS England, Social Prescribing capability specification (2019): full text
What changed
- Numbers: the 71.0% with no intervention is a share of referral cases, not of people; the page now says cases; the decline rates of 32.3% and 30.8% are the highest among age groups only; other groups in the same table are higher, and the page now says so.
Left open
None noted in this check.
Separate check of specific claims: A software brochure and an NHS England specification that already provide for recording attendance were added, and the page no longer says such features go unused.
Separate check of specific claims: A wording slip was fixed, and a claim that other fields have solved this problem is now offered as a possible model.