Social prescribing: less rhetoric and more reality. A systematic review of the evidence

1Social prescribing is a way of linking patients in primary care with sources of support within the community. It is being widely promoted and adopted as means of dealing with some of the pressures on general practice.

2This systematic review assesses the effectiveness of social prescribing programmes relevant to the UK National Health Service setting. We have searched for full publications and grey literature since 2000 and identified 15 evaluations. It is possible that some local evaluations have not been identified, but it is unlikely that any unidentified evaluations would do little to alter the overall picture of a low-quality evidence base with a high risk of bias.

3If social prescribing is to realise its potential, future evaluations must be comparative by design and consider when, for whom, how well and at what cost.

4With estimates of a £30 billion funding gap by 2020, a radical rethink of the way health services are currently delivered remains high on the policy agenda. The Five Year Forward View has stressed that developing innovative approaches to delivering healthcare are integral to the long-term future of the National Health Service (NHS).

5Social prescribing is one such model and is being widely promoted as a way of making general practice (GP) more sustainable. Social prescribing is a way of linking patients in primary care with sources of support within the community. It provides GPs with a non-medical referral option that can operate alongside existing treatments to improve health and well-being. There is no widely agreed definition of social prescribing but the Social Prescribing Network defines it as ‘enabling healthcare professionals to refer patients to a link worker, to co-design a nonclinical social prescription to improve their health and wellbeing.’ Schemes commonly use services provided by the voluntary and community sector and can include an extensive range of practical information and advice, community activity, physical activities, befriending and enabling services. The types of activities offered as part of a social prescribing service can aim to help address the psychological problems and low levels of well-being often manifest in frequent attenders in GP. By addressing these, it is often hoped that there will be a subsequent positive impact on frequency of attendance.

6As early as 1999, the white paper Saving Lives: Our Healthier Nation was advocating that the NHS should make better use of community support structures and voluntary organisations. However, it was in 2006 that the Department of Health advocated the introduction of social prescriptions for those with long-term conditions, and NHS England has since announced the appointment of a national clinical champion for social prescribing. With the current Secretary of State for Health also promoting access to non-clinical interventions that take a more ‘holistic view’, support for social prescribing is significant at the policy level.

7Many localities are now offering or considering implementing social prescribing programmes, but is the apparent enthusiasm justified? As part of a study which aimed to help NHS commissioners make better use of research in their decision-making, we examined the evidence for social prescribing. This systematic review summarises the evidence for the effectiveness of social prescribing programmes relevant to the UK NHS setting.

8The protocol and amendments were registered in PROSPERO (registration number: CRD42015023501).

9Database of Abstracts of Reviews of Effects (DARE), Cochrane Database of Systematic Reviews and NHS Economic Evaluation Database (NHS EED) were searched for relevant systematic reviews and economic evaluations (24 June 2015; no new records added to DARE and NHS EED databases from January 2015 so we did not run updated searches).

10We searched the following databases (initial search 26 June 2015; updated search 5 February 2016): Applied Social Sciences Index and Abstracts (ASSIA), Cumulative Index to Nursing and Allied Health Literature (CINAHL), MEDLINE, Social Care Online and Social Policy and Practice.

11As our focus was on identifying evidence relevant to the UK NHS setting, we also searched for eligible studies in key UK knowledge repositories for health and social care. The websites of National Institute for Health and Care Excellence (NICE), Social Care Institute for Excellence (SCIE) and NHS Evidence were searched for reviews, guidance, evidence briefings or any other papers describing or evaluating social prescribing programmes. Additional searches of the websites of key policy think tanks the Kings Fund, Health Foundation, Nuffield Trust and NESTA were also undertaken. We searched Google to identify grey literature reports of relevant evaluations in UK settings (5 January 2016). Reference lists of retrieved articles were scanned to identify additional studies.

12All the searches were restricted to English language only and published between 2000 to January 2016. The search strategies are available in online supplementary appendix 1.

13Systematic reviews and any published evaluation of programmes where healthcare professionals refer patients from a primary care setting to a link worker or facilitator for any form of social prescription were eligible for inclusion. Studies were eligible regardless of whether a comparison group was included.

14As per the Social Prescribing Network definition, we included only studies where referral was made from a primary care setting to a coordinator, link worker or facilitator of social prescribing (this type of role will be referred to as ‘link workers’ throughout this review). Any activities or interventions being specifically delivered as part of a social prescribing programme were included in the review.

15We excluded studies where referral was made from outside of a primary care setting and or where primary care health professionals refer patients to services delivered as part of mental health or counselling services such as an Improving Access to Psychological Therapies programme. We also excluded evaluations of activities that could be socially prescribed (eg, physical activity programmes or community arts projects) but did not involve referral to a link worker in the first instance.

16The primary outcomes of interest were any measures of health and well-being, including self-reported measures (eg, levels of physical activity or depression scores) and/or measure of usage of health services. We also considered any other outcomes (eg, health service usage) reported in the included evaluations.

17Study selection was performed by one researcher and checked by a second, with any discrepancies resolved by discussion or with recourse to a third researcher.

18Details of the setting, participants, the intervention (type, delivery mode and length of time), type of evaluation and outcomes of evaluation were extracted and quality assessed by one researcher and checked by a second. Discrepancies were resolved by discussion or by recourse to a third researcher.

19We used the Cochrane risk of bias tool to assess the quality of the randomised controlled trial (RCT). To assess the quality of the before and after evaluations, we applied the quality assessment tool developed by the US National Heart, Lung, and Blood Institute for before–after (pre–post) studies with no control group. Our primary focus was on effects. As per our protocol, we have not made a formal quality assessment of studies of a qualitative or descriptive nature.

20We performed a narrative synthesis of the evidence. There were insufficient data to perform meta-analysis for any of the outcomes of interest. No subgroup analyses were planned. The narrative synthesis was intended to move beyond a preliminary summary of study findings and quality to investigate similarities and differences between studies as well as exploring any patterns in the data.

21We identified a total of 431 records through database searching and a further 14 records through other sources. After deduplication, 341 titles and abstracts were screened and 70 full-text papers were assessed for inclusion (see figure 1: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram).

22We excluded 45 studies on eligibility grounds and were unable to access the full text for seven identified records. We also identified two non-systematic reviews of social prescribing schemes. These were excluded as they did not critically appraise included studies and were limited in their synthesis of findings; one review included a number of evaluations that did not meet our inclusion criteria. We checked the reference lists of both reviews to ensure we had identified and included all relevant evaluations.

23We included a total of 15 evaluations (reported in 16 papers) of social prescribing programmes where some form of link worker role was used. The designs included one RCT, one non-RCT, two qualitative studies, four uncontrolled before and after studies and eight descriptive reports of six evaluations, of which, five included some analysis of qualitative data. Details of the included evaluations are presented in table 1.

24In each of the included studies, the link worker (job title variously named) met with the patient to discuss their needs and directed them to appropriate community/voluntary sector sources of support in their locality. The training and knowledge of people fulfilling these types of link worker role varied between projects. In some services this was a paid role, in others these roles were fulfilled by volunteers. Some link workers had good knowledge and existing networks with local services in place and in others they received some basic training and made use of a directory of resources.

25Patients were referred to a range of activities provided by local or national voluntary and community sector organisations. Interventions received included exercise and other physical activities, signposting to housing, welfare and debt advice, adult education and literacy, befriending, counselling, self-help support groups, luncheon clubs and art activities.

26The number of referrals made to social prescribing programmes ranged from 30–1607. Referrals were made by a range of health professionals but primarily GPs. Three of the studies reported that feedback was given to the referrer about the actions taken and the participants' progress in the social prescribing programme.

27Quality assessment and risk of bias for the evaluative designs is presented in table 2. In the RCT, only sequence generation was adjudged to be of low risk of bias; all other criteria were rated as unclear or high risk. The authors reported that the randomisation process was misunderstood in two of the participating practices but random allocation appeared to be maintained. A key inclusion criterion for the Cochrane Effective Practice and Organisation of Care Review Group is that a controlled before and after study must have at least two interventions and two control groups to guard against confounding. Here, the controlled before and after study includes one intervention and one control group, drawn from the same GP. As such, we rated the study as having a high risk of bias and made no further assessment of quality with the Cochrane risk of bias tool. Uncontrolled before and after studies are inherently weak evaluative designs and no included study fulfilled all of the specified quality criteria. In general, evaluations had small sample sizes (<100 participants), significant loss to follow-up (>20%), were lacking in completeness of outcome data and had unclear selection criteria for the study population. Follow-up periods were generally short (immediately post intervention, up to 4 months post intervention). There is therefore a high risk of bias.

28Seven included studies reported the number of people attending an initial appointment with a link worker. Where reported, attendance at this initial appointment with a link worker ranged from 50% to 79%. Participants' attendance at activities to which they were subsequently referred or recommended by a link worker was reported in only two studies and varied from 58% to 100%.

29The RCT, two uncontrolled before and after studies and three descriptive reports measured health and well-being outcomes at baseline and again at up to 6 months after referral to a social prescribing programme; one study reported outcomes at up to 12 months. The measures used were Warwick-Edinburgh Mental Well-being Scale (WEMWBS; ), Hospital Anxiety and Depression Scale (HADS;), General Anxiety Disorder-7 (GAD-7;), Patient Health Questionnaire-9 (PHQ-9;), Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM;), Work and Social Adjustment Scale (WSAS; ), General Health Questionnaire (GHQ-12;) and COOP/WONCA. Table 3 presents findings for studies using validated measures; all report some improvements in health and well-being. However, it is difficult to quantify the size of the observed improvements due to a lack of reported detail, a lack of sufficient control group data and differences in reporting between studies. It is not possible to determine whether any observed improvements were clinically significant. Studies reported short-term outcomes only; there is no evidence about the effect social prescribing has on health and well-being outcomes beyond 6 months.

30One uncontrolled before and after study used a bespoke measure, the Wellspring Well-being Questionnaire, comprising PHQ-9 and GAD-7 tools, and items from Office of National Statistics (ONS's) Well-being Index/Integrated Household Survey and International Physical Activity Questionnaires. A second also used a bespoke measure which used a 5-point scale across eight domains associated with different aspects of self-management such as ‘looking after yourself’ and ‘managing symptoms’. Two further descriptive reports also indicated they used the WEMWBS to measure changes in health and well-being but are poor reported and involve what appear to be very small numbers of respondents. In the two studies using non-validated measures, some positive improvements in outcomes such as depression and anxiety at 3–4 months' follow-up were reported.

31Both comparative evaluations and three uncontrolled before and after studies reported some measure of healthcare usage. This included comparing hospital episode statistics and/or GP record data from 6 to 12 months before intervention with data up to 18 months post intervention. Reported outcomes included frequency of GP consultations, referrals to secondary care, inpatient admissions and accident and emergency (A&E) attendances. Findings were mixed. The RCT reported that the number of primary care contacts were similar between intervention and control groups and that there were fewer referrals to secondary care and more prescription drugs for those in the intervention group compared with the control group. The non-randomised trial reported statistically non-significant reductions in primary care contacts (face-to-face and/or telephone) and referrals to secondary care. The before and after studies reported reductions in secondary care referrals, inpatient admissions and A&E attendances in primary care contact and in face-to-face GP contact but an increase in telephone contact.

32Three before and after studies and five descriptive reports reported on patient experience. Studies used semistructured interviews or survey questionnaires specifically designed for the project evaluation to assess participant experience.

33In six of the studies, participants reported overall satisfaction with social prescribing programmes. General improvements in feelings of loneliness and social isolation, and improved mental and physical health were also observed. Issues that may impact the willingness of patients to participate in socially prescribed activities included confidence, interest in/appropriateness of activities on offer and literacy or travel issues. One qualitative study reported that patients had poor knowledge of the service prior to attending their appointment with the link worker resulting in some feeling that the service did not meet their expectations. Another evaluation identified a similar issue regarding a lack of understanding of the service among participants.

34A small number of studies conducted semistructured interviews with primary care practitioners referring participants to social prescribing programmes and/or link workers. GPs in general found that being able to make a social prescription was a useful additional tool. Key issues identified for successful implementation of social prescribing programmes were central coordination of referrals, resources and training to support coordinators and enabling networking with the voluntary and community sector, and good communication between GPs, participants and link workers: social prescribing is unfamiliar to many GPs and requires good clear explanation to engage participants; delivering feedback on participants' progress encourages GP support for social prescribing.

35The two comparative evaluations reported costs. One found total mean costs were greater in the intervention group (£153) compared with the control group (£133). The other reported no statistically significant differences between the financial and environmental costs of healthcare use between the intervention and control groups.

36One before and after study undertook a cost-benefit analysis using estimated input costs and benefits derived from 12-month outcome data obtained for 108 patients referred to social prescribing (42 of whom were referred to funded voluntary and community service providers). A total NHS cost reduction of £552 189 was generated by multiplying the estimated per-patient cost reduction by the total number of referrals (n=1118) to funded voluntary and community service providers over the 2-year course of the pilot programme. This estimate was compared with total estimated input costs of £1.1 million.

37One other report of an evaluation estimated total running costs of £83 144 for the programme for 1 year.

38This systematic review has examined the evidence to inform the commissioning of social prescribing schemes. Overall, we identified 15 evaluations conducted in UK settings but have found little convincing evidence for either effectiveness or value for money.

39Most of the evaluations of social prescribing activity are small scale and limited by poor design and reporting. Missing information has made it difficult to assess who received what, for what duration, with what effect and at what cost. Common design weaknesses include a lack of comparators (increasing the risk of bias), loss to follow-up, short follow-up durations and a lack of standardised and validated measuring tools. There is also a distinct failure to either consider and/or adjust for potential confounding factors, undermining the ability to attribute any reported positive outcomes to the intervention (or indeed interventions) received. This is particularly important as most referred patients appear to have been receiving other interventions and so we have no way of assessing the relative contributions of the interventions to the outcomes reported. Despite these methodological shortcomings, most evaluations have presented positive conclusions, generating a momentum for social prescribing that does not appear to be warranted.

40Our systematic review appears to be the first to assess the effectiveness of social prescribing programmes relevant to the UK NHS setting. We have searched for full publications and grey literature since 2000 but it is possible that we have not identified some local evaluations. Publication bias occurs when the results of published studies are systematically different from results of unpublished studies. However, we think it unlikely that any unidentified evaluations will be more robust than those included in the review.

41Many of the evaluations presenting positive conclusions were written as descriptive reports with limited or no supporting data presented. As such, they did not adhere to formal reporting standards that would be expected in reports to funding agencies or in academic journal articles. This made extracting any relevant data difficult and it is possible information relevant to outcomes is missed. Even if this shortcoming of data completeness were to be addressed, we believe that it would do little to alter the overall picture of a low-quality evidence base with a high risk of bias.

42Our systematic review has not established that there is clear evidence that social prescribing is ineffective. Rather, we are not yet able to reliably judge which, if any, social prescribing programmes demonstrate a degree of promise and so could be considered further. The use of a link worker is the key feature of social prescribing. How this link worker role was fulfilled varied significantly between projects. So here again, we are not able to reliably judge the type of skills set or level of training and knowledge people require to effectively fulfil this role. For those seeking to commission new or extend existing schemes, this evidence gap is a hindrance rather than a help, especially so given the widespread support and advocacy for social prescribing at the policy level.

43While the tension between rigour and ‘good enough’ evidence has long been recognised, even ‘good enough’ is severely lacking from the social prescribing literature be that in the design or in the conduct of the evaluations themselves. This may in part reflect the way schemes have ‘emerged’ rather than being systematically planned with evaluation built in from the outset. Nevertheless, if social prescribing is to realise its potential, then there is an urgent need to improve the ways by which schemes are evaluated.

44Prospective pathways for undertaking rigorous planned experimental evaluation are well defined, but the opportunity, time and resources needed to employ these in a service context can be limited. However, this does not serve as an excuse for inaction and in the current financial climate we should of course only be investing in those services where we can demonstrate real benefit over existing ways of working. What this should mean for future evaluation of social prescribing is that a more coordinated approach to the planning, implementation and evaluation of new and existing schemes is undertaken. This could and should involve the adoption of a common analytical framework which in turn will facilitate standardised metrics, cross-site comparison and shared learning. The IDEAL framework offers one such pathway to navigate the evaluation continuum that would allow for the iterative development and evaluation of whether social prescribing is likely to succeed in a particular setting and allow for adaptation, refinement and system integration without losing sight of the need for more rigorous testing before wider spread. Whatever analytical framework is adopted, Lamont and colleagues have proposed five essential questions for evaluation which those planning to undertake evaluations of social prescribing programmes would do well to heed. These are:

45▸ Why—clarify aims and establish what we already know from evidence.

46▸ Who—identify and engage stakeholders and likely users of research at outset.

47▸ How—think about study design, using an appropriate mix of methods, and adjust for bias where possible (or at least acknowledge).

48▸ What—consider what to measure (activity, costs and outcomes) and combine data from different sources.

49▸ When—pay attention to timing of results to maximise impact.

50Alongside these, we would also emphasise that rigorous conduct and transparent reporting (regardless of ‘success’ or ‘failure’) are essential. Reporting guidelines such as Standards for QUality Improvement Reporting Excellence (SQUIRE) with its focus on explaining ‘Why did you start?’, ‘What did you do?’, ‘What did you find?’ and ‘What does it mean?’ could readily be applied to ensure that learning is systematically captured in a generalisable format. This in turn would serve to ensure that any future decisions relating to the continuation or wider spread of social prescribing schemes are transparent and evidence informed.

51Social prescribing is being widely advocated and implemented but current evidence fails to provide sufficient detail to judge either success or value for money. If social prescribing is to realise its potential, future evaluations must be comparative by design and consider when, by whom, for whom, how well and at what cost.