The overall approach is based on that of a scoping review or study, a framework for which was provided by Arksey and O’Malley [20]. The following elements of the framework, in particular, informed the approach: identifying all literature regardless of study design, redefining search terms as familiarity with the literature increases, and using an iterative process. However, our methods were also guided by our aim to explore information available on websites about real-world projects or services and compare this with the published literature.

Because of the broad scope and varied terminology associated with social prescribing, and the intention to explore the gray literature in additional to conventional databases, we used a range of search strategies and an iterative approach. We searched the Cochrane Library, MEDLINE, PubMed, and Google scholar, and various sources of gray literature (Google, Yahoo, Bing, greylit.org, Opengrey.eu, and specific NHS websites such as NHS Evidence).

The initial search for published studies combined the following thematic fields to identify social prescribing projects: (1) social prescri* OR refer*; (2) program OR treatment OR management OR education OR support OR physical exercise OR aerobic OR physical activity OR leisure-time OR exercise OR sport OR leisure activit* OR physical fitness OR training OR physical performance OR weight loss OR weight reduction OR BMI OR body weight OR body mass index OR obesity OR overweight OR adiposity OR smoking OR tobacco OR cigarette OR social support OR loneliness; (3) diabetes; (4) primary care OR GP OR general practitioner OR community OR voluntary. In addition, we performed a search on lifestyle intervention trials for type 2 diabetes in MEDLINE and searched the 200 most recent hits for studies on social prescribing to test the effectiveness of our search strategy.

The search for gray literature (websites and related Web-based material) was based on the combination and iterative modification of the terms “social prescribing,” “social prescription,” ”primary care,” ”diabetes,” ”referral,” and “community services.” For instance, we searched the first 100 hits in the search engines for “social prescribing” and “diabetes.” If we found NHS websites with links or references to other projects of potential relevance, we followed these; otherwise we modified the search terms and repeated the process multiple times. An initial set of searches was conducted in December 2014.

The aforementioned searches were repeated in October 2016. At this point, further searches were conducted. Three search strategies for databases were developed based on the results of previous searches including further terms for link workers and “social prescribing interventions” and a term for prediabetes based on the rationale that the interventions would be of a similar nature to those for type 2 diabetes. These strategies are presented in Multimedia Appendix 1. The final search was a further Web-based search via the Google.co.uk search engine using the strategy “social prescribing” AND diabetes AND evaluation. The inclusion of evaluation as a term in this search was due to the fact that in the interim since the previous search, the use of the term social prescribing on websites had become much more extensive. Therefore, it was necessary to focus on identifying projects that had been evaluated. Searches for each of the individual projects located on websites were also carried out by searching using the project name on PubMed.

Various definitions of social prescribing have been proposed, but it is clear that there is still a lack of consensus on its scope and interpretation. Therefore, a broad, pragmatic set of criteria was used to identify studies or projects that were likely to fit within the spectrum of relevant interventions: located in the United Kingdom or Ireland and involving a primary care provider referring patients including those with type 2 diabetes or prediabetes to a third party that is delivering nonmedical services in the community. Studies or evaluations had to include a description of the referral process or be described as a social prescribing service and mention type 2 diabetes or prediabetes either in the inclusion criteria or in the report or publication. In line with the aim of a scoping review, all study designs used in evaluation were included as were all forms of report. There was no formal language restriction, but studies and projects were excluded if they were available only as abstracts, protocols, located outside the United Kingdom or Ireland, or did not provide details regarding the process of referring the patients from primary care to the external health care, community, or voluntary service. The searches were conducted by 2 authors (ML and KP). Initial screening of each set of search results was carried out by 1 author, preliminary selections by each of the 2 authors were then compared to compile a “short-list,” full-texts were consulted, and final selections were made with reference to the third author (MP) where necessary.

For each service, data were extracted by 1 of the authors (ML or KP) into a Microsoft Excel spreadsheet, documenting the name of the service, the name of the organization that carried out the service, the type of organization, the location, the criteria for referral, the type of intervention, outcomes, or objectives, the evaluation of the projects and its findings, the type of publications, and the terminology used in the project regarding social prescription. All extracted data was checked by at least one other author (KP, ML, or MP). For those projects that had undergone an evaluation where the relevant published papers or Web-based report was located, details of the evaluation were extracted. As a range of study designs were used in evaluation and because the main aim was to scope the information available, full appraisal was not carried out, as is the usual approach at this stage [21]. To provide an indication of the quality of the evidence on effectiveness, the following items were extracted: study design (including whether randomized and, where relevant, the control interventions employed), sample size, data collection methods, and outcomes measured. For this purpose, the main study or latest evaluation report was accessed. Details of any additional or related studies are discussed in the text.

A total of 40 projects or services and 24 evaluations were identified through the various searches and sources. A total of 3249 records from databases were screened, and a “shortlist” of 302 potentially relevant studies was compiled. After further screening and exclusion of studies, 29 papers related to 12 projects were selected, which met the inclusion criteria. For the Web-based searches, in addition to iterative searching, 189 documents retrieved by the final Google search were accessed, resulting in a list of 40 possibly relevant projects, of which 11 were excluded (2, not diabetes; 3, only type 1 diabetes; 3, no clear referral from primary care; 3, complex change interventions including but not focused on social prescribing). Twenty-nine projects met the inclusion criteria, with evaluations found for 13 projects (a Web-based report and a published paper were found for 1 project). Three additional project evaluations were located that included people with long-term conditions but, as diabetes was not specifically mentioned, these were not included. The screening process is shown in Figure 1.

Summaries of the evaluated projects or services focusing on the key study or report in each case are shown in Table 1. Details of the additional projects located on the Internet are presented in Table 2. The references to papers, reports, and websites are included for each project or service.

The included projects represented a range of initiatives and interventions and were commissioned by a range of organizations, either solely or in collaboration, including NHS organizations, local government, and charities. Three projects had a national coverage, providing a standardized program (eg, DESMOND) [47] or a service that was locally adopted from an initial version (eg, X-PERT) [79] or a “social prescribing” intervention provided as part of a national program [24]. The remaining projects were limited to a local area, ranging from city districts to counties or regions.

Although all published studies were located using a standard medical database, information on the projects or services was found from a variety of sources. Much of the information was retrieved from websites of the organizations delivering the projects. However, in order to include information on the prespecified aspects and, where evaluation reports were not available, it was necessary to follow links to other resources or websites and to access presentations, minutes of relevant meetings, and reports. The specific sources used are referenced in Table 2.

The studies published in peer-reviewed journals met our broad definition of social prescribing as a referral process from primary care to an external provider of nonmedical services in the community. These studies were not, however, described as social prescribing projects. In fact, only 1 study was located on PubMed that referred to social prescribing [107]. It was potentially relevant, as it was focused on long-term conditions. However, diabetes was not mentioned, and therefore it was not included. The descriptions varied considerably across the projects. Some were referred to as educational programs emphasizing the type of intervention, as “diabetes programs” emphasizing the condition, or GP referral programs focusing on the process. Projects or services were also described as a “community prescription,” a community service, a voluntary, community, and social enterprise undertaking, or a social prescribing project.

The intervention in the majority of projects (n=16) was a structured group education program that included multiple thematic areas such as disease information, disease management, healthy lifestyle, and health consequences of diabetes. For instance, the DESMOND program is a structured education program over 6 hours on 1 or 2 days [47]. In the X-PERT diabetes program, participants receive a handbook and are taught by trained educators about diabetes, weight management and other lifestyle factors, and diabetic complications [77]. A number of services involved referral or recommendation of exercise; 7 schemes were based solely on exercise (“Exercise on Prescription” schemes), whereas others offered access to exercise classes as part of a “menu” of community activities. For example, 1 service offered free physical activity options (eg, Tai Chi, Zumba, Community Gym) for 12 weeks at maximum or an individually tailored exercise program for 12 weeks in combination with personal support, a final meeting, follow-up contacts, and a report that was sent to the GP [101].

Of the interventions that did not involve group education or exercise on prescription, 1 involved trained and quality-assured lifestyle facilitators delivering an “individually tailored behavior change intervention” [22]. Another provided individual support to black and minority ethnic people with type 2 diabetes via “Public Health Improvement Teams” [84]. Other interventions included community-based diabetes prevention programs [48,50,60], a lifestyle and weight management program [76], and a social and horticultural project [64]. The remainder were described as “social prescribing.” In one of these, a personalized signposting service was provided by “Social Prescribing Coordinators” to local services provided by a voluntary, community, and social enterprise (VSCE) group [30]. Similar services were provided by Community Health Champions, Community Prescription Navigators, Social Prescribers or Social Prescribing Workers (or VCS Advisors), link workers, and a practice-based health facilitator (or health manager).

We identified a variety of types of research and evaluation on the interventions, ranging from local projects with regular audits using questionnaires, to national programs tested using multicenter randomized controlled trials (RCTs). Of the 24 evaluations, 10 involved RCTs of behavioral, education, or exercise interventions. The majority were service evaluations that used a combination of approaches, both qualitative and quantitative. This required collection and analysis of monitoring data, measurement of outcomes using validated measures or self-reporting, cost-effectiveness calculations, and methods for gaining an insight into the perspectives of both providers and users of the services through surveys, interviews, and focus groups.

Of the education programs, the DESMOND program, one of the most extensively tested, had undergone a series of clinical trials and other evaluations. These include a large-scale RCT with a subsequent 3-year follow-up study [36,44], a more recent nonrandomized trial comparing health care professional and lay educators [45], and a study on cost-effectiveness [42]. The results revealed that its group education program for patients with newly diagnosed type 2 diabetes was effective at improving some key outcomes but not all. There were positive improvements in illness beliefs, which were still sustained after 3 years, but differences in biomedical or lifestyle outcomes were not apparent after this period [44]. Further research revealed that there was no difference in study outcomes between patients who were referred to 2 health professional educators and those who were referred to a team consisting of a professional educator and a layperson [45]. The cost-effectiveness analysis showed DESMOND to be cost-effective compared with usual care at 12 months [42]. Further studies focused on the facilitators and their skills, and the perspectives of participants in the program, through the use of qualitative approaches [39,46].

A second education-based intervention, X-PERT, has also been tested in an RCT, and a range of benefits were reported on physiological and behavioral outcomes, including improved glycemic control, BMI, diet, and self-management skills at 14 months [77]. A noncontrolled trial of an adaptation of the program designed for Bangladeshi adults with type 2 diabetes, which did not involve GP referral and therefore is not included in the tables, has also been completed but was underpowered due to low attendance rates [108,109]. A program designed specifically for Muslim diabetics was also assessed by retrospective analysis and reported beneficial effects [65].

Pre-post questionnaires and physiological measures were used to assess the CODE (Community Oriented Diabetes Education) program with positive trends but not significant changes in physiological measures. However, there was a significant increase in participants’ knowledge scores, coping ability, motivation to change, and making informed decisions about their diabetes [33].

Other studies investigated the effectiveness of exercise on prescription schemes [26,54,71,75]. Mixed results were reported with effects, if observed, either limited to reduction of anxiety or not maintained over time. One RCT found no difference between an exercise program, instructor-led walking, or advice only [54], while a regression analysis of one scheme explored how the effects of support differed as a function of who provided the support [27]. RCTs were also used to assess the effects of interventions described as behavioral or lifestyle change interventions [22,51,60].

Internal or external evaluation of services described as social prescribing, either ongoing or completed, was reported on a number of the websites. Of these, evaluation reports were found for 6 “social prescribing” services. In other cases, the results of the evaluations were reported in brief on the website, or in the minutes of relevant meetings or other documents. Some services had been evaluated as part of large projects, whereas in several cases, reports were unavailable or could not be located. One further service had been approved by QISMET (the Quality Institute for Self-Management Education and Training), an independent body developed to support self-management providers and commissioners [98]. In some cases, the collection of data was reported but the results were not located by the searches we performed.

The type of evaluation ranged in design. Basic numerical data were collected including numbers referred to and accessing the service, and numbers of people reporting changes in various aspects such as increased physical activity or simply completing the program. Questionnaires and more complex designs, for example, mixed approaches using questionnaires combined with interviews or objective outcome data, were also employed in some cases.

One social prescribing project in inner London that underwent a full evaluation using a mixture of qualitative and quantitative methods also incorporated a matched control group [29]. No significant differences were found between groups based on the quantitative measures, although the qualitative data indicated beneficial effects with positive responses from the participants.

In the case of the other projects described as social prescribing for which full evaluations were found, a pre-post design was used [24,59,66,73]. Although positive outcomes were reported in, for example, physical activity, weight, blood pressure, mood, and social outcomes, (see Table 1 for the range of outcomes reported), this design did not allow firm conclusions on effectiveness to be drawn. One evaluation reported reduced hospital admissions and improved social outcomes [66]. Several evaluations focused on aspects related to implementing such a project in practice by exploring the perspectives of the patients, the health care providers, the role of the link worker, and the project governance to conclude with recommendations for future social prescribing projects [59].

While a formal, in-depth systematic appraisal of the methodology was not carried out, a number of issues were encountered when extracting data from the evaluations. It proved difficult in some reports to find the rationale, or occasionally the actual numbers, for the samples used in evaluations. It was also clear that missing data was an issue, particularly where follow-up was at several time points. This is, perhaps, not surprising considering that the aim of social prescribing and related interventions is to direct the patient to the community and voluntary sector rather than encourage follow-up in primary care. However, what was not always clear was how missing data were treated. Finally, as described earlier, referral was often based on psychological, social, or practical needs; this meant that the primary diagnosis was not often reported nor were the results provided by diagnosis.

Our study revealed limited evidence on social prescribing, specifically for type 2 diabetes in the United Kingdom and Ireland in peer-reviewed literature. Only 1 published study referred to social prescribing and that involved people with long-term conditions with no specific mention of type 2 diabetes. Nevertheless, broader Web-based searches demonstrated the existence of numerous “real-world” projects that investigated potentially relevant community-based interventions and had undergone evaluation using a variety of methods. Analysis of these projects also revealed the diverse ways in which the concept of social prescribing has been implemented in practice. The findings of this scoping study do not allow firm conclusions as to the effectiveness of social prescribing for people with type 2 diabetes. The results of the broader searches provide a clearer picture of the potential effects, possible outcomes, and processes involved, and thus inform research in this area. These evaluations also have the potential to provide valuable information for those organizations planning to implement similar services in future. This study has also revealed the different ways in which social prescribing may be described in practice, and this will aid those searching for evidence in future.

The challenges encountered when designing and conducting this study have led to a number of limitations. First, we attempted to identify evaluations of what is effectively a process, the referral of patients from primary care to a nonmedical, community-based activity, program, or service, rather than an intervention per se. In addition, some studies may not have referred specifically to social prescribing and described the intervention using different wording due to the lack of consensus on terminology in this area, thus precluding their identification. We attempted to address these points by using a range of several search strategies and an iterative process whereby we amended the later strategies according to the results of the initial searches. Although we extended the later database searches by including terms for prediabetes and a range of interventions, because of the recent proliferation of websites on the prevention of diabetes, we did not search specifically for diabetes prevention programs on the Web. Such programs, for example, those launched as part of the national Diabetes Prevention Program [110], are likely to use similar models and interventions to the projects included in this review, but further exploration of this aspect would be valuable. The selection of studies also relied on the referral process being reported clearly. The complexity of both identifying and assessing the evidence on social prescribing is demonstrated by the case of the X-PERT project. The RCT on X-PERT was initially excluded from our search because it did not include referral from primary care to an external service provider, whereas in practice, patients are referred from the general practice to the X-PERT educators via social prescribing. A similar situation occurred with later trials of the DESMOND program. Related to this point, it is debatable whether referring patients to a program as part of an RCT is adequately replicating the social prescribing process in practice. A conventional systematic review of published literature might, therefore, not represent a feasible approach to identify the entire literature and assess the effectiveness of the process.

A second potential limitation was that we limited the search to the United Kingdom and Ireland. The heterogeneity of national health systems, the plethora of diabetes websites, and the varied language used in describing this concept outside the United Kingdom would have proved a considerable challenge, had we attempted to cover the worldwide literature. This does mean that broadly similar projects in other countries were omitted from this review. Third, only information available on the Internet was retrieved and there was no direct contact with those delivering programs or services for further information, confirmation, or clarification. For this reason, it is possible we might have omitted relevant projects or incorrectly interpreted the information we were able to locate. Website information also evolved throughout the course of the study. Finally, by focusing on diabetes and including only those programs or services that specifically referred to people with this condition, we might have omitted relevant services or programs. Conversely, using a broad definition of the intervention resulted in the inclusion of projects and services that, perhaps, do not fit with the original concept of social prescribing as being a mechanism to refer from primary care to the voluntary and community sector. In some cases, the services, although nonmedical and community-based, were still provided by health professionals.

Nevertheless, our broad searches have demonstrated that social prescribing is prevalent in the United Kingdom, and that information is available that is potentially valuable but not currently in the published literature and therefore difficult to locate and access. At this point, our analysis on the range of projects and services, and the evaluations of these, may prove helpful.

It is clear from the findings of this review that different models have been implemented in practice, some of which fit more closely with the original concept of “social prescribing.” It is also clear that although the term “social prescribing” is more widely used in practice in the United Kingdom, it is not a concept that is currently recognized in the main medical databases or in other countries. Even within the United Kingdom, the link workers, who are a crucial part of many social prescribing services, are described in different ways. The findings from this scoping review do, however, suggest a preliminary model for social prescribing such as that presented in Figure 2. This will require further development and refinement.

It has been argued that the term “translational research” does not distinguish between testing of new treatments and research on how to implement new treatments in practice [10]. This is supported by our finding that the process of social prescribing, potentially the crucial step in the implementation process, is neglected in comparison to the intervention itself. A wide range of models of the process were revealed, for example, direct referral by the GP to the community sector, referral to the community via an in-practice link worker, or referral to a link worker based in the voluntary and community sector. Only 1 study attempted to address whether the process of prescribing an intervention had any influence by comparing referral to exercise programs in the community with tailored advice and information on local exercise facilities. None of the evaluations we located investigated, for example, how direct referral by a GP compared with referral via a link worker.

In general, many barriers exist that hamper the adoption of research into diabetes into clinical care at the level of the patient, primary care, health care system, community, and society [111]. Several countries have developed diabetes prevention programs for clinical care [112], whereas there has been limited translational research in the United Kingdom [113]. The effects of lifestyle programs in real-world primary care for patients with prediabetes or T2DM are small [114] and depend substantially on the program and its implementation [115]. Development, evaluation, and reporting of translational research programs need to be adapted to reflect the context and consider generalizability [116]. In exploring the effectiveness of implementation, various measures of success need to be taken into account and interpreted. For example, although in many cases, qualitative data in the evaluations found in our searches revealed a positive patient experience, this has not yet been supported by quantitative outcomes to the same degree. It is also clear that, in evaluating the effectiveness of these programs, a range of generic outcomes are used and the focus is often not on disease-specific indicators such as hemoglobin A1c. This is because social prescribing grew out of a perceived need to address broader social, behavioral, and practical issues that impact on, and may be caused by, the disease. Thus, the focus is on the person who has diabetes rather than on the disease itself. Interest in social prescribing is increasing, and its effects on specific groups such as those with diabetes are likely to be of interest, making it important to provide guidance on how real-world evaluations should be conducted and reported to ensure that the data collected can be more accurately compared and analyzed. The results of our analysis of real-world project evaluation will inform this guidance, while guidance such as that produced by Diabetes UK on commissioning to involve nontraditional providers in the support of people with long-term conditions is also valuable [117].

Our review also suggests a need to improve the reporting of implementation research, specifically service evaluations. Social, cultural, legal, demographic, and other factors that are controlled in trials become relevant influences during the process of implementation [118]. Combining these factors with the diversity in settings, methods, and outcomes of the real-world projects indicates that further development of existing reporting guidelines would be beneficial [119]. However, although standardization of reporting of evaluations would allow greater comparison across services and projects, a service evaluation is produced for a specific organization and to meet the specific requirements of that organization. Thus, increasing standardization of reporting, should this be considered feasible, must be balanced with the need to meet these requirements.

It has been suggested that evidence synthesis plays a pivotal role in developing or refining a framework of translational research [120]. A first step to enable this would be to better understand and define the multiple dimensions of translational research including implementation projects in real-world settings [10]. This might begin with a debate on reporting of different types of translational research, such has been proposed for Web-based health interventions [121]. Additionally, the challenges encountered in identifying relevant projects in this study indicate that it might prove fruitful to develop specific databases or networks to support larger-scale and more rigorous evaluation and comparison of programs and their relative cost-effectiveness.

Finally, a discussion might be necessary on how the principles of systematic reviews of health interventions can be applied to translational research, as the hierarchical model of evidence-based medicine has been shown valid only for certain questions such as efficacy [122]. It has been suggested that the study design alone may be an inadequate marker of the quality of evidence for interventions that are complex, program based, and dependent on context [123]. Consideration must be given to the evaluation process and its ability to detect effects and distinguish between failure of an intervention and failure of the delivery process. It is clear from this review that approaches incorporating a range of methods are necessary for evaluating the implementation of such complex interventions in practice. Novel approaches to synthesizing the evidence in this field are already underway [124].

As part of this discussion, it will be worth considering what role service evaluations of real-world projects play in the evidence base, and whether there are minimum standards for conducting such evaluations. Carrying out formal quality appraisal of these is complex, as they involve a range of data collection approaches, and it is unclear how best to accurately assess and compare the methodology used in individual evaluations. Although standardized tools and checklists exist for designs such as RCTs and qualitative studies, similar tools are not widely available for complex designs. As described earlier, there is also a question of how service evaluations may best be identified, as it is clear that the majority remain in the gray literature and would not be located via conventional systematic search techniques.

This review revealed the range of models of social prescribing and related nonmedical community-based services, and the extent of evaluation that has been carried out to assess the potential benefits of these, which would not have been achieved by searching databases alone. Although the evidence from these evaluations does not prove that social prescribing is an effective measure for implementing nonmedical interventions for patients with type 2 diabetes in the United Kingdom, there is sufficient to indicate the value of further evaluation and comparison, particularly if focused on real-world settings. The findings of this scoping study may inform future evaluation. Further research is necessary to better understand the communication between primary care and community or voluntary sector and to improve the documentation of the final step of translational research, implementing interventions in real-world settings.

Accessing Web-based information provides a potential method for investigating how specific innovative health concepts, such as social prescribing for type 2 diabetes, have been implemented in practice and the full extent of the evaluation of such innovations. Several challenges were encountered including defining the concept, focusing on process plus intervention, and systematically searching varied and evolving Web-based sources. Obtaining sufficient relevant information requires searching for and analyzing information from a range of sources. The methods and findings from this study have already informed a broader scoping exercise on the evaluation of social prescribing projects in the United Kingdom, which is underway. Further exploration of this approach will inform future research on the application of health-related concepts into practice.