Evidence synthesis – Outcomes and instruments used in social prescribing: a modified umbrella review

Abstract

Introduction: Previous social prescribing work highlights a range in the types and number of outcomes used in published studies. We aimed to describe social prescribing outcome core areas and instruments to build capacity for future research and program evaluation.

Methods: This was a modified umbrella review following standard guidelines. We registered the study and searched multiple databases (all languages and years); inclusion criteria were peer-reviewed publications containing outcomes for self-described social prescribing for adults aged 18 years and older. The last search date was 9 July 2023. From the included systematic reviews, we identified primary studies using the same inclusion criteria. For primary studies, we sorted extracted outcomes and instruments into six core areas using a published taxonomy. We located information on instruments’ description and measurement properties and conducted two rating rounds for (1) the quality of systematic reviews and (2) reporting of instruments in primary studies. We conducted a narrative synthesis of reviews, primary studies and outcomes (PROSPERO 2023 CRD42023434061).

Results: We identified 10 systematic reviews and 33 primary studies for inclusion in our review. Outcomes covered most core taxonomy areas, with an emphasis on psychosocial factors (e.g. well-being) and less emphasis on cognition, physical activity, and caregivers and volunteers. We noted few studies provided detailed information on demographic data of participants or measurement properties of instruments.

Conclusion: This synthesis provides an overview and identifies knowledge gaps for outcomes and instruments used in social prescribing interventions. This work forms the basis of our next step of identifying social prescribing–related outcomes that matter most across interested parties, such as individuals providers and decision makers.

Keywords: determinants of health, healthy aging, outcomes research, social prescribing

Introduction

Social prescribing is a health and social model of care with origins in the United Kingdom (UK) and is quickly spreading to many other locations around the globe, including Canada. It has been well described elsewhere,^{Footnote 1Footnote 2} but a key feature of social prescribing is the addressing of people’s unmet nonmedical social needs by connecting them to resources within the community. The most current comprehensive pathway for social prescribing^{Footnote 2} involves health and social providers, or community organizations working with people to identify social needs and making a nonmedical referral to a community navigator, sometimes called a “community link worker.” Together, within a strengths-based approach, the person and link worker identify “what matters most” and the link worker will connect, or even in some cases accompany, people to community assets such as a service, greenspace or network (depending on needs).^{Footnote 2Footnote 3}

Social prescribing is informed by other evidence-based work, for example, community referrals, navigator systems,^{Footnote 4} and the benefits associated with engaging in activities such as physical activity^{Footnote 5} and arts-^{Footnote 6} or museum-based programs.^{Footnote 7} However, we^{Footnote 8Footnote 9} and others^{Footnote 10} have noted definitive evidence is lacking on the effectiveness of social prescribing. One challenge to synthesizing evidence may be due to the previous lack of an accepted definition of social prescribing; however, a new definition is available, co-created through an international modified-Delphi approach.^{Footnote 3} In Canada, social prescribing is in the early stages of development, but is well supported by clinicians’ use of community referrals in practice and the large community-based sector of nonprofit, volunteer and other organizations that support people. The “new” definition and pathways of social prescribing are complemented by the integration of the health and social sectors and the strengths-based and person-centred approaches.^{Footnote 2}

Developing an evidence base for a complex intervention such as social prescribing requires considering many factors. One area we identified from our previous reviews,^{Footnote 8Footnote 9} which may be useful in advancing the science and practice of social prescribing, is related to the outcomes measured within research and program evaluations. Inconsistencies in using and reporting outcomes within trials and programs present challenges when determining the effectiveness of an intervention (such as social prescribing), ensuring person-centred, meaningful and important outcomes are included, and, later, when combining data for evidence syntheses.

Two previous studies used mapping review methods to identify and categorize outcomes in social prescribing.^{Footnote 11Footnote 12} While both reviews provide important knowledge and perspectives, we proposed to go further and identify outcomes used in social prescribing studies and compare them against a recently published taxonomy of outcomes for health and social interventions from Dodd and colleagues.^{Footnote 13} Reviewing currently used outcomes provides the opportunity to determine if any core areas or domains are missing from data collection and to describe current reporting practices.

Therefore, in this modified umbrella review, we aimed to describe outcomes, domains and instruments used in previous social prescribing studies for adults and older adults. The United Kingdom (UK) National Health Service (NHS) developed a Social Prescribing Common Outcomes Framework,^{Footnote 14} but to our knowledge there is not a core outcome set for social prescribing research. We proposed to address this knowledge gap,^{Footnote 15} and describe outcomes used in previous social prescribing research in advance of developing a core set.^{Footnote 16} Taken together, we approached this work to provide practical guidance for choosing outcome measures, with the overall aim of contributing to the science that underpins social prescribing.

Methods

We conducted a modified umbrella review following the guidelines outlined by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement^{Footnote 17} and the Joanna Briggs Institute (JBI) standards for conducting an umbrella review, or review of reviews.^{Footnote 18} We modified the review process by screening (via two authors, independently) each primary study within included reviews to confirm it met our inclusion criteria. We made the changes for two main reasons: (1) there was a wide age range and grey literature within the identified systematic reviews, and some of the primary studies did not meet our inclusion criteria; and (2) there was an overlap of primary studies across included reviews.

We registered the protocol with the international Prospective Register of Systematic Reviews (PROSPERO; 2023 CRD42023434061)^{Footnote 19} before conducting searches with our final strategy. Our two exploratory research questions were: (1) What outcomes have been used in social prescribing research for community-dwelling adults aged 18 years and older? and (2) How do the identified instruments map onto the taxonomy of outcome core areas and domains?

Eligibility criteria

We included peer-reviewed systematic reviews and primary studies from all languages and all years that synthesized quantitative data for self-described social prescribing interventions for adults aged 18 years and older, and across the continuum of settings, such as hospitals primary care, community settings, etc. We included programs that have been described as “social prescribing,” such as arts-based or museum-based programs (known as “arts-on-prescription” or “museum-on-prescription”). We included evidence for adults and older adults for this synthesis, as this was our focus for developing a core set of outcomes. Further, younger and older populations have distinct needs and services, and therefore we did not include outcomes from younger age groups in this synthesis.

We excluded publications that did not provide any outcome measures. We made the decision to include only peer-reviewed literature, because we wanted to compare reporting in the primary studies, which may be more likely to follow research reporting guidelines.

The following were our criteria based on the population, intervention, comparator and outcome (PICO) framework. Population: we focussed on peer-reviewed evidence that included adults aged 18 years and older receiving social prescribing. Intervention: we included reviews and primary studies that self-identified as social prescribing. Comparator: studies included in the reviews could have any or no comparator. Outcomes: as our aim was to identify all possible outcomes, we did not place any limits on this component.

Information sources and search strategy

We searched the databases listed below; the date of our last search was 9 July 2023. One author (MCA) ran all of the searches and uploaded identified citations into Covidence systematic review software (Veritas Health Innovation, Melbourne, AU). We also conducted a forward and backward (reference list) search for peer-reviewed publications from all included reviews based on citations downloaded from Web of Science or Google Scholar. The databases, along with keywords, were:

• Ovid MEDLINEÒ and Epub Ahead of Print, In-Process, In-Data-Review & Other Non-Indexed Citations, Daily and Versions and Embase. Keywords: ("social prescribing" or "social prescription"). ab,ti. AND "systematic review".ab,ti.
• EBM Reviews - Cochrane Database of Systematic Reviews. Keywords: "social prescribing" OR "social prescription" AND "systematic review"
• EBSCO (APA PsycArticles, APA PsycInfo, CINAHL Complete, Social Work Abstracts, SPORTDiscus) social prescribing or social prescription (title) AND systematic review
• Epistemonikos "social prescribing" OR "social prescription" AND "systematic review" title and abstract
• Google Scholar "social prescribing" OR "social prescription" AND "systematic review" title

Selection process

All systematic reviews identified through search strategies were independently screened at Level 1 (titles and abstracts) and Level 2 (full text) by two authors (IKS, MCA) based on the inclusion criteria described above. We reviewed the primary studies from each included review and categorized them as peer-reviewed or unpublished studies/evaluations. We then reviewed the primary studies (e.g. a separate round of Level 1 and 2 screening) to decide if they met our inclusion criteria.

Data collection process

We extracted the following information for each review: author, publication year, systematic review question(s) and aims, population, setting, demographic information, summary of findings, and outcome measures or instruments. We also compared across reviews to look for overlap of primary studies to better understand data contributing to findings. For each included peer-reviewed primary study, we extracted the following information: author, publication year, population, setting, social prescribing intervention, and descriptive and outcome data collected. For this phase, one author (IKS or MCA) extracted information from studies in Covidence and Excel, and two other authors (EE and HA) reviewed and confirmed extracted findings. In the case of discrepancies between reviewers, a third review author (AC) made the final decision.

Sorting process

For each primary study, we extracted data on quantitative outcomes and sorted them based on a published taxonomy;^{Footnote 13} we chose this taxonomy because it was developed to use in determining core outcome sets. The original taxonomy has five core areas: death, physiological and clinical, life impact, resource use, and adverse events. Within the core areas there are 38 categories or domains. Two authors (IKS, MCA) independently sorted outcomes into core areas and domains following the guidelines provided by the taxonomy,^{Footnote 13} with two modifications: (1) we changed the domain “psychiatric outcomes” to “mental health” in the physiological/clinical core area; and (2) we moved the domain “delivery of care” to its own core area.

One author (MCA) created the first table of sorted outcomes from the previous step, and then all other authors (IKS, EE, HA, AC) reviewed the table. We also reviewed and compared the extracted outcomes with the NHS Social Prescribing Common Outcomes Framework,^{Footnote 14} which has four main areas: impact on the person (control and well-being, physical activity, management of daily life activities, connection); impact on community groups (confidence, impact of taking referrals, impact of social prescribing, changes in number of volunteers, capacity of the volunteer sector, and support needed); impact on the health and social care system (provider visits, medications, “morale of staff in general practice and other referral agencies”^{Footnote 14,p.30}); and other data collection (referrals, “equality monitoring,”^{Footnote 14,p.31} contacts with link workers, satisfaction).^{Footnote 14} Finally, for extracted instruments, we located information on measurement properties for a similar population (community-dwelling adults), when possible.

Assessment of systematic reviews and primary studies

We used the JBI critical appraisal tool^{Footnote 18} to analyze systematic reviews included in this synthesis. For each primary study, we compared the outcome reporting against one criterion proposed by the Consolidated Standards of Reporting Trials (CONSORT) 2020 Extension for Outcomes:^{Footnote 20} “Item 6a.8. Provide a description of the study instruments used to assess the outcome (e.g. questionnaires, laboratory tests) along with reliability, validity and responsiveness in a population similar to the study sample.”^{Footnote 20,p.2254} We reviewed each study to locate the term within the publication; this could include either  stating an instrument was valid, reliable and/or responsive, or providing a reference or measurement statistic. We also compared extracted demographic information for each primary study with the criteria proposed by PROGRESS-Plus^{Footnote 21}: age, gender/sex, ethnicity/culture/language/race, education, occupation, place of residence, religion, social capital and socioeconomic status.

Synthesis methods

We conducted a narrative synthesis, including compiling and sorting a list of outcomes and instruments contained within reviews and primary studies.

Potential review biases

We considered bias throughout our review process. A priori, we tried to address potential bias by following standard procedures and registering and updating our protocol. We planned for conflicts of interest, such as if an author on this synthesis were also to be an author on an included review or primary study. In this case, another author, not in conflict of interest, would provide a rating of the review.

Team membership consisted of trainees and researchers, and two of the co-authors had clinical training. Our team also spanned the age range from young adult to older adult; team members had experience with multiple research methods, and authors with experience in systematic reviews mentored less experienced team members. However, none of the authors had direct, lived experience with social prescribing as defined in this work.

We acknowledge that we only included peer-reviewed studies, and that this may create a publication bias, but this was intentional in order to identify and evaluate the reporting for outcome measures for studies that usually follow standard reporting guidelines (e.g. CONSORT 2020, or similar statements based on different study designs). However, by not including unpublished literature, we may have missed some outcomes, in particular as they may relate to implementation of a program (e.g. via a process evaluation).

Results

Study selection

After two rounds of Level 1 and 2 screening, we included 10 systematic reviews (Figure 1A) and 33 primary studies (Figure 1B).

Figure 1. PRISMA flow diagram for umbrella review (A) and individual studies (B)

Figure 1A : Descriptive text

PRISMA flow diagram for umbrella review

Step number	Name of step	Number of elements at the start of the step	Details	Number of elements removed before next step
1	Identification	(n = 918)	Citation searching (n = 762) Studies from databases/registers (n = 156) MEDLINE (n = 60)  Epistemonikos (n = 48)  MEDLINE and Embase (n = 17) EBSCO (n = 16)  Google Scholar (n = 14)  Embase (n = 1)	Duplicate studies (n = 313)
2	Screening	(n = 605)	Studies screened	Studies excluded (n = 575)
3	Screening	(n = 30)	Studies sought for retrieval	Studies not retrieved (n = 0)
4	Screening	(n = 30)	Studies assessed for eligibility	Studies excluded (n = 20), as follows: Not defined social prescribing (n = 7) Not a systematic review (n = 6) No outcomes (n = 5) Wrong study aims (n = 2)
5	Inclusion	(n = 10)	Systematic reviews included	N/A

In the end, (n = 10) systematic reviews were included.

Figure 1B : Descriptive text

PRISMA flow diagram for individual studies

Step number	Name of step	Number of elements at the start of the step	Details	Number of elements removed before next step
1	Identification	(n = 110)	Primary studies from included reviews	Duplicates removed (n = 32)
2	Screening	(n = 78)	Studies screened	Studies excluded (n = 29)
3	Screening	(n = 49)	Studies sought for retrieval	Studies not retrieved (n = 0)
4	Screening	(n = 49)	Studies assessed for eligibility	Studies excluded (n = 16), as follows: Social prescribing not stated (n = 9) Wrong study design (n = 6) Conference abstract (n = 1)
5	Inclusion	(n = 33)	Primary studies included in review	N/A

In the end, (n = 33) primary studies were included in review.

Note: PRISMA template from Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Int J Surgery. 2021;88:105906. https://doi.org/10.1016/j.ijsu.2021.105906

Systematic review characteristics (n = 10)

There were six systematic reviews with authors located in the UK,^{Footnote 10Footnote 22Footnote 23Footnote 24Footnote 25Footnote 26} and one study each from Canada,^{Footnote 8} Germany,^{Footnote 27} Ireland^{Footnote 28} and Portugal.^{Footnote 29} Table 1 provides a summary of the systematic reviews, including the study aims and the number of included primary studies. Overall, participants’ descriptive information was missing from several systematic reviews;^{Footnote 10Footnote 25Footnote 26} however, this may be because the information was missing from some of the primary studies.

Table 1. Summary of information for the 10 included systematic reviews

First author Location Year	Number of information sources	Number of peer-reviewed studies Total and included in outcome synthesis	What systematic review aimed to determine	Findings
BickerdikeFootnote 10 UK 2016	16 reports 15 studies	5 2Footnote 48Footnote 59	“… effectiveness of social prescribing programs relevant to the UK NHS setting” (p. 2)	Most studies had positive findings Included studies had methodological limitations including possible confounding factors “[A] lack of standardized and validated measuring tools” (p. 15)
CooperFootnote 23 UK 2022	17 reports 13 studies	13 14Footnote 7Footnote 34Footnote 35Footnote 36Footnote 37Footnote 38Footnote 39Footnote 40Footnote 41Footnote 43Footnote 45Footnote 46Footnote 52Footnote 57	“… effectiveness, and active ingredients of UK-based social prescribing interventions targeting mental health” (p. 2)	Positive outcomes in 12/13 studies Methodological limitations with high attrition rates “[S]ubstantial variability in outcome measures” (p. 11)
CostaFootnote 29 Portugal 2021	13 studies	13 11Footnote 36Footnote 39Footnote 43Footnote 45Footnote 46Footnote 47Footnote 48Footnote 49Footnote 50Footnote 51Footnote 52	“[T]o locate and summarize evidence regarding the effectiveness of SP targeting the adult population assisted in primary health-care settings.” (p. 3)	8/10 studies reported positive physical or psychological well-being outcomes Evidence regarding the effectiveness of SP continues to be unclear “More studies are needed to establish the adequate and more standardized outcome measurement tools” (p.14)
KielyFootnote 28 Ireland 2022	9 reports 8 studies	9 2Footnote 50Footnote 52	“… effects on health outcomes and costs of social prescribing link workers … for people in community settings focussing on people experiencing multimorbidity and social deprivation” (p. 1)	Only identified social prescribing may improve self-rated health “[O]verall certainty of the evidence for our selected outcomes, which was low or very low for most outcomes.” (p. 11) “… there was a lot of variation in outcomes included and how they were measured ...” (p. 11)
NapieralaFootnote 27 Germany 2022	68 reports 53 studies	33 18Footnote 30Footnote 31Footnote 35Footnote 38Footnote 39Footnote 41Footnote 42Footnote 44Footnote 45Footnote 47Footnote 48Footnote 49Footnote 50Footnote 51Footnote 52Footnote 55Footnote 58Footnote 60	“… effectiveness of SP for facilitating psychosocial support with an international focus” (p. 2)	Uncontrolled trials had positive findings Limitations of evidence due to confounding and other possible sources of bias “Most frequently studied outcomes represent the domains of mental health and well-being, loneliness, quality of life, general health, self-efficacy, and health care utilization.” (p. 8)
PercivaFootnote 18 Canada 2022	7 studies	7 7Footnote 7Footnote 33Footnote 44Footnote 49Footnote 51Footnote 54Footnote 56	“… effect of social prescribing for older adults within primary care” (p. 2)	“There were some positive effects of social prescribing on physical and psychosocial outcomes (e.g. social participation, well-being)” (p. 1) Limitations noted in methodology (study design, small sample sizes, missing data) Outcome measures reported included: physical and psychosocial outcomes and health resource use
PeschenyFootnote 46 UK 2019	16 studies	7 4Footnote 48Footnote 49Footnote 51Footnote 52	“… outcomes of social prescribing programs based on primary care and involving navigators” (p. 664)	“The existing evidence for improvements in health and well-being, health-related behaviours, self-concepts and daily functioning is mixed.” (p. 670) “… the quality of the majority of included studies was poor … sparse data on numbers of participants and loss of follow-up, a non-comprehensive sampling strategy, and a lack of information on data collection and analysis methods.” (p. 670) “Although we found that studies are interested in similar outcome areas, the tools to measure these outcomes varied between studies.” (p. 670)
ReinhardtFootnote 24 UK 2021	9 studies	2 1Footnote 39	“… impact of social prescribing (SP) programs on loneliness among participants and the population” (p. 204)	“[S]ocial prescribing models designed to address loneliness have been largely viewed as helpful by both participants and service providers” (p. 211) “[V]ariability and paucity of evidence and lack of control group comparisons make it difficult to draw conclusions” (p. 211) “We demonstrate a gap between social prescribing design and social prescribing evaluation and illuminate a lack of impact assessment in relation to social care. We also note a lack of consensus on what the impact of a person-centred approach such as social prescribing should be.” (p. 12)
ThomasFootnote 25 UK 2021	8 studies	8 1Footnote 61	“… SP that engage communities in co-design and co-production leading to improvements in well-being as well as examine barriers and facilitators to SP intervention development” (p. 3)	“Well-being outcomes across the included studies were reported to have been an increase in confidence, empowerment, and self-sufficiency as well as reduction in social isolation.” (p. 10) “[A]ll included studies were of a low-quality standard.” (p. 11) Missing data (e.g. study duration and participant numbers) “[V]arious data collecting methods were used within each study” (p. 11)
VidovicFootnote 26 UK 2021	51 studies	33 13Footnote 30Footnote 32Footnote 38Footnote 39Footnote 40Footnote 42Footnote 44Footnote 46Footnote 47Footnote 50Footnote 51Footnote 52Footnote 53	“…impact of SP on addressing loneliness, social isolation, well-being, and connectedness… at the individual, system, and community levels.” p. 1	“[A] majority of studies in this report find change following a social prescribing intervention, but not necessarily change due to a social prescribing intervention.” (p. 14) “Given this small number of studies that establish causality, conclusions regarding the impact of social prescribing are tentative, at best.” (p. 14) “[V]ariability in the types of measures used in regard to the four outcomes …” (p. 11). “[W]e recommend standardizing measurement and data collection to help deliver stronger, more reliable, and more rigorous evidence.” (p. 14)

Almost all of the reviews aimed to look at the general effect or impact of social prescribing, with some reviews reporting more specific criteria such as social prescribing from one location (UK),^{Footnote 10Footnote 23} or focussed on loneliness,^{Footnote 24Footnote 26} mental health,^{Footnote 23} primary care^{Footnote 8Footnote 22Footnote 29} or older adults.^{Footnote 8} One review focussed on social prescribing interventions that used a co-design or co-production approach.^{Footnote 25} The review by Vidovic and colleagues^{Footnote 26} provides an overview of outcomes used in social prescribing for four key measures: loneliness, social isolation, well-being and connectedness. Systematic reviews consistently noted positive outcomes but with limited evidence due to small sample sizes and methodological challenges with primary studies (Table 1). Most systematic reviews commented on the variability in the outcome measures.^{Footnote 10Footnote 22Footnote 23Footnote 25Footnote 26Footnote 28Footnote 29}

Primary study characteristics (n = 33)

Table 2 provides an overview of the 33 primary studies. Overall, 29 (88%) studies were conducted in the UK; three studies were from Australia,^{Footnote 30Footnote 31Footnote 32} and one was from South Korea.^{Footnote 33} Studies were published between 2008 and 2022, with 19 (58%) studies published from 2019 onwards, but before the publication of the social prescribing definition by Muhl and colleagues in 2023.^{Footnote 3} Four studies^{Footnote 34Footnote 35Footnote 36Footnote 37} were based on arts-on-prescription, “… part of mainstream social prescribing provision in primary health care.”^{Footnote 37,p.1} Two studies were based on museums-on-prescription.^{Footnote 7Footnote 38} Two studies^{Footnote 39Footnote 40} were based on the social cure approach, which “suggests that social identities can enhance health/well-being through psychological resource provision.”^{Footnote 40,p.387} We noted overlap for studies included in the 10 systematic reviews. Over half of the primary studies 19 (58%) appeared in more than one review, with a range in the number of times they were included: two;^{Footnote 7Footnote 30Footnote 35Footnote 36Footnote 40Footnote 41Footnote 42Footnote 43} three;^{Footnote 7Footnote 44Footnote 45Footnote 46Footnote 47} four;^{Footnote 48Footnote 49Footnote 50} five;^{Footnote 39Footnote 51} and six.^{Footnote 52}

Table 2. Summary of information for the 33 primary studies

First author Location Year	Study characteristics	Target population Age	N	Program description
AggarFootnote 30 Australia 2020	Retrospective analysis of pre- and post-intervention design	Adults and older adults 18–65 years	175	12-week social prescribing program with link workers Social services; enrolment in social and therapeutic activities included arts and crafts, yoga and relaxation, equine therapy and social groups
AggarFootnote 31 Australia 2021	Exploratory, quantitative, longitudinal design	Adults and older adults 18–65 years	13	10-week; facilitated by mental health social worker Arts and crafts group led by practising artist/instructor
BirdFootnote 53 UK 2019	Mixed methods design	Adults with long-term conditions [type 2 diabetes, pre-diabetes and/or hypertension] and who were inactive 18–70 years or older	602	12-week program that included a GP, activity program manager and exercise specialists who delivered the program Physical activity intervention
CarnesFootnote 52 UK 2017	Mixed methods design with a matched control group	People from general practices who were frequent attenders and/or socially isolated	486	Up to 6 sessions GP referral to social prescribing coordinator who co-created an action plan with well-being goals Sometimes community referrals were initiated Support via social prescribing volunteers
CroneFootnote 34 UK 2013	Prospective longitudinal follow-up design (observational)	Adults Mean (SD) age: 54 (16) years	157	10-week art program (e.g. poetry, ceramics, drawing, mosaic, painting) GP or provider referral Within a GP clinic or community location
CroneFootnote 35 UK 2018	Prospective longitudinal follow-up design (observational)	Adults Mean (SD) age: 51.2 (15.9) years	818	10-week art program (e.g. poetry, ceramics, drawing, mosaic, painting) GP or provider referral Within a GP clinic or community location
ElstonFootnote 44 UK 2019	Pre- and post-intervention design	Adults and older adults with two or more long-term conditions > 50 years	Pre-intervention: 1046 Post-intervention: 82	12-week, strengths-based, with coordinators (link workers) GP or other people/services who make the referral Coaching, advocacy and navigation of and access to local health, social and economic services
FosterFootnote 41 UK 2021	Concurrent mixed methods design	People who were experiencing, or at risk of, loneliness Mean (SD) age: 65.5 (19.3) years	2250	12-week social prescribing service with referrals from any source and link worker support Navigated access to community activities and resources
GiebelFootnote 54 UK 2021	Pre- and post-intervention design	People with a diagnosis of dementia (any subtype and age) and their family carer Mean (SD) age: 74.0 (8.0) years	25	Referred by psychiatrist or GP Collaboration with local NHS Trust, clinics, council, clinical commissioning group, other organizations, unpaid dementia carers and dementia care navigators Classes were ongoing, 60-minute sessions at local community centres Variety of activities such as low-impact exercises, local walks, tai chi, relaxation techniques, mindfulness and games, etc. Assessment at 3 and 6 months
GrayerFootnote 48 UK 2008	Pre- and post-intervention design	Adults with a psychosocial problem Mean (SD) age: 43.1 (14.6) years	75	Referrals from primary health care team Graduate primary care mental health workers met with people to identify needs and make referrals to community resources Assessment at 3 months
HoltFootnote 42 UK 2020	Multi-level, repeated measures design	Adults with anxiety and depression Mean age: 47 years	66	12-week arts-on-prescription workshops Led by artist and health provider
HowarthFootnote 55 UK 2020	Case study	Adults and older adults Between 30 and 85 years Age range: most people above 60 years	9	12-week, nature-based social prescription Referrals to gardening Provides a space to grow, reflect, and meet
JonesFootnote 57 UK 2013	Pre- and post-intervention design	Adults and older adults with low-level mental, long-term health conditions, low levels of physical activity, diet-related health, low income and/or social isolation Age range: 18–70 years or older	687	Group of 10 (out of 15) community organizations delivering activities such as leisure, exercise, cooking, befriending, arts and crafts activities Group activities in 10-week blocks of 2-hour sessions Self-referral, health care agency or another agency Baseline assessment occurred months 1–3, with follow-up between months 4–6
JonesFootnote 56 UK 2020	Pre- and post-intervention design	Participants aged 55+, their families, staff, the NHS and local government Mean (SD) age: 73.8 (8.9) years	66	Health Precinct, a partnership with a health board and public health 16 weeks in duration, with a focus on exercise goals, with allied health or nursing Common referral path via an exercise referral scheme or cardiac rehab Activities to encourage social participation and physical activity
KelleziFootnote 39 UK 2019	Mixed methods approach with longitudinal survey	GPs, health care providers, service users > 29 years	630	Used a “social cure” approach, e.g. group membership’s effect on social life, health and well-being Referrals from GP practices and person meets with health coach; link workers connect people to social groups Assessments completed 4 months apart
KimFootnote 33 South Korea 2021	Pre- and post-intervention design	Older adults aged > 65 years Mean (SD) age: 82.0 (5.9) years	10	Once weekly for 10-week program: music storytelling, a self-help group and gardening Review by public health doctor Two program coordinators attended sessions
KimberleeFootnote 58 UK 2016	Pre- and post-intervention design	Adults 36-40 years	70	People referred from local GPs Person collaborates with staff to co-produce goals Can also access community resources Program was ongoing, but assessments occurred at baseline and at least 3 months later
LoftusFootnote 49 UK 2017	Quality improvement design	Older adults (65 years+) with a chronic condition with frequent GP visits or multiple medications Mean (SD) age: 72.1 (6.8) years	28	12-week program Referrals from GP to social prescribing coordinator Variety of activities such as social clubs, counselling, arts program, falls prevention, exercises classes, etc.
MaughanFootnote 59 UK 2016	Observational study	Adults with a mental health condition	55	Aimed to connect primary care and community services Program duration between 6 and 18 months (maximum 20 times) Asset map exercise to identify resources with a focus on mental health awareness and lifestyle change
MercerFootnote 50 UK 2019	Quasi-experimental cluster-randomized controlled trial	Adults aged > 18 years Mean (SD) age: 49.0 (16.0) years 56.0 (15.0) years	214	Community link worker/coordinator connected practices (referring GPs and nurses) and local community organizations Intervention was needs-dependent Baseline and follow-up questionnaire 9 months later
MortonFootnote 45 UK 2015	Pre- and post-intervention design	Adults Mean (SD) age: 52.0 (11.0) years	136	Most people referred by health providers, but some referrals from charitable organizations or others Provided series of free activities such as meditation, arts and crafts, etc.
PeschenyFootnote 46 UK 2019	Pre- and post-intervention design	Adults from primary care Mean (SD) age: 51.2 (15.7) years	56	12 sessions to free programs usually organized by third sector organizations GP referral to navigator Services include advice, physical activity, gardening, social activities, stress management and creative activities
PeschenyFootnote 60 UK 2021	Pre- and post-intervention design	Adults from primary care Mean (SD) age: 50.3 (16.8) years	68	12 sessions to free programs usually organized by third sector organizations GP referral to navigator Services include advice, physical activity, gardening, social activities, stress management and creative activities
PoulosFootnote 32 Australia 2019	Pre- and post-intervention design	Older adults (65 years+) Mean (SD) age: 78.1 (8.0) years	127	Arts-on-prescription classes (6–8 participants per class) were held weekly for 8–10 weeks Health provider referral Community care worker or volunteer was available during each class to assist participants and artists Courses available: visual arts, photography, dance and movement, drama, singing and music
SumnerFootnote 36 UK 2020	Cohort study	Adults Mean (SD) age: 51.1 (15.9) years	1297	8–10-week arts-on-prescription intervention Referrals from GPs and other providers Courses available: creative arts such as painting, ceramics, playwriting and mosaics
SumnerFootnote 37 UK 2021	Cohort study	Adults Mean (SD) age: 50.5 (15.7) years	245	Two 8-week arts-on-prescription interventions Referrals from GPs and other providers Courses available: creative arts such as painting, ceramics, playwriting and mosaics
SwiftFootnote 61 UK 2017	Case study	Primary care patients	> 5000	Community Wellbeing Practices program GP referrals Community navigators with up to 12 months of contact Provides access to educational courses, hobby and interest groups and volunteering opportunities
ThomsonFootnote 7 UK 2018	Repeated measures design	Older adults at risk of loneliness and social isolation 65–94 years	115	10 weekly sessions of museum-based programs Referrals from health and social care, and third sector organizations
ThomsonFootnote 38 UK 2020	Exploratory sequential mixed methods design	Adult mental health service users 44–70 years Mean age: 53 years	46	10 weekly sessions Referred via community partners (mental health nurse) Outdoor horticultural activities and indoor nature-based creative activities Delivered by a horticultural specialist, an arts tutor and a museum volunteer
van de VenterFootnote 43 UK 2015	Pre- and post-intervention mixed methods design	People with mild-to-moderate mental health problems 27–73 years Mean age: 43 years	44	20-week arts-on-referral program 4 artist-facilitated groups, held at GP practices or community centres Included: painting, textiles, music, photography and film
VogelpoelFootnote 51 UK 2014	Mixed methods design	Older people with sensory impairments 61–95 years Average age: over 80 years	12	12-week program for people with sensory impairments GP referral Practical arts workshop program Visual and tactile arts facilitator Travel support via transport options including community transport, taxis, buses and service transport
WakefieldFootnote 40 UK 2022	Multimethod longitudinal design	Adults with long-term health conditions or who reported being isolated/lonely/anxious 17–85 years Mean (SD) age: 52.7 (14.8) years	63	Used a “social cure” approach, e.g. group membership’s effect on social life, health and well-being Referrals from GP practices and person meets with health coach; link workers connect people to social groups Baseline and follow-up assessment at 6–9 months
WoodallFootnote 47 UK 2018	Mixed methods design	Adults and older adults 16–94 years Mean age (SD): 53.1 (18.0) years	342	6-session program, although most people completed it at 16 weeks (mean: 10 weeks) Self-referral or via referrals from GPs, health, social care and other relevant providers Coordinators provided information on local groups and services

Assessment of systematic review methods and instrument reporting

Most systematic reviews were in agreement with the JBI critical appraisal criteria,^{Footnote 18} with all but two reviews^{Footnote 24Footnote 26} scoring eight or higher (out of 11 possible points). The question (from the JBI tool) rated with the most “no” or “unclear” responses was related to reporting the research question based on PICO format. Of primary studies, most publications did not provide detailed information on instruments’ basic measurement properties. Validity was most often mentioned or referenced,^{Footnote 7Footnote 30Footnote 31Footnote 32Footnote 33Footnote 34Footnote 36Footnote 37Footnote 41Footnote 42Footnote 43Footnote 44Footnote 47Footnote 48Footnote 50Footnote 51Footnote 53Footnote 54Footnote 55Footnote 56Footnote 57Footnote 58Footnote 59} followed by reliability^{Footnote 33Footnote 36Footnote 38Footnote 42Footnote 43Footnote 48Footnote 53Footnote 54Footnote 57} and responsiveness.^{Footnote 37Footnote 42Footnote 45Footnote 60} Few studies provided specific information, such as a measurement statistic (e.g. Cronbach alpha or kappa).

Results of synthesis

Categorization of outcomes

Many extracted instruments were patient-reported outcomes measures (PROM) focussed on well-being, with variability in the number and types of outcomes used; there were some patient-reported experience measures (PREM; e.g. satisfaction). Figure 2 and Table 3 provide an overview of outcomes and information from primary studies. Overall, almost all taxonomy core areas were represented, except death, but most domains contained instruments from only a few studies (e.g. physical functioning such as physical activity), with some exceptions. Specifically, over half of primary studies used at least one of six different well-being instruments, although most studies used one of two outcome instruments: Warwick-Edinburgh Mental Well-being Scale (WEMWBS)^{Footnote 32Footnote 34Footnote 35Footnote 36Footnote 37Footnote 42Footnote 43Footnote 44Footnote 45Footnote 47Footnote 51Footnote 53} and Short Warwick-Edinburgh Mental Well-being Scale.^{Footnote 41Footnote 54Footnote 55Footnote 60Footnote 61} Many studies had a specific level of  cognition as an inclusion criterion, but only two studies^{Footnote 56Footnote 62} reported a related instrument; however, the WEMWBS aims to “capture a wide conception of well-being, including … cognitive-evaluative dimensions.”^{Footnote 63,p.2} Only one study reported on adverse events (unintended outcomes).^{Footnote 56} Table 4 lists identified instruments within primary studies, alongside a description and some measurement properties.

Figure 2. Overview of outcomes and information from primary studies based on the published taxonomy

Source of published taxonomy: Dodd S, Clarke M, Becker L, Mavergames C, Fish R, Williamson PR. A taxonomy has been developed for outcomes in medical research to help improve knowledge discovery. J Clin Epidemiol. 2018;96:84-92. https://doi.org/10.1016/j.jclinepi.2017.12.020
Abbreviations: ED, emergency department; GP, general practitioner.

Table 3. Outcomes extracted from primary studies, categorized using the published taxonomy, by core area and domain, showing instrument and citing primary study

Core area	Domains
Physiological/clinical	General outcomes (n = 4) Burden of multimorbidityFootnote 50 General healthFootnote 52 Number/presence of chronic conditionsFootnote 33Footnote 35Footnote 37Footnote 52 Pain scaleFootnote 30	Mental health (n = 9) Center for Epidemiological Studies Depression scale (CES-D-7)Footnote 57 General Anxiety Disorder-7 (GAD-7)Footnote 37Footnote 58 Geriatric Depression Scale Korean Version (GDS-K)Footnote 33 General Health Questionnaire-12 (GHQ-12)Footnote 48 Hospital Anxiety and Depression Scale (HADS)Footnote 45Footnote 50Footnote 52 Kessler Psychological Distress Scale (K10)Footnote 30Footnote 31 Patient Health Questionnaire-9 items (PHQ-9)Footnote 58Footnote 61 Patient Health Questionnaire-8 items (PHQ-8)Footnote 37 Short Mood ScaleFootnote 42		Metabolism and nutrition (n = 2) BMIFootnote 53Footnote 57 Energy expenditureFootnote 46	Musculoskeletal and connective tissue (n = 1) Muscle strengthFootnote 53
Life impact	Physical functioning (n = 6) FrailtyFootnote 32 General Practitioner Physical Activity Questionnaire (GPPAC)Footnote 57 International Physical Activity Questionnaire (IPAQ)Footnote 46Footnote 53Footnote 58 Number of creative, lifestyle or regular activitiesFootnote 32Footnote 50Footnote 52 Participation in sportFootnote 53 Rockwood Clinical Frailty Scale (RCFS)Footnote 44	Social functioning (n = 4) Community belongingFootnote 39Footnote 40 Friendship ScaleFootnote 58 Number of social groupsFootnote 39Footnote 40 Social Participation ScaleFootnote 33		Role functioning (n = 3) Capacity to workFootnote 30 Paid employmentFootnote 31 Work and Social Adjustment Scale (WSAS)Footnote 48Footnote 50	Emotional functioning/well-being (n = 18) Campaign to End Loneliness Measurement ToolFootnote 47Footnote 56 Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM)Footnote 48 General Self-Efficacy Scale (GSE)Footnote 45 General Self-Efficacy Scale (GSE) Korean VersionFootnote 33 Life satisfactionFootnote 57 Museum Well-being Measure for Older Adults (MWM-OA)Footnote 7Footnote 38 ONS Well-beingFootnote 58 Patient Activation Measure (PAM)Footnote 44 Perceived creativityFootnote 32 Rosenberg Self-Esteem ScaleFootnote 56 Rosenberg Self-Esteem Scale Korean VersionFootnote 33 Self-efficacyFootnote 31 Southwest Well-being QuestionnaireFootnote 57 UCLA Loneliness ScaleFootnote 30Footnote 31Footnote 33Footnote 41 Short-form UCLA Loneliness Scale (ULS-8)Footnote 39Footnote 40 Warwick–Edinburgh Mental Well-being Scale (WEMWBS)Footnote 32Footnote 34Footnote 35Footnote 36Footnote 37Footnote 42Footnote 43Footnote 44Footnote 45Footnote 47Footnote 51Footnote 53 Short Warwick–Edinburgh Mental Well-being Scale (SWEMWBS)Footnote 41Footnote 54Footnote 55Footnote 60Footnote 61 Well-being StarFootnote 44
	Cognitive functioning (n = 2) 6 Item Cognitive Impairment Test (6CIT)Footnote 56 Korean Mini-Mental State Examination (K-MMSE)Footnote 33	Global quality of life (n = 1) World Health Organization Quality of Life–abbreviated 26 item (WHOQOL-BREF)Footnote 30Footnote 31		Perceived health status (n = 4) EuroQol-5 Dimensions (EQ-5D)Footnote 30Footnote 31Footnote 40Footnote 47Footnote 50Footnote 56Footnote 61 General Health Score (GHS)Footnote 52 Investigating Choice Experiments for the Preferences of Older People Capability Measure (ICE-CAP A)Footnote 50 Measure Yourself Medical Outcome Profile 2 (MYMOP2)Footnote 52	Personal circumstances (n = 2) Index of Multiple Deprivation (IMD)Footnote 34Footnote 35Footnote 36Footnote 37Footnote 46Footnote 50Footnote 52Footnote 54Footnote 60 Camberwell Assessment of Need Short Appraisal Schedule (CANSAS-P)Footnote 30Footnote 31
Delivery of care	Adherence/compliance (n = 2) AttendanceFootnote 32Footnote 33Footnote 34Footnote 35Footnote 36Footnote 38Footnote 43Footnote 46Footnote 53Footnote 54Footnote 56Footnote 58 CompletionFootnote 34Footnote 35Footnote 55	Patient/carer satisfaction (n = 5) Client Satisfaction Questionnaire (CSQ)Footnote 48 ExpectationsFootnote 30 Perceived benefitsFootnote 32 Program satisfactionFootnote 30Footnote 33Footnote 35Footnote 47Footnote 52Footnote 57Footnote 61 Satisfaction with social supportFootnote 30		Process, implementation and service outcomes (n = 12) AcceptabilityFootnote 48 AdoptionFootnote 53 Community link evaluationFootnote 48 ImplementationFootnote 53 Goal achievementFootnote 44 Program maintenanceFootnote 53 ReachFootnote 53 Referral source (i.e. who made the referral)Footnote 30Footnote 32Footnote 34Footnote 35Footnote 39Footnote 41Footnote 44Footnote 45Footnote 52Footnote 57 Referral reasonFootnote 34Footnote 35Footnote 36Footnote 37Footnote 42Footnote 43Footnote 49Footnote 53Footnote 55Footnote 56Footnote 60 Referral type (new or re-referral)Footnote 34Footnote 35Footnote 37 Program engagementFootnote 34Footnote 35Footnote 50Footnote 52Footnote 57 Uptake of referralFootnote 34Footnote 35Footnote 49Footnote 50Footnote 56
Resource use	Economic (n = 6) Environmental impactFootnote 59 GP consultationsFootnote 39Footnote 47Footnote 48Footnote 49Footnote 52Footnote 58Footnote 59 GP home visitsFootnote 49 GP phone callsFootnote 39Footnote 49 Health care utilizationFootnote 30 Social return on investmentFootnote 41Footnote 56Footnote 58		Hospital (n = 2) Emergency department visitsFootnote 44Footnote 52 HospitalizationsFootnote 31	Need for further intervention (n = 7) Community physiotherapy, occupational therapy, nursingFootnote 44 Community referralsFootnote 44Footnote 48 Complex referralsFootnote 44 GP referralsFootnote 52Footnote 59 Health and social care servicesFootnote 56 Medication prescriptionsFootnote 48Footnote 49Footnote 52Footnote 59 Mental health referralFootnote 48	Societal/carer burden (n = 1) Social supportFootnote 40
Adverse events	Adverse events (n = 1) Unintended outcomesFootnote 56

Table 4. List of instruments used in the primary studies with a description of the instrument and measurement properties, when available, for a general adult population

Instrument	Description	Measurement properties
6 Item Cognitive Impairment Test (6CIT)Footnote 78	Brief cognitive screening tool	“[C]orrelates well with the [Mini-Mental State Examination] MMSE and is thus a valid measure of cognitive impairment. The 6CIT is especially useful in the identification of milder dementia; although its sensitivity is only approximately 80%, this is considerably better than that of the MMSE, whose sensitivity ranges from 50% to 65%, depending on cutoff.”Footnote 78,p.938-9
Burden of MultimorbidityFootnote 79	Measures disease burden based on self-reported chronic conditionsFootnote 79 “… respondents selected from a list of 21 common chronic diseases and additionally were allowed to add medical conditions not already on the list. They rated each condition on a five-point scale from 1 (interferes with daily activities “not at all”) to 5 (interferes with daily activities “a lot”). The total score representing level of morbidity was thus the sum of conditions weighted by the level of interference assigned to each.”Footnote 79,p.2-3	“[V]alidation against medical records revealed that median sensitivity relative to a ‘gold standard’ of chart review was 75% (range Footnote 35%–100%) and median specificity was 92% (range 61%–100%) …”Footnote 79,p.3
Camberwell Assessment of Need Short Appraisal Schedule Patient-rated (CANSAS-P)Footnote 80	“[I]nstrument for the assessment of needs in people with severe mental health problems”Footnote 80,p.114	“The CANSAS-P exhibited comparable detection of needs with its predecessor, better identification of domains that are problematic for patients to respond to, good test–retest reliability, especially for unmet needs, and generally positive evaluations by patients.”Footnote 80,p.113
Campaign to End Loneliness Measurement ToolFootnote 81	Consists of three questions framed from a positive perspective, focussed on friendships and relationships	“A reliability analysis indicated excellent internal consistency of the scale with a Cronbach’s alpha = 0.901.” Footnote 81,p.3Footnote 82
Center for Epidemiological Studies Depression Scale (CES-D-7)Footnote 83	This 20-item instrument asks questions to rate symptoms of depression (related to sleep, appetite, loneliness) experienced in the past weekFootnote 84	“The internal consistency, test–retest reliability and validity were high for all sex and age groups…”Footnote 84,p.283
Client Satisfaction QuestionnaireFootnote 85	An 8-item questionnaire on people’s satisfaction with program/service received	“It possesses a high degree of internal consistency and correlates with therapists’ estimates of client satisfaction.”Footnote 85,p.204 “In that study, the scale’s coefficient alpha was 90 when used with clients after their first service contact (n = 213), and 94 when used in a 90-day follow-up with the same clients (n = 113). On the other hand, the CSQ has been found to bear only moderate to low relationships with measures of outcome after a very brief or short-term follow-up interval.”Footnote 85,p.205
Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM)Footnote 86	This is a Footnote 34-item instrument of subjective well-being, symptoms, function and riskFootnote 86	“Internal and test–retest reliability were good (0.75–0.95), as was convergent validity with seven other instruments, with large differences between clinical and nonclinical samples and good sensitivity to change.”Footnote 86,p.51
EuroQol-5 Dimensions (EQ-5D)Footnote 87	This instrument consists of two components: (1) participants are asked to rate (scale 1–5) five domains of health (mobility, self-care, usual activities, pain/discomfort and anxiety/depression); and (2) a visual analogue scale: 0 (worst)–100 (best) for self-rated health	“The EQ-5D-5L is a reliable and valid generic instrument that describes health status which can be applied to a broad range of populations and settings. The assessment of responsiveness, in particular, needs further and more rigorous exploration. Rather large ceilings persist in general population samples, reflecting the conceptualization of the EQ-5D instrument, which focusses on limitations in function and symptoms, and does not include positive aspects of health such as energy or well-being.”Footnote 88,p.668
General Anxiety Disorder-7 (GAD-7)Footnote 89	A 7-item self-report scale to identify general anxietyFootnote 89	“The internal consistency of the GAD-7 was excellent (Cronbach α = .92). Test–retest reliability was also good (intraclass correlation = 0.83)”Footnote 89,p.1094; “There was a strong association between increasing GAD-7 severity scores and worsening function on all 6 SF-20 scales.”Footnote 89,p.1094 “The GAD-7 may be particularly useful in assessing symptom severity and monitoring change across time, although its responsiveness to change remains to be tested in treatment studies.”Footnote 89,p.1095
General Health Questionnaire-12 (GHQ-12)Footnote 90	The 12-item GHQ-12 screens for general (nonpsychotic) mental health problems among primary care patientsFootnote 90	“Reliability: The Cronbach’s alpha of the GHQ-12 for bimodal scoring (0-0-1-1) was 0.76, indicating satisfactory internal consistency.”Footnote 91,p.7 “Validity coefficients for the GHQ-12 … were generally high, with the mean area under the ROC curves being 0 ± 88, with a fairly narrow range …”Footnote 90,p.194
General Practitioner Physical Activity Questionnaire (GPPAQ)Footnote 92	Four questions about weekly physical activity completed in the previous year related to work, exercise and vigorous exercise, and number of stairs climbed	“The repeatability of the physical activity index was high (weighted kappa = 0.6, p < 0.0001). There were positive associations between the physical activity index from the questionnaire and the objective measures of the ratio of daytime energy expenditure to resting metabolic rate (p = 0.003) and cardiorespiratory fitness (p = 0.001).”Footnote 92,p.407
General Self-Efficacy Scale (GSE)Footnote 93	A 10-item scale to measure “the strength of an individual’s belief in his or her own ability to respond to novel or difficult situations and to deal with any associated obstacles or setback”Footnote 93,p.35	“High internal consistency ratings … ranged from .82 to .93”Footnote 93,p.35; “concurrent validity … [was] found with self-esteem (.52).”Footnote 93,p.36
General Self-Efficacy Scale–Korean VersionFootnote 94		“Cronbach’s alpha for the entire GSE scale Korean version is 0.81.”Footnote 94,p.844 “Significantly, the EFA revealed two factors, coping and confidence; however, Cronbach’s alpha for the confidence subscale was 0.54, indicating that it was not reliable. Therefore, it is advised to use either the full GSE scale–Korean version or the subscale coping alone.”Footnote 94,p.851
Geriatric Depression Scale Korean Version (GDS-K)Footnote 62	A 30-item screening test for depression for older people in community and clinical settings	“Cronbach’s coefficient alpha for the GDS-KR was 0.90 and the test–retest reliability was 0.91 (p < 0.01).”Footnote 62,p.232
Hospital Anxiety and Depression Scale (HADS)Footnote 95	A 16-item (8 items for anxiety and 8 items for depression) self-assessment scale for medical outpatient clinics	“Cronbach’s alpha for HADS-A varied from .68 to .93 (mean .83) and for HADS-D from .67 to .90 (mean .82).”Footnote 96,p.69 “Correlations between HADS and other commonly used questionnaires were in the range .49 to .83.”Footnote 96,p.69
Investigating Choice Experiments for the Preferences of Older People Capability Measure (ICE-CAP A)Footnote 97	Measures capability that “is primarily concerned with the evaluation of individual advantage based on a person’s ability to achieve ‘functionings’ in life that are valuable to them.”Footnote 98,p.2 Measures five areas: “stability (‘settled and secure’), attachment (‘love, friendship and support’), autonomy (‘independent’), achievement (‘achieve and progress’) and enjoyment (‘enjoyment and pleasure’)”Footnote 98,p.3	“The reliability of the capability questions, which takes into account higher level of inherent variability, is in the range of 0.52 (autonomy) to 0.61 (stability). The reliability of the health status questions is somewhat higher, in the range of 0.60 (usual activities) to 0.79 (mobility).”Footnote 99,p.627 “We found that the reliability of a simple measure of adult capability (the ICECAP-A) was slightly lower than that for a commonly used health functioning measure (the EQ-5D-3L) but not obviously affected by age, sex or education”Footnote 99,p.628
International Physical Activity Questionnaire (IPAQ)Footnote 100	Self-report measure of physical activity; “The objective was to develop a self-reported measure of physical activity suitable for assessing population levels of physical activity across countries.”Footnote 100,p.1381	“Overall, the IPAQ questionnaires produced repeatable data (Spearman’s rho clustered around 0.8), with comparable data from short and long forms. Criterion validity had a median rho of about 0.30, which was comparable to most other self-report validation studies.”Footnote 100,p.1381
Kessler Psychological Distress Scale (K10)Footnote 101	A 10-item screening instrument for nonspecific psychological distress	“There is a significant association between scores on the K10 and scores on the GHQ and SF-12, measures of symptoms and disability respectively, and between the K10 and the number of consultations for a mental problem in the previous 12 months. These findings support the validity of the K10 as a measure of psychological distress.”Footnote 101,p.496-7
Korean Mini-Mental State Examination (K-MMSE)Footnote 102	Translation of the Mini-Mental State Examination, a 30-item screening tool for cognitive impairment	“The sensitivity of the K-MMSE was 48.7%, with a specificity of 89.9%. The incidence of false positive and negative results totalled 10.1% and 51.2%, respectively. In addition, the positive predictive value of the K-MMSE was 87.1%, while the negative predictive value was 55.6%.”Footnote 102,p.177
Measure Yourself Medical Outcome Profile 2 (MYMOP2)Footnote 103	A short, generic, patient-specific measure of health. People are asked to state and rate one or two symptoms and “one activity of daily living, which they have come for help with, and which they consider to be the most important in affecting their lives.”Footnote 103,p.28	“The responsiveness index relating to minimal clinically important change was greater, for all MYMOP scales except well-being, than the level of 0.8 nominated as ‘high’ by previous work … [the] well-being scale was less responsive, but practitioners reported that it was clinically useful, especially in chronic disease.Footnote 103,p.1018 “MYMOP’s validity was supported by its ability to detect different degrees of change in relation to change scores and in acute and chronic conditions, and by its correlations with SF-Footnote 36 scores. Although the issue of clinical usefulness was clouded by follow-up being postal and not related to clinical follow-up, interviews provided important information on the effect of using the instrument in the consultation.”Footnote 103,p.1018
Museum Well-being Measure for Older Adults (MWM-OA)Footnote 104Footnote 105	“[M]useum-focussed measure to evaluate the well-being benefits of museum and gallery events and activities”Footnote 104,p.44	Internal consistency reliability was constructed for positive (α = 0.81) and negative emotion (α = 0.82). “Comparison of pretest-posttest differences showed highly significant increases in positive scores and decreases in negative scores with medium to large effect sizes.”Footnote 105,p.29
Patient Activation Measure (PAM)Footnote 106	A 13- or 22-item measure of activation, focussed on a person’s knowledge, skills and belief in their ability to self-manage health/health care	“Assessments of the 22-item PAM using national sample data show a high level of reliability with infit values ranging from .71 to 1.44. All but one of the outfit statistics are between .80 and 1.34.”Footnote 106,p.1020 “The results indicate considerable evidence for the construct validity of the PAM. Those with higher activation report significantly better health as measured by the SF 8 (r = .38, p < 0.001), and have significantly lower rates of doctor office visits, emergency room visits and hospital nights (r = −.07, p < 0.01).”Footnote 106,p.1021
Patient Health Questionnaire-9 items (PHQ-9)Footnote 107	A 9-item depression questionnaire from the full PHQ for “making criteria-based diagnoses of depressive and other mental disorders commonly encountered in primary care”Footnote 107,p.606	“The internal reliability of the PHQ-9 was excellent, with a Cronbach’s α of 0.89 in the PHQ Primary Care Study and 0.86 in the PHQ OB-GYN Study.”Footnote 107,p.608 Test–retest reliability of the PHQ-9 was also excellent. “PHQ-9 score ≥10 had a sensitivity of 88% and a specificity of 88% for major depression.”Footnote 107,p.606
Patient Health Questionnaire-8 items (PHQ-8)Footnote 108	This is the PHQ-9 questionnaire without the last question.	“Correlation between PHQ-9 and PHQ-8 scores was r = 0.997. Sensitivity and specificity for the PHQ-9 (54%, 90%) and PHQ-8 (50%, 91%) to detect major depression were similar.”Footnote 109,p.163
Rockwood Clinical Frailty Scale (CFS)Footnote 110	Instrument for clinicians to observe patients (using clinical judgment) “that evaluates specific domains including comorbidity, function and cognition to generate a frailty score ranging from 1 (very fit) to 9 (terminally ill).”Footnote 110,p.1	“Reliability of retrospective with prospective CFS scores was high (κ [kappa] = 0.89).”Footnote 111,p.1009 “Precision and interrater reliability of the comparison of retrospective CFS scores were high with narrow interquartile ranges and κ = 0.85.”Footnote 111,p.1009
Rosenberg Self-Esteem Scale (RSES)Footnote 112	A 10-item scale to measure two facets of self-esteem, self-competence and self-liking (using both positive and negative statements).Footnote 112	“Internal consistency reliability: Cronbach coefficient a was .91 for the overall sample and ranged from .84 (66+ age group) to .95 (unemployed working group), with an average of .90.Footnote 112,p.72
Rosenberg Self-Esteem Scale–Korean VersionFootnote 113		“The analysis showed that item eight was a misfit, and items three, nine, and ten were challenging to answer. It was found that, as the response range, a 4-point Likert scale was appropriate, and most of the items between the old and young adults functioned differently. As a result, RSES is not suitable for use for the older Korean adults because it did not satisfy the construct validity.”Footnote 113,p28984
Short Mood ScaleFootnote 114	A 6-item scale that measures changes in mood over the day, with a specific focus on calmness, valence and energetic arousal	“The results suggest that the proposed three factors calmness, valence, and energetic arousal are appropriate to assess fluctuations within persons over time. However, calmness and valence are not distinguishable at the between-person level. Furthermore, the analyses showed that two-item scales provide measures that are reliable at the different levels and highly sensitive to change.”Footnote 114,p.258
Social Participation ScaleFootnote 115	“… eight items evaluated on a five-point scale to measure the level of older adults’ social participation in a variety of interpersonal relationships”Footnote 33,p.5	The Cronbach alpha of the instrument was 0.85.Footnote 33
Social SupportFootnote 116	A 10-item questionnaire to assess perceived social supportFootnote 116	The Cronbach alpha of the instrument was 0.81.Footnote 116
South West Well-being Questionnaire (SWWBQ)Footnote 57	“Designed to be used for before-and-after evaluation, the South West Well-being Questionnaire … consisted of sets of validated and original measures that covered general health, social well-being, personal well-being, mental ill health, healthy eating and physical activity.”Footnote 56,p.1952	“Post analysis of the study data indicated acceptable internal consistency for the scales: 1. CES-D-7 baseline Cronbach’s α 0.853; follow-up Cronbach’s α 0.715 2. WEMWBS-7 baseline Cronbach’s α 0.885; follow-up Cronbach’s α 0.849 3. SWB-6 baseline Cronbach’s α of 0.714; follow-up Cronbach’s α of 0.708.”Footnote 56,p.1953
UCLA Loneliness Scale (ULS-20)Footnote 117	A 20-item measure of subjective loneliness and isolation	“The measure has high internal consistency (coefficient alpha = .96) and a test–retest correlation over a two-month period of .73. Concurrent and preliminary construct validity are indicated by correlations with self-reports of current loneliness and related emotional states, and by volunteering for a ‘loneliness clinic.’”Footnote 117,p.290
ULS-8 (Short-form UCLA Loneliness Scale)Footnote 118	An 8-item version of the UCLA Loneliness Scale	“Internal consistency reliability was high … Cronbach’s (1951) alpha = 0.84”Footnote 118,p.74 “An eight-item short-form of loneliness, ULS-8, was selected on the basis of results from exploratory factor analysis, and it was found to be highly correlated with the ULS-20 (r = .91).”Footnote 118,p.77
Warwick–Edinburgh Mental Well-being Scale (WEMWBS)Footnote 63	A 14-item measure of well-being in the general population “capturing affective-emotional aspects, cognitive-evaluative dimensions, and psychological functioning” of well-beingFootnote 63,p.2	“A Cronbach’s alpha score of 0.89 (student sample) and 0.91 (population sample) suggests some item redundancy in the scale. WEMWBS showed high correlations with other mental health and well-being scales and lower correlations with scales measuring overall health… Test–retest reliability at one week was high (0.83).”Footnote 63,p.1
Short Warwick–Edinburgh Mental Well-being Scale (SWEMWBS)Footnote 119	A 7-item version of the WEMWBS, to measure mental well-being with the focus on functionFootnote 119	“In this exploratory analysis, SWEMWBS demonstrates construct validity and sensitivity to change as a clinical outcome measure for patients with common mental disorders (CMDs) in primary care, demonstrating inverse correlation and comparable sensitivity to change over a course of clinical treatment when compared to two widely used clinical outcome measures.”Footnote 119,p.7
Well-being StarFootnote 120	A 7-domain scale, each scoring 1–10, “with a long-term health condition, to support and measure their progress in living as well as they can”Footnote 121,p.1	“Internal consistency: internal consistency was good (Cronbach’s α = .82).”Footnote 121,p.1 “Responsiveness to change: Wilcoxen Signed Rank Test revealed a statistically significant increase in all outcome areas …  A large effect size was found for the Lifestyle area, medium effect sizes for Looking after yourself, Managing symptoms, Work, volunteering and other activities and Feeling positive. The effect sizes for the other three areas were small–medium.”Footnote 122,p.2
World Health Organization Quality of Life–abbreviated (WHOQOL-BREF)Footnote 123	A quality-of-life and general health measure, contains 26 questions, designed to be used cross-culturally	“Domain scores produced by the WHOQOL-BREF correlate highly (0.89 or above) with WHOQOL-100 domain scores (calculated on a four-domain structure). WHOQOL-BREF domain scores demonstrated good discriminant validity, content validity, internal consistency and test–retest reliability.”Footnote 123,p.551
Work and Social Adjustment Scale (WSAS)Footnote 124	A 5-item self-report scale of ability to function (work, home, etc.) related to an identified problem (like the impact of mental health).	“Cronbach’s α measure of internal scale consistency ranged from 0.70 to 0.94. Test–retest correlation was 0.73. Interactive voice response administrations of the WSAS gave correlations of 0.81 and 0.86 with clinician interviews. Correlations of WSAS with severity of depression and obsessive-compulsive disorder symptoms were 0.76 and 0.61, respectively. The scores were sensitive to patient differences in disorder severity and treatment-related change.”Footnote 124,p.461

When comparing extracted outcomes (Table 3) with the NHS Social Prescribing Common Outcomes Framework,^{Footnote 14} most categories were included by one or more primary studies. We observed that family caregivers, volunteers and voluntary groups were mentioned in many primary studies.^{Footnote 32Footnote 33Footnote 38Footnote 41Footnote 52Footnote 54Footnote 56Footnote 57Footnote 61} Two primary studies^{Footnote 54Footnote 56} included family caregivers within data collection, but we did not locate specific measures to evaluate volunteers’ experience or information related to societal or cost outcomes, despite the inclusion in the NHS framework. It is notable only six primary studies^{Footnote 46Footnote 50Footnote 52Footnote 53Footnote 57Footnote 58} reported on physical activity (also mentioned in the NHS framework), and these studies were all published before 2020. However, the NHS Social Prescribing Common Outcomes Framework^{Footnote 14} was dated June 2020 and pertains to the health and social care system in the UK. Therefore, not all included primary studies may have followed the framework (as a guideline) because of where the study was located (e.g. outside of the UK) and when it was published, as only 13 (39%) studies were published in 2020 or later.

Demographic information

Reporting varied for information provided in primary studies (n = 33). All primary studies but two^{Footnote 59Footnote 61} reported participants’ age and gender/sex. Thirteen primary studies collected information on ethnicity/culture/language/race;^{Footnote 7Footnote 30Footnote 39Footnote 41Footnote 42Footnote 43Footnote 47Footnote 48Footnote 52Footnote 53Footnote 54Footnote 57Footnote 58} 12 primary studies reported on occupation/employment;^{Footnote 30Footnote 35Footnote 36Footnote 37Footnote 39Footnote 42Footnote 46Footnote 48Footnote 50Footnote 52Footnote 58Footnote 60} seven primary studies reported on socioeconomic status (based on place of residence);^{Footnote 34Footnote 35Footnote 36Footnote 37Footnote 50Footnote 53Footnote 54} four primary studies reported on education;^{Footnote 33Footnote 40Footnote 52Footnote 58} and four primary studies mentioned disability.^{Footnote 30Footnote 51Footnote 57Footnote 58}

Discussion

We provide an overview of the outcomes and instruments used in peer-reviewed literature for social prescribing with adults aged 18 years and older, and highlight variability in the type and number of instruments used within studies. Our findings support and extend previous work^{Footnote 11Footnote 12} by comparing identified outcomes with a published taxonomy.^{Footnote 13} Many instruments were PROMs, alongside some PREMs, with less emphasis on physical activity, cognition or people providing unpaid care, such as family members and volunteers; only one study reported on unintended outcomes (e.g. harms). We further identified, similar to the recent mapping review,^{Footnote 12} missing information related to equity: most studies only provided basic participant descriptive information, yet these data inform the development of equity within program development and delivery. Further, this evidence synthesis is a reminder for more comprehensive reporting of outcomes, given the recent development of reporting guidelines.^{Footnote 20} Overall, this synthesis could be used to guide future research studies and program evaluations. It will be used to support our next phase—to conduct a modified Delphi study to determine outcomes important for people who impact or are impacted by social prescribing.^{Footnote 15}

General interpretation

Almost all taxonomy core areas and domains were included across social prescribing primary studies, but there was a strong emphasis on well-being and mental health. Noteworthy is the lack of studies measuring physical and daily activity to quantify any behaviour change associated with participation resulting from the prescription. However, it is possible, based on our inclusion criteria, that we excluded studies focussed on exercise and physical activity but that did not call their program “social prescribing.” There were included studies that focussed on other activities (e.g. arts, museum, connection) which could result in incidental physical activity. However, the studies did not routinely capture “what” people are doing within programs (e.g. incidental physical activity). Arts- or museum-based programs are not necessarily promoted as physical activity, but they are opportunities to add movement into everyday life.

Another identified gap was the measurement or collection of information on people’s cognition. Although more than 20 of the primary studies reported populations as middle-aged or older (40+ years), including nine studies of people aged over 60 years, only two studies reported using an instrument to directly evaluate the effect or impact of social prescribing on cognition.^{Footnote 33Footnote 56} There is an interplay between mental health (e.g. depression, anxiety), physical health and cognition. Depression in later life can increase the risk of dementia^{Footnote 64} or frailty.^{Footnote 65} Conversely, life satisfaction may prevent a decline in some measures of cognition.^{Footnote 66} In 2019, there were approximately 57 million people (all ages) living with dementia globally;^{Footnote 67} this number is projected to rise to 152.8 million people by 2050.^{Footnote 67} Cognitive function and unmet social needs may be related, possibly due to the consequences of negative environmental factors, such as food insecurity or fewer neighbourhood resources.^{Footnote 68} Further, a decline in cognition can challenge functional independence and completion of simple daily tasks.

These factors, at a minimum, should be considered when co-creating an action plan to address unmet needs. In an ideal situation, it would be beneficial to see if social participation (as an aim of social prescribing) could mitigate the risks for cognitive decline. Some social prescribing activities, such as arts- and museum-based programs, may encourage concurrent cognitive and physical activities, which when combined in other research has been effective at promoting cognitive health.^{Footnote 69} Taken together, better understanding a person’s cognition could assist when connecting them with a link worker, introducing community resources and services, and determining the effectiveness of the social prescription.

The NHS Social Prescribing Common Outcomes Framework highlights collecting information on volunteers,^{Footnote 14} but there were fewer evaluations of the impact or effect of family caregivers, volunteers and volunteer organizations,^{Footnote 41Footnote 54Footnote 56} despite their being mentioned in many primary studies.^{Footnote 32Footnote 38Footnote 41Footnote 52Footnote 54Footnote 56Footnote 57Footnote 61} Despite the important contribution of volunteers personally and economically,^{Footnote 70} there are fewer published studies for volunteering and social prescribing; when they are available, they are focussed on volunteering as a “prescription.”^{Footnote 71Footnote 72}

There were also few mentions of family caregivers in primary studies,^{Footnote 54Footnote 56} who play an essential role in providing supportive care. At a personal and societal level, the costs of family caregiving are high. There is the likelihood of caregivers experiencing high risk of physical and mental health challenges resulting in reduced quality of life.^{Footnote 73} Further, in 2017, a report from the Canadian Imperial Bank of Commerce (CIBC) estimated caregiving costs Canadians CAD 33 billion annually for direct and indirect costs, such as out of pocket expenses (including paying for other providers) and time away from work; most affected are women and people with lower income.^{Footnote 74} Given the benefits of volunteers and family caregivers at a personal and societal level, it is important to support, tailor, track and evaluate this important contribution to health and social models of care.

Social prescribing has a central theme of connection, for people with unmet social needs, family members, volunteers, providers and community organizations. The creation and sustainment of relationships between and across interested parties depends on effective communication and trust,^{Footnote 22} among other factors. There are relational strategies and techniques to build intra- and interpersonal trust of people, providers and implementation teams,^{Footnote 75} which can be used to generate effective changes in the adoption and sustainability of programs or clinical practices.^{Footnote 76} We did not locate outcomes on relationships, but may have missed these data because the information is available in the unpublished literature or in studies using different methods. Future research could consider measurement of the development, strength and sustainability of relationships for people receiving social prescribing, and for people who deliver, manage and make decisions for its delivery across the continuum of care and sectors.

Strengths and limitations

This work has many strengths to contribute to the science and practice of social prescribing. Despite the comprehensive approach in this synthesis, we recognize several limitations. First, we made the decision to include only peer-reviewed studies that described their program as “social prescribing.” Social prescribing is a relatively new care model, but similar programs have existed for decades. However, using this criterion means we excluded studies that align with this model but do not call themselves social prescribing. Conversely, it could also mean we included studies that called their program social prescribing when it may not have been as closely aligned to the definition that is now published.^{Footnote 3} In the recent social prescribing mapping review, the authors noted the challenges with screening studies to determine if the intervention was social prescribing.^{Footnote 12} In our previous reviews^{Footnote 8Footnote 9} we had a similar experience, and thus decided to only include studies described as social prescribing.

Second, we only included peer-reviewed studies when searching for outcomes. We made this decision because many systematic reviews noted data were missing across studies (Table 1), and we wanted to compare reporting in the peer-reviewed primary studies, which may be more likely to follow research reporting guidelines. We acknowledge this means we may have missed other outcomes, especially as there are many social prescribing studies published in the grey literature.

Third, we only included outcomes that were captured using quantitative strategies, and we may have missed information that was obtained via interviews and focus groups. Despite the important and rich data obtained through these methods, our findings may not have changed substantially, as studies in the current review included outcomes from almost all taxonomy domains. Nonetheless, concepts such as social connectedness may be better explored through qualitative methods, to better understand the effect of a complex intervention such as social prescribing. Our work highlights what is or could be measured—it does not limit how the outcome or domain should be measured.

Fourth, our work is only descriptive and does not provide any information as to which outcomes should or should not be included in evaluating social prescribing interventions. This was intentional, because determining the scope and priorities of future evaluation should be a collaborative process based on needs, preferences and supporting information, which together with interested parties (such as people, families, providers and decision makers) can be used to advance the science and practice of social prescribing.

Conclusion

We recognize it is impossible to measure everything in one study, but a core set of outcomes would benefit the field. Although the NHS has already provided outcomes to include in social prescribing evaluations,^{Footnote 14} there remains the need to expand the list, standardize what and how we measure outcomes, and provide more information when describing people and processes for social prescribing. Specifically, consideration should be given to equity-considered guidelines such as PROGRESS-Plus^{Footnote 21} to describe communities and people receiving and delivering social prescribing. It is also important to provide more information on unintended outcomes and the rationale and instrument measurement properties (reliability, responsiveness and validity, at a minimum).^{Footnote 77} As there is now an international definition of social prescribing,^{Footnote 3} it is important to use it to guide interventions and how they align (or do not). The current work is intended to prompt interest and action in the continued development of the science and practice underpinning social prescribing.

Acknowledgements

Professor Ashe gratefully acknowledges the support of the Canada Research Chairs Program. Dr. Chudyk acknowledges the support from the Canadian Institutes of Health Research through the Patient-Oriented Research Awards – Transition to Leadership Stream, Phase 2 Award. Dr. Esfandiari acknowledges the postdoctoral fellowship provided by the Edwin S.H. Leong Centre for Healthy Aging.

Funding

We gratefully acknowledge funding for this project from The University of British Columbia Health Innovation Funding Investment Award and Social Sciences and Humanities Research Council.

Conflicts of interest

The authors declare no conflicts of interest.

Authors’ contributions and statement

• MCA, EE—conceptualization.
• MCA, IKS, HA, AC, EE—methodology, formal analysis.
• MCA, IKS—writing—original draft.
• MCA, IKS, HA, AC, EE—writing—review and editing.
• MCA, IKS—visualization.

All authors read and agreed to the published version of the manuscript.

The content and views expressed in this article are those of the authors and do not necessarily reflect those of the Government of Canada.

Previous | Table of Contents | Next