In-person interventions to reduce social isolation and loneliness: An evidence and gap map

The EGM in brief

This report discusses the effectiveness of in-person interventions to reduce social isolation and loneliness in people across all ages. It emphasizes that the available evidence is not consistent and varies in support of interpersonal and community-based delivery interventions and health-related outcomes (loneliness, wellbeing, depression/anxiety), with most evidence coming from high-income countries.

What is this EGM about?

Social isolation and loneliness are becoming more significant public health issues associated with increased deaths and poorer health outcomes.

In-person interventions can help people who are at risk of feeling socially isolated and lonely by teaching them how to maintain their existing connections or form new ones. Policy makers are interested in knowing which interventions are effective in reducing social isolation and loneliness.

What is the aim of this EGM?

To present available evidence from systematic reviews and primary studies about the effects of in-person interventions to reduce social isolation and/or loneliness across all ages and settings.

What studies are included?

The EGM included 513 articles comprising 421 primary studies and 92 systematic reviews that evaluate the effectiveness of in-person interventions for reducing social isolation and/or loneliness. The studies had to report on the effect of in-person interventions and could be conducted in any setting.

What are the main findings of this gap map?

The evidence in this EGM varies with more evidence for interpersonal and community-based delivery interventions than interventions delivered at societal levels. Most measured outcomes include health-related outcomes of loneliness, well-being, and depression/anxiety. Specifically, there are evidence clusters for community-based interventions of group activities and interpersonal delivery interventions for changing negative thinking and feelings about social relationships. Most evidence comes from high-income countries including the US, UK, and Australia, with none from low-income countries. Most systematic reviews included in this EGM are of critically low quality, with less than 5% of high or moderate quality.

Less than 5% of included reviews reported outcomes about how well the interventions were implemented. Only two primary studies and one review assessed interventions delivered at the societal level and only one review assessed interventions for age-friendly communities.

While all age groups were represented, more reviews and primary studies were interested in older adults (60 years or above, 63%) than young people (24 years or below, 35%). Very few described how they identified people who were at-risk and even fewer assessed whether the effects of the interventions were different for populations experiencing inequities.

What do the findings of the map mean?

This EGM is a step toward identifying effective in-person interventions for reducing social isolation and loneliness for all age groups. Though a large body of evidence exists, with much of this research from more recent years, it shows variations, with a lack of high quality reviews. The map contributes to building an evidence base in the field by which funders and researchers can identify gaps in the evidence and then address these as future research priorities.

How up-to-date is this EGM?

We searched for relevant studies up until 17 February 2022.

The problem, condition, or issue

Social isolation is the objective lack or paucity of social contact and infrequent interactions with others (Badcock et al., 2022; Donovan et al., 2020; Leigh-Hunt et al., 2017). Loneliness is a related concept defined as the subjective, negative feeling of inadequate meaningful connections resulting from an unmet need or discrepancy between desired and actual social relationships (Perlman et al., 1981; Prohaska et al., 2020). Loneliness has two components: an emotional component (unpleasant, negative feeling) and a social cognition component (perception of social disconnection from other people with a desire to be connected) (Badcock et al., 2022). Loneliness can also be a transient normal experience or chronic with negative physical and mental health consequences (Akhter-Khan et al., 2020; Qualter et al., 2015).

The two concepts, social isolation, and loneliness, are distinct; social isolation is objective and associated with social contact while loneliness is subjective and related to social connectedness (O'Rourke et al., 2018). One may occur without the other, although they are related and may also co-occur. People may have a large social network and feel lonely, while some with a small network may not.

Social isolation and loneliness can occur across all age groups and are associated with serious health consequences including anxiety and depression, cardiovascular disease, and premature mortality (Cené et al., 2022; Leigh-Hunt et al., 2017). One German cohort study of nearly 5000 middle-aged and older adults found that the risk of mortality is exacerbated when social isolation and loneliness co-occur (Beller et al., 2018). Poor relationships are associated with 32% increased risk of stroke, 29% increased risk of coronary heart disease (Cené et al., 2022; Valtorta et al., 2016), and 26% increased risk of all-cause mortality (Donovan et al., 2020; Holt-Lunstad et al., 2015). Incident depression and dementia have a bi-directional relationship with both social isolation and loneliness (Cené et al., 2022; Donovan et al., 2020) although several studies have reported that dementia is associated with loneliness than social isolation (Cené et al., 2022; Holwerda et al., 2014; Rafnsson et al., 2020). Both social isolation and loneliness are associated with negative health-related behaviors such as smoking and physical inactivity (Cené et al., 2022; Menec et al., 2020). The negative health impacts of social isolation and loneliness have been shown to increase health and social care service use (Cotterell et al., 2018; Windle et al., 2012). These negative impacts occur when contextual and risk factors affecting social relationships persist and individuals do not use appropriate coping strategies to address them (Akhter-Khan et al., 2022; Elder et al., 2012).

Since the onset of the COVID-19 pandemic, movement restriction policies have made social isolation and loneliness prominent global issues and a public health priority (Galvez-Hernandez et al., 2022; Holt-Lunstad 2022; WHO, 2021). The prevalence of severe loneliness increased by 15% and social isolation by 13% in adults 18 years or older across 101 countries during the pandemic (O'Sullivan et al., 2021). Small increases in the prevalence of loneliness were also observed in a recent synthesis of longitudinal studies during the COVID-19 pandemic (Ernst et al., 2022). The prevalence is hard to measure across the lifespan because of the lack of standardized measurement instruments and definitions, and the use of different cut-off points and age categories (Holt-Lunstad, 2022; Prohaska et al., 2020). A recent systematic review and meta-analysis on the prevalence of loneliness pre-COVID-19 pandemic across 113 countries (Surkalim et al., 2022) showed varying rates for adolescents (9.2% to 14.4%), young adults (1.8% to 9.4%), middle-aged adults (2.4% to 12%), and older adults (4.2% to 24.2%) depending on the country. A prevalence study in a German population-based adult cohort showed that social isolation increases with age from 5.4% (95% confidence interval [CI]: 4.7 to 6.0) in the youngest age group (18 to 39 years) to 21.7% (95% CI: 19.5 to 24.0) in the oldest age group (70 to 79 years) (Röhr et al., 2021). The global prevalence of social isolation in community-dwelling older adults was found to be 25% (95% CI: 21 to 30) (Teo et al., 2022). Studies have found a u-shaped association between age and loneliness with the highest levels of loneliness in younger and older people (Lasgaard et al., 2016; Yang et al., 2011), but few studies have focused on younger people. Although social isolation and loneliness are global issues, most of the studies were conducted in high income countries, especially in Europe, with very few in low-middle-income countries (Fakoya et al., 2020; Surkalim et al., 2022).

Both social isolation and loneliness are linked to less social support and can be triggered by situational factors such as adversity, significant life changes or transitions, such as moving away from home, starting a new job, becoming a parent, illness, and the death of a spouse or parent (Badcock et al., 2022; Elder et al., 2012; Lim et al., 2020; Qualter et al., 2022). They are associated with risk factors including individual factors (e.g., personality, maladaptive cognition, poor health, disability or mobility impairment, cognitive impairment), interpersonal or social factors (e.g., peer victimization or discrimination, poor relationship quality, quantity of friends or social contacts, living alone), socio-environmental factors (e.g., neighborhood deprivation, inaccessible location of residence, housing, cultural prejudice), and demographic factors (e.g., age, gender, educational level, low socio-economic status, unemployment) (Badcock et al., 2022; Elder et al., 2012; Lim et al., 2020; O'Sullivan et al., 2021; Qualter et al., 2022).

Many systematic reviews have evaluated the effectiveness of interventions to reduce social isolation or loneliness with conflicting findings demonstrating a need for better quality research (Masi et al., 2011; Victor et al., 2018; Williams et al., 2021). A number have focused on older adults, but social isolation and loneliness affect people across the life span, including young people (Qualter et al., 2015; Surkalim et al., 2022), with interventions designed specifically for them (Eccles et al., 2021). Most of the reviews have focused on people living in the community or long-term care settings (Fakoya et al., 2020; Grenade et al., 2008). There is limited research addressing social isolation and/or loneliness for patients in clinical settings (NASEM, 2020). Studies that consider hospitalized patients focus on screening and detection of loneliness and social isolation, the impact of social isolation and loneliness on health service use and which interventions may be used; hardly any focus on assessing the effectiveness of interventions to reduce social isolation and loneliness (Grenade et al., 2008; NASEM, 2020; Proffitt et al., 1993; Razai et al., 2020; Zamir et al., 2018).

The impact of interventions has been found to differ depending on population characteristics such as coping skills, needs, degree of loneliness, and contextual factors like age, socioeconomic status, health condition, and place of residence (Fakoya et al., 2020). Therefore, there is no one-size-fits-all approach, and it is important to tailor appropriate interventions to individuals’ needs and contexts (Akhter-Khan et al., 2020; Fakoya et al., 2020; Mann et al., 2017).

There are health equity issues related to social isolation and/or loneliness such as the gap in evidence from low-middle income countries (Surkalim et al., 2022), limited access to interventions caused by disabilities and lack of transportation, or limited programs in rural areas compared to urban areas (Dassieu et al., 2021; NASEM, 2020; Qualter et al., 2022). Social isolation and/or loneliness related to structural inequities (e.g., intersectional discrimination across race, gender, socioeconomic status; age-based discrimination and ethnic minorities), have a negative impact on health outcomes (Dassieu et al., 2021).

This current evidence and gap map identifies areas where evidence is available, as well as any gaps in research related to in-person interventions for social isolation and loneliness across any age.

The intervention

Different types of interventions for reducing social isolation (Dickens et al., 2011; Findlay, 2003), loneliness (Cohen-Mansfield et al., 2015; Eccles et al., 2021; Hagan et al., 2014; Jarvis et al., 2020; Mann et al., 2017; Masi et al., 2011; Veronese et al., 2021), or both social isolation and loneliness (Cattan et al., 2005; Gardiner et al., 2018; Poscia et al., 2018) have been described and assessed in several systematic and scoping reviews. However, there is a lack of a standardized framework for describing these interventions (Fakoya et al., 2020; Prohaska et al., 2020). Interventions for reducing social isolation and loneliness are often complex with multiple and interacting components, working through different potential mechanisms of action (Fakoya et al., 2020; Gardiner et al., 2018).

Several approaches have been used to categorize interventions in some reviews. The interventions have been categorized by the format or delivery mode or type as one-on-one or group-based (Cohen-Mansfield et al., 2015; Dickens et al., 2011; Fakoya et al., 2020; Findlay, 2003; Hagan et al., 2014; Masi et al., 2011; Poscia et al., 2018), or technology or non-technology (in-person) based (Eccles et al., 2021; Masi et al., 2011). They have also been categorized by the type, or strategy, being classified as interventions for social skills training, enhancing social support, enhancing social interaction or social cognition training (Masi et al., 2011). Other terms have been used as a rationale for categorization, such as the focus, nature or goal of the intervention (Cohen-Mansfield et al., 2015; Fakoya et al., 2020; Masi et al., 2011); the purpose, intended outcomes, and mechanisms by which interventions target social isolation and loneliness (Fakoya et al., 2020; Gardiner et al., 2018).

The scoping review by Mann et al. classifies interventions as direct (i.e., targeting loneliness and related concepts) or indirect (targeting wellbeing which impacts loneliness), and also articulates various levels of engagement for those delivering the interventions following the socio-ecological model: individual level, relationship and community level, and societal level (Mann et al., 2017).

Another scoping review (O'Rourke et al., 2018) classified interventions for reducing loneliness by their components into nine types: one-to-one personal contact, activity group, animal contact, skills course interventions, or varied/non-specific interventions, reminiscence, support group, model of care and public broadcast.

This evidence and gap map focuses on in-person interventions that are non-technology based and delivered face-to-face since there is a gap map on digital interventions for older adults (Welch et al., 2022b).

Why it is important to develop the EGM

The existing body of evidence for interventions to mitigate social isolation and/or loneliness is rapidly growing and characterized by small, low-quality trials, with inconsistent terminology and conclusions on their effectiveness (Eccles et al., 2021; Fakoya et al., 2020; Prohaska et al., 2020; Veronese et al., 2021) as well as equity-related issues. With the rapid growth of evidence in this sector, this evidence and gap map will demonstrate areas where evidence is available and areas where there are gaps that researchers, decision and policymakers could use to help select interventions and prioritize future research to fill gaps in research and to understand which interventions might work in which conditions. It will also improve the discoverability of evidence on different types of interventions and enhance their use for informed decision-making by stakeholders including health and social care providers, policymakers, citizens, caregivers, and patients.

Existing evidence and gap maps and/or relevant systematic reviews

There is rapidly expanding research on alleviating social isolation and/or loneliness since the COVID-19 pandemic. Several relevant systematic reviews have been conducted and included in three umbrella reviews (Boulton et al., 2021; Jarvis et al., 2020; Veronese et al., 2021). Boulton et al. found mixed evidence of effectiveness on loneliness for remote befriending, social support, and low intensity psychosocial interventions. Jarvis assessed various interventions addressing loneliness in older adults and found limited effect, with the greatest effect in a social cognition intervention. Veronese et al. reported low or very low-quality evidence of three interventions (meditation/mindfulness, social cognitive training, and social support) that reduced loneliness. A scoping review of reviews showed the lack of studies conducted in low-middle income countries (Fakoya et al., 2020).

One evidence and gap map has been conducted on remotely delivered digital interventions including befriending, social support, and low-intensity psychosocial interventions for social isolation and loneliness in older adults (Boulton et al., 2021). It showed mostly low-quality reviews and few studies on older people who are not caregivers or who do not have a particular chronic illness. Our group is currently working on another gap map on digital interventions for older adults with a broader scope of interventions and including caregivers (Welch et al., 2022; Welch et al., 2022b), but there is currently no mapping of evidence for in-person interventions to reduce social isolation and loneliness across all ages.

Population targeted by interventions

Social isolation and loneliness are complex public health issues and their occurrence across the lifespan is influenced by individual contextual and risk factors, needs, expectations, and coping skills which are all inter-related and influence relationship ties (Akhter-Khan et al., 2022; Elder et al., 2012; Gardiner et al., 2018; O'Rourke et al., 2018; Qualter et al., 2022). Contextual and risk factors such as structural changes that may cause displacement (e.g., moving schools or wars), living situations (e.g., living alone or in a care facility such as orphanage, long-term care home), resources (available activities or social supports) affect people's motives, expectations, coping skills, and social relationships. Coping skills and social supports may be a protective factor if they allow people to promote their wellbeing or resilience. On the other hand, inadequate coping skills and social supports may be a risk factor for social isolation or loneliness.

Based on a public health approach, interventions may target anyone regardless of risk (universal), or target subpopulations at high risk (selective) or socially isolated or lonely people (indicated) (Springer et al., 2007). Categorizing target populations into these three orders gives a clearer picture and understanding of whom to prioritize and how to allocate resources more efficiently.

Risk factors

Social isolation or loneliness may be triggered in both young and old across the life span by situational factors such as significant life events or transitions (e.g., adversity, moving away from home, retirement, death of a spouse, friend, or relative) and may be associated with risk factors including physical and mental health factors (e.g., poor health, maladaptive cognition or cognitive decline, disability or impaired mobility, personality), interpersonal or social factors (e.g., living alone, peer victimization, social contacts, relationship quality), socio-cultural or environmental factors (e.g., neighborhood deprivation, inaccessible location of residence, cultural individualism, social discrimination, and stigma) and demographic factors (e.g., age, gender, socio-economic status) (Dahlberg et al., 2022; Elder et al., 2012; Lim et al., 2020; NASEM, 2020; Qualter et al., 2022).

Needs assessment

Socially isolated and lonely individuals across the lifespan may have unmet needs that can be due to low social support or disruption in social interactions with people at different levels—individual, group, community, or societal (Elder et al., 2012; Holt-Lunstad, 2022; NASEM, 2020; WHO, 2021). These needs may include basic needs (housing, nutrition and food security, and healthcare), social and emotional needs (social connections and companionship), financial, education (skills development and learning), communication, caregiver needs, home modifications and maintenance, domestic assistance, mobility, personal care, respite care and civic engagement (meaningfulness and status, the need for having a purpose in later life or being able to contribute usefully to society) (Abdi et al., 2019; Bedney et al., 2010; Bee et al., 2014; Eccles et al., 2021).

Social support is a major component of social connection and may be provided to meet different needs. Social support can take the forms of instrumental/tangible, informational, emotional, or belonging support (Elder et al., 2012; Holt-Lunstad, 2022; NASEM, 2020). It can be perceived, or actual support provided through social connections with other people and through different sectors including health, transportation, housing, work, nutrition, environment, education, leisure: arts and entertainment (Holt-Lunstad, 2022).

Loneliness is also associated with social relationship expectations that are influenced by personal, social, cultural, and historical contexts and include the availability of social contacts (proximity), feeling cared for and relying on others (support), feeling close, understood, and listened to (intimacy), sharing interests and enjoyable experiences (fun), having opportunities to contribute meaningfully (generativity) and feeling valued and actively included (respect) (Akhter-Khan et al., 2022). A discrepancy between expected and actual social relationships will result in loneliness.

It is important to assess and understand individuals’ specific contexts, risks, expectations, and needs to tailor appropriate interventions to reduce social isolation or loneliness (Akhter-Khan et al., 2022; Eccles et al., 2021; Fakoya et al., 2020; Jopling, 2020; Lim et al., 2020; Mann et al., 2017; NASEM, 2020).

Interventions

Interventions to reduce social isolation and/or loneliness are more effective when targeted to the individual's specific experience and context (such as triggers, risk factors, and accessibility to resources) (Eccles et al., 2021; Fakoya et al., 2020; Prohaska et al., 2020; Veronese et al., 2021). Interventions aimed at reducing social isolation may have different components than those aimed at reducing loneliness (O'Rourke et al., 2018). The interventions can be delivered through various modes to bring about changes at different levels—individual, relationship, community, and societal levels (Holt-Lunstad, 2022; Lim et al., 2020; Mann et al., 2017; Ogrin et al., 2021; WHO, 2021). In addition, one component of an intervention may target multiple factors (O'Rourke et al., 2018). It is therefore challenging to categorize interventions by the risk factors they target following the socio-ecological framework or by their mechanisms. In order to have mutually exclusive categories and subcategories on the evidence map, we classified interventions as follows, based on who is providing the intervention and where the intervention is provided.

Self-delivery: These are self-guided interventions that focus on addressing personal characteristics (biological and sociodemographic factors, e.g., socio-economic status) that may trigger social isolation or loneliness, through strategies that change the attitudes, beliefs, and behaviors of individuals. The purpose of these interventions is to affect changes in an individual that would result in reductions in isolation and loneliness. They can be facilitated with training, guidance, or available resources from the healthcare professionals. Examples include self-guided social cognitive interventions like cognitive behavioral therapy, mindfulness, or reminiscence therapy.

Interpersonal delivery: These are interventions that focus on building close meaningful personal relationships with family and others in the community, for example, friends, colleagues, neighbors, volunteers. They can be delivered by healthcare or social care professionals, volunteers, or other people in the individual's social network. The purpose/aim of these interventions is to affect changes in an individual or a specific relationship or network of relationships that would result in reductions in isolation and loneliness. They are accomplished through mechanisms including changing cognition, social skills training and psychoeducation, healthcare support, and social support. Examples include cognitive behavioral therapy, family therapy, social prescribing or friendship enrichment program.

Community-based delivery: These interventions are delivered by healthcare or social care professionals, community workers/volunteers, or other people within the same community setting. They can be delivered through neighborhood organizations, community-based healthcare, and social services or facilities. They focus on addressing risk factors in social settings and increasing opportunities for social interactions with others such as connecting to community group activities or peer support groups; neighborhood approaches like networks, transportation, meals on wheels; intergenerational approaches. They may also provide a supportive environment or encourage participation by improving access to amenities within the community e.g., built environment, age-friendly communities, and volunteering. These interventions can affect changes in an individual, a specific relationship or network of relationships, or changes in the community that would result in reductions in isolation and loneliness.

Societal level delivery: Focus on policies and laws that address societal risk factors like discrimination and stigma, socio-economic inequality or may seek to change social norms that prevent social connection such as policies addressing housing, employment, transportation and the environment. These interventions affect change(s) in broader society, resulting in reductions in isolation and loneliness. Examples include public awareness campaigns, coalition and partnership initiatives, or family-friendly policies.

Multi-component/complex: These are combinations of multiple components within the interventions involving the same/different types of delivery modes in the same study, for example, reminiscence therapy and group activities.

Mechanisms

Different pathways or mechanisms related to contextual or risk factors, motives, expectations, and coping skills have been identified through which interventions may reduce social isolation or loneliness. The interventions may improve and maintain existing relationships or enable people to create new connections (Akhter-Khan et al., 2022; Jopling, 2020; Mann et al., 2017) by addressing contextual or risk factors (Akhter-Khan et al., 2022; Lim et al., 2020; Ogrin et al., 2021). Some interventions aim to change one's outlook or negative social cognition (Akhter-Khan et al., 2022; Jopling, 2020; Mann et al., 2017; Masi et al., 2011; Ogrin et al., 2021) while others provide support to enhance social interactions (Mann et al., 2017; Masi et al., 2011; Ogrin et al., 2021). Some interventions involve building skills, purposeful activity, or implementing a philosophy of care within a facility (Akhter-Khan et al., 2022). Some interventions are complex and may address social isolation or loneliness through multiple or poorly specified mechanisms (Akhter-Khan et al., 2022; Holt-Lunstad, 2018; Lim et al., 2020).

Process indicators and outcomes

The potential of interventions to reduce social isolation or loneliness have been assessed through their acceptability, adherence, and feasibility. These process indicators determine progress toward outcomes such as health and psychosocial outcomes (e.g., loneliness, social isolation, social connectedness), indicators of social connections (e.g., social support, social engagement, social cohesion), as well as cost and cost-effectiveness outcomes. See Glossary of terms (Supporting Information Appendix 1).

We used this conceptual framework to define and code the intervention and outcome categories and subcategories for the two-dimensional matrix in the evidence and gap map.

Types of study design

We included on-going (protocols, trial registrations) and completed systematic reviews and primary studies with any study design that has a control group: randomized controlled trials, non-randomized studies including control before-after, and statistical matching quasi-experimental studies.

The inclusion of systematic reviews was based on the population, intervention, comparison, outcome (PICO) framework and if they met at least four of the five criteria of a systematic review as defined by Moher et al. (2015). That is, they describe adequate search methods used to identify studies, eligibility criteria for selection of studies, methods of critical appraisal of included studies, sufficient details or characteristics of included studies, and synthesis or analysis of findings of the included studies.

Eligible quasi-experimental designs included quasi-randomized studies, regression discontinuity designs, natural experiments, non-equivalent comparison group designs and interrupted series designs with at least three data points before and after a discrete intervention (Waddington et al., 2014).

We excluded any study designs with no control group such as longitudinal cohort studies and cross-sectional studies, and those studies with interrupted time series designs with less than six data points.

We included mixed methods studies, but exclusive qualitative research was excluded.

We included studies irrespective of their publication status (e.g., protocols, working papers).

Types of intervention/problem

We included any intervention which aims to reduce social isolation and/or loneliness that is delivered in-person regardless of the intensity, duration, and frequency of administration. We excluded digital or technology-based interventions.

Included interventions may be one-on-one or group based and were categorized based on our conceptual framework as self-delivery, interpersonal delivery, community-based delivery, societal-level delivery as well as multi-component or complex interventions. See Table 1 for subcategories and examples.

Comparison interventions included no interventions, other interventions, or usual care.

If reviews included a subset of interventions that is not eligible, we only coded studies with the eligible interventions.

Types of population

We considered any age group, people with or at risk of social isolation or loneliness, or the general population, whether based on case finding or screening for vulnerability or not.

Age groups include:

<10 years (children)

Types of outcome measures

We considered the following types of outcomes:

health and psychosocial outcomes,

We considered adverse effects of interventions such as psychological distress, safety, and others as described by the studies. Different measuring tools have been used for loneliness, social isolation, and related outcomes. See Table 2 for outcome categories and measurements.

We did not use outcomes as eligibility criteria; however, eligible studies and reviews must have assessed interventions with a primary objective to reduce social isolation and/or loneliness, and could have reported outcomes other than social isolation and loneliness. Studies and reviews were not included if the primary aim of the interventions assessed was not to reduce social isolation and/or loneliness. Those that assessed the effects of interventions on social isolation and/or loneliness as a primary outcome or considered other indicators of social connections, such as social support, social engagement, social cohesion, and social capital were also included.

Studies and reviews assessing the effect of interventions on indicators of psychological wellbeing such as quality of life, anxiety or depression, with a focus on mental health rather than social isolation or loneliness were excluded.

Other eligibility criteria

Report structure

We followed the reporting structure of Campbell EGMs with the standard headings: abstract, plain language summary, background, methods, results, discussion, and conclusion.

This report includes the description of the study flow with included studies, excluded studies, and any studies awaiting assessment as well as the PRISMA study flow diagram. We also present the conceptual framework, tables, and figures summarizing the distribution of included primary studies and systematic reviews across all the coding categories. These categories include study designs, publication status, quality of systematic reviews, types of interventions, types of outcomes, population characteristics, settings, geographic distribution.

The evidence and gap map has interventions as the row dimension and outcomes as the column dimension. We use bubbles of different sizes to represent included primary studies and systematic reviews and different colors to distinguish primary studies and methodological quality of systematic reviews. The filters used in the map are based on the number of included studies and coded information. See a sample of the map in Figure 2.

OPEN IN VIEWER

Figure 2

Sample map.

Filters for presentation

Additional dimensions of interest used as filters include: 1.

Study characteristics: the publication status of included studies, study design, countries where studies were conducted, World Bank classification by income (low income economies, lower-middle income economies, upper-middle income economies, high income economies), and WHO regions (African Region, Region of the Americas, South-East Asian Region, European Region, Eastern Mediterranean Region, Western Pacific Region), and setting (personal home, independent living/residential home, assisted living, long-term care/nursing home, orphanages, schools, workplaces, community centers, art gallery or museums, medical facilities, prisons), equity focus (study aimed at/focused on disadvantaged across any PROGRESS-Plus factors) and equity analysis (assessing any differences in effects (benefit or harm) across any PROGRESS-Plus factors).

Dependency

We treated multiple reports of the same study as one study. A study with multiple outcomes and interventions was shown multiple times on the map (once for each outcome or intervention identified). Primary studies were mapped regardless of whether they are included in multiple systematic reviews. Systematic reviews were mapped to interventions and outcomes based on their PICO question.

Screening and study selection

Pairs of reviewers used Machine learning text mining in Eppi-Reviewer web-based software program (Thomas et al., 2020) to screen titles and abstracts independently (EG, SD, EB, VB, TH, AW, AA, PD, JH, RD, SA, RI, LM, AAA, AJ, and FJ). We initially screened 10% of the titles and abstracts. The priority screening function developed a classifier based on the probability of inclusion determined from the preliminary screening results and presented the most likely studies to be included first. We manually screened all the articles to ensure all potentially eligible studies were captured for the full text screening stage which was also conducted by two reviewers independently.

We did not screen the reference lists of included systematic reviews as planned to identify additional studies that may have been missed in the database searches.

All screening was done following the eligibility criteria (see Appendix 3).

Data extraction and management

We developed and pilot tested a data extraction code set in Eppi-Reviewer for data collection (see Supporting Information Appendix 4). We used a set of included studies for testing. The same studies were coded by all the reviewers and the coding was assessed for agreement. Any discrepancies were discussed, and description of the coding criteria was modified for clarity as necessary. After the pilot test, members of the team (EG, SD, EB, VB, TH, AW, AA, PD, JH, RD, SA, RI, LM, AAA, AJ, and FJ) individually extracted and coded data. Some non-English papers (including Chinese, Mandarin, Korean papers) were coded by CWY, RY, and TAMTM. Automation and text mining was not used for coding.

We coded the study characteristics (study design, publication status, methodological quality assessment of systematic reviews), categories and subcategories of interventions and other intervention characteristics (focus, sectors, goals, and risk factors targeted), outcome domains and subdomains, population characteristics (using PROGRESS-Plus acronym), settings, locations (by country, WHO region, and World Bank income classification).

We coded how populations were recruited and whether they were selected based on disadvantages across any PROGRESS-Plus factors.

We also coded whether differential analysis across any PROGRESS-Plus factors was conducted in the studies and systematic reviews to understand any equity issues.

We did not contact authors of studies or systematic reviews for any missing information given the expected size of the map (over 300 studies).

Tools for assessing risk of bias/study quality of included reviews

Pairs of reviewers used the AMSTAR 2 tool (Shea et al., 2017) to assess the quality of systematic reviews independently. Any disagreements were resolved by discussion. Primary studies were not assessed for risk of bias or methodological quality following guidance for evidence maps (Snilstveit 2016; White et al., 2020).

Methods for mapping

We used the EPPI-Mapping tool (Digital Solution Foundry and EPPI_Centre, 2020) to develop the evidence and gap map. It is a tool to generate and visualize maps as an interactive HTML file on a web browser. The map is generated using exported data from EPPI-Reviewer, the software for managing and coding data.

Results of the search

Our search retrieved 15,698 records from databases and 100 from other sources. We screened 9704 titles and abstracts after duplicates and other ineligible studies were removed by automation tools. We identified 1244 potentially eligible articles and retrieved full texts of 1183 records that were assessed and 513 were included. See Figure 3 for the PRISMA flow diagram.

OPEN IN VIEWER

Figure 3

PRISMA flow diagram.

Excluded studies

A total of 670 studies were excluded during the full-text review phase due to an inappropriate aim (n = 327), intervention (n = 78), or study design (n = 192), and some as duplicates (n = 73) (Figure 3).

A list of key excluded studies can be found in the table for “Characteristics of excluded studies.” Three studies (Lee, 2014; Roessler, 2012; Scazufca, 2022) were excluded as the aim was not to reduce either social isolation or loneliness. Lee (2014) assessed the effects of a recall program on depression and self-esteem. Roessler et al. (2012) examined the effects of a physical activity and group counseling intervention on the quality of life in woman with polycystic ovary syndrome. Scazufca et al. (2022) assessed the effects of a psychosocial collaborative care intervention on depression in older adults who were disadvantaged socioeconomically.

One study (Lin et al., 2020) was excluded as the intervention group received horticultural therapy through 3D virtual reality.

Four studies were excluded for an incorrect study design (Fokkema & van Tilburg, 2007; Low et al., 2015; Mächler et al., 2022; Purewal et al., 2017). Fokkema and van Tilburg (2007) was a literature review of intervention to reduce loneliness that did not report using a systematic search. Low et al. (2015) was a primary study with no control group that assessed the effect of the Lifestyle Engagement Activity Program (LEAP) in older adults in aged care. Purewal et al. (2017) was a systematic review of companion animal studies that looked at association as opposed to effectiveness. Mächler et al. (2022) was a primary study that investigated the views of patients following participation in a holistic care program which aims to increase social activity instead of measuring effectiveness.

Studies awaiting classification (if applicable)

We were unable to find the full texts of 63 articles that were potentially eligible. See references of Studies awaiting classification.

Interventions

To be included in this EGM, interventions had to be focused on addressing either loneliness (reviews: n = 33, 36%; primary studies: n = 272, 65%), social isolation (reviews n = 28, 30%; primary studies: n = 82, 19%) or both (review: n = 31, 34%; primary studies: n = 67, 16%).

The included interventions were delivered using a group format (reviews: n = 74, 80%; primary studies: n = 335, 80%) or one-on-one (reviews: n = 51, 55%; primary studies: n = 103, 24%). For articles that included both group and one-on-one formats (reviews: n = 43, 47%; primary studies: n = 23, 5%), both options were coded. Few articles did not specify the format (reviews: n = 10, 11%; primary studies: n = 6, 1%).

The included articles involved interventions delivered across various sectors including clinical and population health, transport, housing, work, nutrition, environment, education, leisure: arts and entertainment, and spiritual care. Clinical and population health was the most reported sector (reviews: n = 74, 80%; primary studies: n = 295, 70%), followed by leisure (reviews: n = 40, 43%; primary studies: n = 112, 27%) and education (reviews: n = 22, 24%; primary studies: n = 52, 12%). Environment (reviews: n = 1, 1%; primary studies: n = 6, 1%), housing (reviews: n = 2, 2%; primary studies: n = 5, 1%) and transport (reviews: n = 2, 2%; primary studies: n = 0, 0%) were the least reported sectors. For interventions delivered across multiple sectors, all relevant sectors were coded (Figure 5).

OPEN IN VIEWER

Figure 5

Intervention sector reported in included articles.

The interventions were organized using the following categories according to who was administering the intervention: self-delivery interventions (reviews: n = 3, 3%; primary studies: n = 11, 3%), interpersonal delivery interventions (reviews: n = 72, 78%; primary studies: n = 284, 67%), community-based delivery interventions (reviews: n = 61, 66%; primary studies: n = 176, 42%), and societal level delivery interventions (reviews: n = 1, 1%; primary studies: n = 2, 0.48%). Some interventions involved multiple delivery strategies and were coded as complex (reviews: n = 5, 5%; primary studies: n = 47, 11%) in addition to all relevant delivery strategies. For example, an intervention that combined cognitive behavioral therapy and social support group was coded as both an interpersonal delivery and a community-based delivery intervention, as well as a multi-component/complex intervention (Figure 6).

OPEN IN VIEWER

Figure 6

Intervention delivery method reported in included articles.

Each intervention category included multiple subcategories. The few interventions that used a self-delivery mode were further categorized into two subcategories: self-guided interventions for changing maladaptive cognition (reviews: n = 3, 3%; primary studies: n = 6, 1%) and self-guided social skills training and psychoeducation (reviews: n = 0, 0%; primary studies: n = 5, 1%) (Figure 7).

OPEN IN VIEWER

Figure 7

Self delivery interventions reported in included articles.

Interpersonal delivery interventions had four subcategories: changing cognition, social skills training and psychoeducation, healthcare support, and social support. For both reviews and primary studies, changing cognition (e.g., cognitive behavioral therapy, psychotherapy) was the most reported (reviews: n = 52, 57%; primary studies: n = 151, 36%), while healthcare support was the least reported (reviews: n = 10, 11%; primary studies: n = 30, 7%). Some studies used multi-component interventions that involved multiple interpersonal delivery interventions (reviews: n = 40, 43%; primary studies: n = 32, 8%) so all relevant categories were coded (Figure 8).

OPEN IN VIEWER

Figure 8

Interpersonal delivery interventions reported in included articles.

Interventions with a community-based delivery mode were organized into five subcategories: group activities, support groups, neighborhood approaches, age-friendly communities, and volunteering. Group activities were most common for both reviews and primary studies (reviews: n = 40, 43%, primary studies: n = 106, 25%). Interventions to promote age-friendly communities (reviews: n = 1, 1%, primary studies: n = 0, 0%) and volunteering interventions (reviews: n = 8, 9%, primary studies: n = 2, 0.48%) were the least common. Some studies used multi-component interventions that involved multiple community-based delivery interventions (reviews: n = 22, 25%; primary studies: n = 9, 2%) so all relevant categories were coded (Figure 9).

OPEN IN VIEWER

Figure 9

Community-based delivery interventions reported in included articles.

Societal-level delivery interventions focusing on policies and norms encompassed four subcategories: public health policy that promotes social cohesion and inclusion, public education and awareness of social relationships, policies on the design of neighborhoods, and funding relevant research. Evidence across all four subcategories was very sparse with no reviews or studies reporting on public education and awareness of social relationships or funding relevant research. Only one primary study reported on public health policy that promotes social cohesion and inclusion (n = 1, 0.24%) and one on policies on the design of neighborhoods as an intervention (n = 1, 0.24%). Only one review reported about policies on the design of neighborhoods (n = 1, 1%).

Outcomes

The outcome framework consisted of four categories, each with a differing number of subcategories. These categories included: health and psychosocial outcomes, indicators of social connections, cost and cost-effectiveness outcomes, and process indicators or implementation outcomes. For studies that looked at more than one outcome, all relevant outcomes were coded. Most reviews and primary studies reported at least one health and psychosocial outcome (reviews: n = 91, 99%; primary studies: n = 417, 99%). Few reviews and primary studies reported cost and cost-effectiveness outcomes (reviews: n = 3, 3%; primary studies: n = 6, 1%) or process indicators or implementation outcomes (reviews: n = 3, 3%; primary studies: n = 34, 8%) (Figure 10).

OPEN IN VIEWER

Figure 10

Outcomes reported in included articles.

Health and psychosocial outcomes consisted of eight subcategories: loneliness, social isolation, social connectedness, quality of life/well-being, anxiety/depression, self-esteem or self-efficacy, adverse effects, and health services use. Loneliness was the health and psychosocial outcome reported most often (reviews: n = 67, 73%; primary studies: n = 325, 77%), followed by anxiety/depression (reviews: n = 48, 52%; primary studies: n = 181, 43%), and quality of life/well-being (reviews: n = 42, 46%; primary studies: n = 179, 43%). Adverse effects (reviews: n = 2, 2%; primary studies: n = 7, 2%) and health services use (reviews: n = 4, 4%; primary studies: n = 16, 4%) were the least reported health and psychosocial outcomes (Figure 11).

OPEN IN VIEWER

Figure 11

Health and psychosocial outcomes reported in included articles.

Indicators of social connections consisted of the following subcategories: social support, social engagement, social cohesion, and social capital. Social support (reviews: n = 36, 39%; primary studies: n = 82, 19%) and social engagement (reviews: n = 14, 15%; primary studies: n = 20, 5%) were the most reported indicators of social connections. Social cohesion (reviews: n = 0, 0%; primary studies: n = 2, 0.48%) and social capital (reviews: n = 2, 2%; primary studies: n = 6, 1%) were reported in few articles (Figure 12).

OPEN IN VIEWER

Figure 12

Indicators of social connections reported in included articles.

Cost and cost-effectiveness outcomes category had three subcategories: cost-effectiveness, cost per participant, and healthcare or social care utilization costs. Cost effectiveness (reviews: n = 3, 3%; primary studies: n = 5, 1%) was the most assessed, while healthcare or social care utilization costs (reviews: n = 1, 1%; primary studies: n = 2, 0.48%) and cost per participant (reviews: n = 1, 1%; primary studies: n = 2, 0.48%) were assessed in very few articles (Figure 13).

OPEN IN VIEWER

Figure 13

Cost and cost-effectiveness outcomes reported in included articles.

Process indicators or implementation outcomes consisted of the following subcategories: acceptance, adherence, and feasibility. More primary studies than reviews looked at these process indicators, with only one review reported to have looked at adherence, feasibility, and increased awareness of community services (n = 1, 1%). Acceptance (n = 21, 5%), feasibility (n = 15, 4%), and adherence (n = 14, 3%) were most common among the primary studies (Figure 14).

OPEN IN VIEWER

Figure 14

Process indicators or implementation outcomes reported in included articles.

Region

This EGM identified publications from 50 different countries across the globe. For the primary studies, the countries where the interventions were conducted were coded. For reviews, the countries specified for their included studies were coded. It should be noted that if a review included multiple studies conducted in the same country, that country was only coded once. If one primary study was included in multiple reviews the country was coded for each review.

The most common countries were the USA (reviews: n = 46, 50%; primary studies: n = 113, 27%), the UK (reviews: n = 32, 35%; primary studies: n = 35, 8%), China (reviews: n = 14, 15%; primary studies: n = 44, 10%), and Australia (reviews: n = 15, 16%; primary studies: n = 31, 7%) (Figure 16).

OPEN IN VIEWER

Figure 16

Geographic heatmap of included primary studies.

The location in which the interventions took place were further categorized using the World Health Organization (WHO) regions and the World Bank income classification. Most studies were conducted in the European Region (reviews: n = 59, 64%; primary studies: n = 138, 33%), followed by the Americas (reviews: n = 47, 51%; primary studies: n = 126, 30%), and then the Western Pacific Region (reviews: n = 42, 46%; primary studies: n = 124, 29%). Few studies were conducted in the African Region (reviews: n = 1, 1%; primary studies: n = 1, 0.24%). Over a quarter of the reviews included studies from multiple regions (reviews: n = 25, 27%) so all relevant regions were coded (Figure 17). Some reviews did not report the countries of their included studies and therefore the relevant WHO Region and World Bank Income Classification could not be determined (n = 15, 16%).

OPEN IN VIEWER

Figure 17

World Health Organization regions reported in included articles.

The World Bank Income Classification coding matches that of the WHO Region coding, as most studies were conducted in the European Region and the Americas, which consist mostly of high-income countries. Countries with high-income economies (reviews: n = 72, 78%; primary studies: n = 320, 76%) were reported most often, followed by upper-middle income countries (reviews: n = 28, 30%; primary studies: n = 67, 16%), then lower-middle income countries (reviews: n = 15, 16%; primary studies: n = 34, 8%). No studies took place in low-income countries. All relevant income classifications were coded for reviews that included primary studies from multiple countries with various income classes (reviews: n = 29, 31%) (Figure 18).

OPEN IN VIEWER

Figure 18

World Bank income classification of countries in included articles.

Settings

Many articles did not specify the setting (reviews: n = 56, 61%; primary studies: n = 134, 32%) that the intervention was administered in. Of those that did specify the setting, community settings, such as a community center or a park, were the most common (reviews: n = 24, 26%; primary studies: n = 90, 21%). Long-term care, or nursing homes were the second most reported setting (reviews: n = 23, 25%; primary studies: n = 53, 13%). Only one review reported the setting to be a prison (n = 1, 1%), and only 2 primary studies reported the setting to be an orphanage (n = 2, 0.48%). For studies involving interventions administered in multiple settings (reviews: n = 32, 34%; primary studies: n = 24, 6%), all relevant settings were coded (Table 4).

Participants needs

The needs of the participants in the included studies were identified. Social and emotional needs were the most prevalent (reviews: n = 89, 97%, primary studies: n = 401, 95%), followed by learning a new activity or language (reviews: n = 14, 15%; primary studies: n = 42, 10%) and clinical/health needs (reviews: n = 5, 5%; primary studies: n = 55, 13%). No reviews or primary studies reported participant needs for respite care. For studies involving populations with multiple needs (reviews: n = 22, 34%; primary studies: n = 123, 29%), all relevant needs were coded (Table 5).

Participant characteristics (PROGRESS-Plus)

We documented all PROGRESS-Plus factors described for participants in the included reviews and primary studies. Among the PROGRESS-Plus factors, age (reviews: n = 61, 65%; primary studies: n = 387, 92%) and gender or sex (reviews: n = 58, 63%; primary studies: n = 356, 85%) were the most reported. No reviews reported the religion of participants (Table 6).

Participant ages were arranged into groups based on life stages, such as childhood, adolescence, or middle-aged adulthood. Most reviews and primary studies included older adults between the ages of 60–75 (reviews: n = 55, 60%; primary studies: n = 226, 54%). Young adults aged 25–44 were the age group that was the least represented (reviews: n = 24, 26%; primary studies: n = 119, 28%) among adults. Among young people, adolescents aged 10–24 were the most represented (reviews: n = 25, 27%; primary studies: n = 135, 32%), while children under the age of 10 were included in the least articles (reviews: n = 11, 12%; primary studies: n = 34, 8%). All age groups were represented and more than half of the studies involved multiple age groups (reviews: n = 63, 70%; primary studies: n = 263, 63%). Around a third of the reviews and some primary studies did not report the age of participants (reviews: n = 33, 35%; primary studies: n = 39, 9%) (Figure 19). Overall, more reviews and primary studies combined focused on older adults ≥60 years (n = 307, 60%) than young people ≤24 years (n = 176, 34%).

OPEN IN VIEWER

Figure 19

Age of participants reported in included articles.

Of the studies involving children or adolescents (reviews: n = 25, 27%; primary studies: n = 151, 36%), no reviews and few primary studies reported parental status (reviews: n = 0, 0%; primary studies: n = 12, 8%). Less than 5% of the primary studies reported children to have single parents (n = 6, 4%) or working parents (n = 6, 4%).

Health condition of participants

Participants were reported as having specific health conditions in 226 articles (reviews: n = 41, 44%; primary studies: n = 185, 44%). Since depression is commonly associated with social isolation and loneliness, we presented it separate from other mental disorders. For reviews, mental health disorders such as schizophrenia, psychosis, bipolar disorder, neuropsychiatric disorders, attention deficit hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), or post-traumatic stress disorder (PTSD), but not including depression, were most reported (n = 14, 15%), followed by non-communicable diseases such as cardiovascular diseases, diabetes, or cancer (n = 12, 13%). No reviews reported participants with communicable diseases, psychological or social frailty, care dependence, discharge from hospital, or those in end-of-life/palliative care. For primary studies, non-communicable diseases were most reported (n = 63, 15%), followed by depression (n = 42, 10%). Discharge from the hospital and end-of-life or palliative care were reported in one primary study each (n = 1, 0.24%). All conditions were coded for studies involving participants with multiple conditions (reviews: n = 24, 26%; primary studies: n = 40, 16%) (Table 7).

Equity analysis

All but two reviews and 14 primary studies reported an equity focus on populations disadvantaged across PROGRESS-Plus. The most frequently reported focus was age with nearly 60% of reviews and primary studies focused on age (reviews: n = 53, 57%; primary studies: n = 247, 59%). The next most reported equity foci were loneliness (reviews: n = 40, 43%; primary studies: n = 204, 49%), health status (reviews: n = 20, 22%; primary studies: n = 116, 27%), social isolation (reviews: n = 32, 34%; primary studies: n = 52, 12%), and living situation (reviews: n = 16, 17%; primary studies: n = 69, 16%), respectively. Religion was the least reported focus, with only one primary study reporting it as a focus (n = 1, 0.24%). Two thirds of the articles focused on populations disadvantaged across multiple progress factors (reviews: n = 59, 63%; primary studies: n = 281, 67%) (Table 8).

Included articles used various strategies to identify participants at risk for social isolation or loneliness. Case-finding was the most common strategy for reviews (n = 11, 12%), while outreach was the most common for primary studies (n = 90, 21%), followed by case-finding (n = 77, 18%). Some of the articles utilized multiple strategies to recruit participants that were disadvantaged (reviews: n = 8, 9%; primary studies: n = 34, 8%). Most reviews and a quarter of the primary studies did not specify a strategy (reviews: n = 70, 76%; primary study: n = 105, 25%) (Figure 20).

OPEN IN VIEWER

Figure 20

Recruitment strategies for at-risk populations reported in included articles.

Though many articles reported an equity focus, few conducted an analysis of the effects across PROGRESS-Plus. More primary studies than reviews performed a subgroup or regression analysis. Of the primary studies that performed an analysis, analysis by gender or sex was the most common (n = 13, 3%), followed by health status (n = 9, 2%), social capital (n = 7, 2%), and age (n = 5, 1%). For the reviews, analysis by age and health status (both n = 2, 2%) were most common. No reviews or primary studies performed an analysis to assess differences in effects across occupation or education. Few articles conducted an analysis of the intervention effectiveness across multiple PROGRESS-Plus factors (reviews: n = 1, 1%; primary studies: n = 10, 2%) (Table 9).