Mental disorders are the leading cause of overall disease burden for young people worldwide [1]. Three quarters of mental health problems emerge before 25 years of age [2], and there is evidence that mental disorders are becoming more prevalent in young people [3, 4]. Depression is a common mental disorder which has a range of impacts on individuals, as well as their families and communities. Health-related impacts of depression include suicide, physical illness, and premature mortality, while non-health related outcomes include poorer education and employment outcomes [5].

It is widely documented that loneliness is associated with an increased risk for depression [6, 7]. For example, recent research indicates that measures of loneliness in young adults are predictive of future antidepressant use [8]. Loneliness is defined as a discrepancy between the quantity and quality of desired social relationships and actual social relationships [8], and has recently been recognised as a major public health concern for societies [9].

Research indicates that experiences of loneliness are highest in younger adults and older people [10, 11]. A recent cross-temporal meta-analysis of 437 independent samples of young adults indicated that there have been increases in loneliness since 1976, implying that loneliness is a rising concern among young adults [12]. More recently, data shows that young adults experienced greater feelings of loneliness and depression during the COVID-19 pandemic than other age groups. These findings were seen across numerous high-income countries [7, 13,14,15,16], with a US study reporting that increases in loneliness accounted for much of the increase in depression among young adults during the pandemic [17].

Early adulthood is a unique developmental period characterised by social, environmental, educational, economic, cognitive, and psychological change [8]. Contemporary cohorts of young adults have also grown up alongside significant societal developments which have changed the nature of human relationships, such as greater mobility opportunities and changes in communication due to technological advancements [12]. The culmination of these unique experiences may explain young adults’ vulnerability to both loneliness and mental health difficulties.

Given the relationship between loneliness and depression, social interventions which aim to increase social connections, or reduce loneliness, may be important in protecting against depression in young adults. This is further supported by the claim that social capital, as a major social determinant of mental health, is protective against depression [18, 19]. Most literature concerning social connection interventions currently focuses on children and adolescents [20,21,22], older people [23], or clinical samples with pre-existing mental health problems [24,25,26] only. A lack of focus on social connection interventions for young adults in the general population is apparent. This was recently reflected by a 24-year-old participant in a qualitative study, in which she described early adulthood as “a time in life that often gets overlooked” [27].

It is important to address this gap given the unique challenges presented in early adulthood and current prevalence of mental health problems and loneliness. Systematically reviewing the effectiveness of social connection interventions in reducing depression among young adults is necessary to address this gap. The aims of the current review were to (1) identify and present interventions which address social connection/loneliness in young adults, and (2) describe the effectiveness of these interventions in changing depression and social connection/loneliness outcomes.

Methods

A systematic review was conducted, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (Supplementary File 1). A protocol was pre-registered on PROSPERO (CRD42023395595). In the protocol, it was originally anticipated that heterogeneity in outcome measures would mean that meta-analysis would not be possible. Following study selection and data extraction, it was evident that meta-analysis was possible for depression and loneliness outcomes and was therefore added to the review.

Data source

Three databases were searched: PubMed, Scopus, and PsycInfo. The searches were conducted from 01 January 2000 to 01 January 2023 and were limited to English language. The search strategy included terms such as: “young adults”, “social interventions”, “social connectedness”, “depression”, and “randomized controlled trials” (see Supplementary File 2 for the full search strategies).

Inclusion/exclusion criteriaStudy designs

Peer-reviewed quantitative studies assessing the effectiveness of an eligible intervention were included. This included randomised controlled trials (RCTs), quasi-experiments, and pre-post studies with a comparison group. Qualitative studies, reviews, commentaries, editorials, and book chapters were excluded.

Participants

Young adults aged 18–24 years were included, and all other age groups were excluded. Participants from any setting (e.g., educational, occupational, community, others) in the general population were eligible for inclusion. Clinical patient populations explicitly recruited from clinical settings like mental health services were excluded.

Interventions

Interventions aiming to increase the quantity or quality of social connections or reduce loneliness in young adults in the general population, were included. These interventions could be in-person or online but needed to explicitly state that they aimed to increase social connections or reduce loneliness. For example, social media platforms were not automatically included as “social interventions” because, while some social media platforms (e.g., Facebook, Instagram) may have initially been designed with the intention to socially connect users through two-way interaction, they have evolved over time to become marketing tools with in-built designs that commonly encourage one-way non-interactive engagement (e.g., scrolling a social media feed with products, advertisements, and entertainment) [28].

Interventions which exclusively used psychological approaches such as cognitive behavioural therapy (CBT), mindfulness, or acceptance and commitment therapy (ACT), were excluded.

Comparator

Studies had to include a comparison group. Eligible interventions could be compared to either no intervention or to a different intervention.

Outcomes

To be eligible for inclusion, studies had to include both our primary and secondary outcomes of interest. The primary outcome we assessed was the effect of interventions on changes in the severity, course, or prevalence of depression or depressive symptoms. Studies needed to have a measure of depression at baseline and after the intervention, using a valid and reliable tool. Secondary outcomes we assessed were measures of social connection(s) and/or loneliness, which also had to be measured using a valid and reliable scale.

Context/settings

No restrictions were placed on the context/settings. Studies were included from low-, middle-, and high-income countries. Interventions could be conducted in a range of settings, such as college/university, sporting clubs, or in community settings more widely.

Study screening and selection

The electronic database search results were downloaded to Endnote 20.4.1. All titles and abstracts were screened by one reviewer (CVA). Applying the inclusion criteria, studies were either excluded or progressed to a full-text screening stage if they were potentially relevant. One reviewer screened all full-text articles (CVA), and a random 10% sub-sample were screened by a second reviewer (TKO). The two reviewers had 100% agreement. The reference lists of the included articles were also screened to ensure no studies were missed. The corresponding author of each included article was contacted to request any additional studies meeting the eligibility criteria.

Data extraction

Data extraction was conducted in Microsoft Excel, using a form designed and tested by the study authors. Data extracted included: first author, year of publication, study design, location, description of population (including age, gender, ethnicity), description of intervention, intervention delivery mode (online or in-person), total number of participants (at baseline and at follow-up), depression measure used, social connection/loneliness measure used, main findings related to depression, main findings related to social connection/loneliness outcome, and funding information. All data was independently extracted by two authors (CVA and LM) and discrepancies were addressed by a third author (TKO).

Risk of bias assessments

Risk of bias assessments were conducted using two tools. The Cochrane Risk of Bias tool for Randomised Controlled Trials (RoB 2.0) was used for studies that used a randomised controlled trial design. The Risk of Bias in Non-Randomised Studies (ROBINS-I) tool was used for studies that used a non-randomised design. All studies were independently assessed by two reviewers (CVA and LM). Any discrepancies between the reviewers were resolved through discussion, and the other reviewers (TKO and JDM) were consulted to resolve final uncertainties.

Narrative synthesis

Narrative synthesis was conducted and summaries of the included interventions, depression outcomes, and social connection/loneliness outcomes were tabulated. To compare intervention types, studies were grouped by their intervention delivery modality (e.g., in-person or online) and by the study sample type (e.g., higher education students or not).

Statistical analysis

Random effects meta-analyses were performed, utilising standardised mean differences (SMD) to account for different depression and loneliness measures used across studies, with an assessment of I2 for heterogeneity. The metan suite of commands were utilised in STATA/MP 18.0 for the analysis. To assess for publication or small study biases, we created and visually inspected a funnel plot. Egger’s test was not used as there were less than 10 studies [29]. We presented overall effects as well as subgroup effects for the different intervention types, separated by delivery modality and study sample type. Sensitivity analyses were also conducted and reported by excluding the studies which were assessed as having serious risk of bias.