Following removal of non-English language and duplicate articles, 2852 articles remained for screening by title and abstract (Figure 1). 138 articles underwent full-text screening during which 118 were excluded, leaving 20 articles. Three additional articles were included after searching reference lists of full-text articles. 19 quantitative and four qualitative studies were included in the quality assessment.
Findings from quantitative research
Quality assessment of quantitative studies is presented in Figure A1 (randomised control trials, n= 11) and Figure A2 (non-randomised designs, n=8) in Additional file 3 [see Additional file 3]. The study by Mohamed et al. (29) was found to have an overall critical risk of bias arising mostly from missing data; this article was removed from further review.
Of the remaining 18 studies, eight were judged as having overall high or serious risk of bias and ten were judged as having overall moderate risk or some concerns. Of the studies judged to have a high or serious risk of bias, most (n=6) were due to a lack of blinding of participants. As all measurements of loneliness were self-reported, knowledge of the intervention may have influenced participants’ responses. Therefore, findings of studies judged as having high or serious risk of bias were interpreted cautiously.
All studies other than Mascaro et al. (30), in which all analyses were conducted using blinding, were judged as either moderate risk or some concerns for risk of reporting bias. No further articles were removed due to concerns of risk of bias.
Of 18 studies using quantitative measurements of loneliness, 11 were randomised control trials (RCT), three were controlled pre-test, post-test designs and four were pre-test, post-test designs without a control group. 11 of the 18 studies were published within the past ten years, the remainder being conducted between 1985 and 2010.
The characteristics of each study are presented in Table 1. The majority of the studies were conducted in the U.S. (n= 14); one additional study was conducted in each of the following countries: UK (31), Turkey (32), Canada (21) and China (13). All participants were university students, however, some articles specified subgroups, including students with autistic spectrum disorder (33), graduate students (34), medical students (20, 30), elementary education students (32), psychology students (35, 36), first year students (37, 38), students displaying elevated levels of loneliness at baseline (13, 21, 31, 39) or students self-identifying as stressed, anxious, or depressed, or with a pre-existing mental health condition (31). The sample sizes in the included studies ranged from 2 (40) to 112 (36).
Tools used to measure reductions in feelings of loneliness were relatively consistent across the research, with most (n=12) using the Revised UCLA Loneliness Scale or UCLA3. The UCLA3 consists of 20 items assessing subjective feelings of loneliness and feelings of social isolation, using questions such as “How often do you feel close to people?”. Participants rate each item on a scale of 1 (Never) to 4 (Often) (41). Other tools used to measure feelings of loneliness included one item of the PERMA profiler (36), a three-item loneliness Likert scale (35), The University of the Philippines Loneliness Assessment Scale (UPLAS) (42) and the Chinese College Student Loneliness Scale (13).
Measurement points varied. A pre- and post-test measurements only were taken in five studies. An additional mid-test measurement was implemented in six studies (including Barber (22) in which multiple mid-test measurements were taken). A further follow-up measure was taken in six studies. The length of time between post-test measurements and follow-up measurements varied from two weeks (43) to around six months (20, 37, 44).
13 studies explored the effectiveness of a single intervention whereas five studies compared the effectiveness of multiple interventions. In total, 18 studies covered 25 interventions. While the aim of three studies was solely to reduce levels of loneliness (13, 35, 39), most used loneliness as one measure of intervention effectiveness among others, such as happiness (32, 36), anxiety (31, 32, 42), adjustment to college (37), academic achievement (38), self-esteem (21, 22, 33, 40), confidence (44), overall wellness (20, 34), empathy (20), depression (21, 31, 40, 43) and compassion (30).
The nature of the interventions followed common themes: some provided psychoeducation (21, 35, 36, 39, 43), some implemented social support groups in which students could discuss problems (20, 33, 37, 38, 44), some provided an opportunity for social interaction with other students outside of a support group, with or without a specific activity or topic of discussion (22, 31, 32, 40, 42) and others implemented reflective exercises such as mindfulness (13, 22, 30, 34, 35).
The majority (n= 21) of interventions were implemented in a group setting, whereas a smaller number (n= 4) were implemented on an individual. Two group interventions were delivered online (22, 40) as was one individual intervention (22). Of the group interventions, group sizes varied from two participants (40) to 112 participants (36).
Time commitments required from students to partake in each intervention varied. The longest duration of any intervention session was 180 minutes (32), with others lasting between 30 minutes and 120 minutes (Table 1). Overall, support group interventions appeared to require longer sessions whereas individual interventions were found to use shorter sessions.
Of the studies that included a control condition (n=14), most (n=8) did not implement an intervention for this group. Instead, measurements coincided with those taken from the intervention group. Six studies, however, created an alternative intervention to act as a control condition. These included the provision of general information (35, 37, 38), an educational course not aimed at reducing loneliness (32, 36) and a self-help discussion (39).
Of 25 interventions, 15 interventions found a statistically significant effect in decreasing loneliness scores (Table 2). These included three psychoeducation interventions (21, 36), four reflective exercises (13, 22, 30, 34), four social support interventions (33, 37, 38, 44) and four social interaction interventions (22, 32, 40, 42). Seven interventions showed a reduction in loneliness score, however this was not deemed to be statistically significant (20, 31, 35, 43).
Two out of four individual interventions (performed with or without the use of the internet) showed statistical significance (22, 34), the two not showing statistical significance were psychoeducation type (43). Of 21 group interventions, 13 showed statistical significance (13, 21, 22, 32, 33, 36, 38, 40, 42, 44).
Of the interventions performed online, all three were effective in reducing loneliness scores regardless of whether they were performed in an online group or individually (22, 40). Two of these interventions involved chatting to another online (22, 40), and the other involved typing journal entries (22).
Three of ten psychoeducation interventions showed statistical significance (21, 36), as did four of five social support interventions (33, 36-38, 44), and four of five reflective interventions (13, 22, 30, 34). Three reflective exercises utilised meditation and mindfulness (13, 30, 34) and one utilised writing in a journal (22). Four of five social interaction interventions demonstrated a statistically significant reduction in loneliness scores (22, 32, 40, 42), one of which utilised non-human interaction with a dog and facilitator (42).
The largest proportion of interventions failing to show a statistically significant reduction in loneliness score were in the psychoeducation category, of which, five of seven were published in 1996 or earlier (39, 43).
Studies involving students with pre-existing mental health problems or feelings of loneliness (diagnosed or self-declared) included a range of interventions including social interaction (31), reflective (13), and psychoeducation interventions (21, 39, 43). However, only three interventions out of nine targeting these students demonstrated a statistically significant reduction in loneliness score (13, 21).
Although no study compared the effect of a single intervention across subgroups, reductions in feelings of loneliness appeared consistent across groups. For example, reductions in loneliness were seen across students of all educational disciplines, in those with autism spectrum disorder (ASD) (33) and in students from varied geographic locations. Similarly, reductions in feelings of loneliness were reported in studies sampling mostly female students (21) and in studies sampling mostly male students (33).
Findings from qualitative research
Quality assessment of qualitative studies is presented in Figure A3 (n=4) in Additional file 4 [see Additional file 4]. Three studies (45-47) had an overall moderate risk of bias. Horgan et al. (48) was found to have a serious risk of bias due to missing data. The primary objective was to assess a decrease in depressive symptoms, however, only a small number of participants (7 out of 118) provided forum posts at follow up, some of which related to the reduction of feelings of loneliness.
All qualitative studies were judged as having moderate risk of bias due to confounding factors. This arose from a lack of randomisation in the allocation process.
All interventions explored through qualitative research (n=4) were delivered as part of a group, with one (48) delivered in an online format (Table 3). Two studies recruited medical students (45, 46), one recruited black, male students (47) and one recruited students scoring highly on the C-DES scale of depressive symptoms (48). Three studies were conducted in the USA (45-47) and one in Ireland (48).
Clark et. al. (47) collected data through semi-structured, open ended interviews as opposed to questionnaires or forums (45, 46, 48). Horgan et al. (48) and Brodkin et al. (45) primarily used quantitative methods to measure outcomes other than loneliness, for example, reductions in depressive symptoms. These were included as reductions in loneliness were commented on qualitatively.
Data were collected at different points across the research: Brodkin et al. (45) adopted pre-test and post-test measurements, as did Horgan et al (48). Clark et al (47) and Goetzel et al. (46) collected data post-test only. Although this design prevents a comparison to baseline levels of loneliness, participants were asked to give their opinions of the intervention, not their levels of loneliness.
The types of interventions implemented included support groups and increasing social interaction (Table 3). Support groups included the “Human Dimension Group” which adopted a face-to-face discussion around student difficulties with support from a facilitator (46). An online forum allowed students to support each other virtually with access to online resources (48). The “Black Male Scholars Programme” improved social interaction by encouraging students to play sports together and by enlisting speakers on a number of topics (47). A parenting and professionalism course also improved social interaction in which students learned about parenting (45).
Table 4 summarises the findings of interventions using qualitative data collection. The results of a study sampling students of black African or Caribbean descent showed that bringing students together who have a shared cultural experience can reduce the feelings of loneliness (47). Feedback for the Black Males Scholar Programme was generally positive across participants and focussed on reducing alienation of ethnic minorities in universities.
Horgan et al. (48) reported that, although primarily aimed at reducing depressive symptoms, the online forum resulted in participants feeling less lonely due to being able to talk to others in a similar situation to theirs. Being able to tell someone about their problems improved feelings of loneliness: “It's about empathy and the realization that you're not alone. Being able to tell someone/people, even if it is just anonymously putting it out there into the virtual domain, is supposed to be cathartic in some way. And it kinda’ is.” (48)
Studies investigating medical students showed that support groups were beneficial, although Brodkin et al. (45) specified that the intervention group reported no increase in loneliness as opposed to a decrease. Feedback for both interventions was generally positive across participants.