Author year [reference] | Effective | Noted differences across groups / characteristics |
---|---|---|
Effective | ||
Collins 2006 [25] | Yes – reduction in mean loneliness with statistically significant improvement | Greatest reduction of loneliness observed among ethnic minorities, amongst whom lowest income reported significantly less loneliness than highest income. Minority participants with highest education had significantly greater improvement compared to second and third levels but were not significantly different from those at lowest education level |
Creswell 2012 [26] | Yes – reduction in mean loneliness and statistically significant model interaction | More participants dropped out of intervention group; difference was marginally significant. No significant differences between groups regarding baseline demographics or pairwise comparison. No significant differences between dropouts in primary treatment trial and treatment completers |
Gaggioli 2014 [30] | Yes – reduction in mean loneliness with statistical significance for general and emotional | The decrease in emotional but not social loneliness could be due to the nature of the intervention, in particular reminiscing to promote feelings of togetherness and intimacy |
Larsson 2016 [38] | Yes – statistically significant reduction in mean loneliness for both intervention-control sequences, and negative percentage change | No significant difference between T3 and T2 for group 1 [I/C]. Majority of participants were women living alone, representing those more vulnerable to loneliness. However, despite this study population characteristic, none of the participants reported high baseline loneliness, indicating the intervention was tested on a group with low to moderate loneliness. Interaction between intervention and sequence did not achieve statistical significance |
Ehlers 2017 [27] | Yes – reduction in mean loneliness and statistically significant latent change score | Persons with higher levels of baseline loneliness at demonstrated greater decreases over the course of the intervention. Mediation model found greater decrease in stress explained greater reductions in loneliness, and increased social support was directly related to decreased loneliness. Collectively changes in social support and stress explained around 26% of the variability in change in loneliness. Mode of exercise intervention did not account for individual differences in loneliness with similar change in loneliness observed across different intervention conditions |
Bouwman 2017 [22] | Yes – reduction in mean loneliness and statistically significant linear regression coefficients | Number of lessons did not affect loneliness in the full group, and only slightly increased loneliness in the light program. Regulative coping is more effective than active coping in alleviating today’s loneliness. Higher levels of loneliness when the assignment did not go well (full program). Practicing (e.g. assignments) is more effective than just reading about coping strategies |
Cohen-Mansfield 2018 [24] | Yes – reduction in mean loneliness, significant decline in loneliness for intervention as compared to control group | Provided option of group or individual based on pilot work which found some people were not comfortable in groups, or not initially willing to participate in groups. Individual sessions allow work on specific barriers and solutions, group sessions allow participants to practice and share solutions. Significant effect for interaction of intervention group by time. Baseline loneliness and number of group sessions attended are significant predictors of the final loneliness score. Impact of group setting likely reflects both the impact of the group and that those who attended group sessions were more ready to enhance social activities and tackle loneliness |
Hwang 2019 [32] | Yes – qualitative and quantitative decreases in loneliness stated | UCLA loneliness score showed significant decrease in loneliness and de Jong Gierveld showed significant decrease in emotional loneliness. No significant change on Lubben score, possibly due to the short intervention period |
Kall 2020†[33] | Yes – reduction in mean loneliness, positive effect size, and statistically significant model outputs for intervention and waitlist | Nonsignificant relationship between post-treatment loneliness and ‘dose’ (i.e., loneliness not significantly related to number of completed modules or average treatment time from therapist)†|
Ghanbari 2021 [31] | Yes – significant reduction in mean loneliness, and difference between intervention and control | NR |
Fong 2021 [29] | Yes –statistically significant model outputs for reduced loneliness | Greater baseline loneliness was reflected in greater loneliness at follow-up. The higher the level of education the larger the reduction in loneliness at follow-up |
Kall 2021* [34] | Yes – reduction in mean loneliness, CBT favoured over waitlist with statistical significance. No - IPT did not present statistically significant results | Found significant heterogeneity in the initial level of loneliness and slope |
Nazari 2021 [39] | Yes - reduction in mean loneliness with statistically significant difference between intervention and control after the study | No statistically significant difference across gender, while other studies have shown that women are lonelier than men |
Kotwal 2021 [36] | Yes – statistically significant reduction in loneliness score | Suggests the one-to-one intervention was more successful than previous interventions using telephone-based support, gatekeepers or clinical case managers. Suggest this could be due to the social experience rather than treatment with medical providers; flexibility of number, frequency and goal of sessions; motivational interviewing and companionship to promote safety; focus on unique needs of participants |
Can’t tell if effective | ||
Steven 2000 [43] | Can’t tell - reduction in mean loneliness for both intervention and control groups, greater mean change score for intervention than control | NR |
Rolandi 2020 [40] | Can’t tell – non-significant differences for total loneliness score, some benefit to specific feelings | Between-group differences observed for individual UCLA scale items e.g. cross-sectional analysis for feeling left out highlighting potential benefit of social network site use for specific loneliness feelings |
Caputi 2021 [23] | Can’t tell – reduction in mean loneliness and statistically significant model outputs in short-term but not long-term | Significant negative effect of verbal ability. Reduced loneliness among ToM training group likely due to engagement in discussions about different perspectives while no-ToM group discussed non-social stories. Higher vocabulary scores predicted lower loneliness thus language has a protective role against high perceived loneliness |
Kanter 2021 [35] | Can’t tell – reduction in mean loneliness, statistically significant intervention effect at start but lost by final day | The effect of the intervention on loneliness increased over the intervention period and was strongest on the last day of the intervention, reducing after the intervention was discontinued. No covariates were significantly associated with differing intervention effects. Characteristics associated to higher odds of missing or incomplete surveys include: assigned to intervention, being farther along in the study, younger age, living farther north. Low participating participants more likely: younger, non-white, unmarried, income <$10,000 per year |
Shapira 2021 [42] | Can’t tell – reduction in mean loneliness but only achieved statistically significant difference at T1 | Significant main effect of time-by-group interaction indicating groups differed in loneliness post-intervention. Main effect of time did not reach statistical significance. Lack of continued decrease after 1-month follow up (T2) implies group contacts can lead to a decline in loneliness once interactions become less frequent. |
Not effective | ||
Kramer 2022 [37] | No – no decrease in loneliness | None of the demographic characteristics were significantly associated with loneliness. Number of chat messages correlated with, but did not predict, loneliness |
Fields 2021 [28] | No – overall no quantitative change in loneliness | Authors suggest the lack of quantitative change could be due to additional pre-existing contextual factors in the daily lives of participants (e.g. physical disability, lack of close friends or living relatives etc.) making loneliness more systemic and harder to change. Additionally, could be confounding from tackling the digital divide alongside loneliness |
Sandu 2021 [41] | No | No significant relationship between change in loneliness and either duration or number of calls. However, did notice a trend towards significance in relationship between UCLA loneliness score and increasing duration of calls |