Ethics approval for this study was obtained from the University of New South Wales Human Research Ethics Committee (HC190764). All procedures performed in the study involving human participants were in accordance with the ethical standards of the institutional or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. As such, informed consent to participate was obtained from all participants prior to data collection.

The study was a 2-arm parallel, double-blind, randomized controlled trial conducted at the Black Dog Institute (BDI), Australia. Participants were recruited from the community via targeted advertisements posted on Facebook between May 11 and May 22, 2020 (Multimedia Appendix 1). Eligibility criteria were as follows: aged between 18 and 25 years, fluent in English, own and have access to a smartphone, currently residing in Australia, and having experienced suicidal thoughts in the past 12 months. Participants were excluded if they reported being diagnosed with a psychotic disorder ever or had attempted suicide in the 1 month prior to trial registration due to safety considerations. Participants undergoing other mental health treatments were not excluded from the trial. The study was approved by the University of New South Wales Human Research Ethics Committee (HC190764) and required written opt-in consent from participants.

After completing the baseline survey, participants were assigned randomly (2:2) to the intervention or control condition within a block design (four per block). Randomization was performed using a computer-generated algorithm integrated into BDI’s digital trial management platform. All study participants and investigators were masked from condition assignment until the completion of the final survey.

The trial was run online using this trial management platform, which also supported the recruitment portal, including online consent, screening, registration, data collection, and access to the apps (see [25] for full details). Participants were required to complete these closed (password-protected) surveys within 7 days at baseline (T0, day 0), post intervention (T1, day 43), and 3 months post intervention (T2, day 132). After completing the baseline assessment, participants were sent a link (via email and SMS text message) to download the free app randomly allocated to them and informed that they had 6 weeks to access this app. No reminders/prompts or support were provided to use the apps.

LifeBuoy is a brief self-guided smartphone intervention consisting of seven models grounded in dialectical behavior therapy and acceptance and commitment therapy. Each module takes approximately 5 minutes to complete, and participants had the opportunity to return to them as frequently as needed over the course of 6 weeks. Participants were required to complete each module sequentially before moving on to the next. Each module introduced and practiced knowledge and skills relating to distress tolerance, emotional regulation, and goal setting through interactive learning exercises. A detailed description of LifeBuoy has been published elsewhere along with screenshots of the app [25].

A matched-attention placebo app (LifeBuoy-C) was developed for the control condition (details published elsewhere) [25]. LifeBuoy-C provided generalized mental health information in a similar format and time expectancy to the intervention app to control for potential digital placebo effects [38]. LifeBuoy-C mirrored the module structure, user interface, and graphic design of LifeBuoy but contained no therapeutic content.

To determine whether the loneliness levels of participants at baseline moderated the effect of the LifeBuoy and LifeBuoy-C apps on suicidal ideation (SIDAS) and depressive symptoms (PHQ-9) across time (T0 to T1; T1 to T2), piecewise linear mixed models for repeated measures analyses were used. Thus, loneliness (TILS), two time variables reflecting T0 to T1 and T1 to T2, condition (LifeBuoy, LifeBuoy-C), and relevant interaction terms between these variables were specified as fixed effects in models, along with an intercept as a random effect. An identity covariance matrix was used, and degrees of freedom were estimated using Satterthwaite’s method. The linear mixed model approach can include all participants in the analyses, even those with missing follow-up data points, under the missing-at-random assumption, and therefore accords with the intention-to-treat principle.

To examine within the LifeBuoy condition whether app engagement affected the relationship between loneliness and each of the key outcomes (suicidal ideation, depression), piecewise linear mixed models for repeated measures analyses were conducted similar to the above approach described. In this case, app engagement (the number of modules completed), loneliness (TILS), two time variables reflecting T0 to T1 and T1 to T2, and relevant interaction terms between these variables were specified as fixed effects in models, along with an intercept as a random effect. Broader engagement data (number of modules completed by condition and median time spent in the app by condition) has been reported elsewhere [25]. Safety data and harms have also been reported elsewhere [24].

For all analyses, P values ≤.05 were considered significant. For all linear mixed models examined, residual plots indicated no distinct patterns, and the distribution of residuals in histograms approximated normality, indicating model assumptions were met. SPSS 25.0 (IBM Corp) was used for all analyses. Attrition analyses were conducted and reported elsewhere [25].

In total, 455 young adults completed the baseline survey and were randomized to either the intervention condition (n=228) or the control condition (n=227). No participants formally withdrew or were excluded from the study. Demographic and clinical profiles of the total sample are described elsewhere, along with a CONSORT (Consolidated Standards for Reporting Trials) trial flow diagram [25]. The mean loneliness score at baseline was 7.17 (SD 1.83), with no significant differences between the intervention and control conditions (7.19 vs 7.16; F=0.89; t₄₅₃=0.16; P=.44). The mean SIDAS score was 22.61 (SD 8.18), which indicates clinically relevant levels of suicidal ideation (scores ≥21), with 136 (59.6%) participants in the intervention condition and 129 (57.1%) participants in the control condition meeting this criterion. The mean PHQ-9 score was 17.15 (SD 5.64), indicating that the sample, on average, met the clinical threshold for moderate-severe depression (scores ≥15). A total of 152 (66.7%) participants in the intervention condition and 156 (69.3%) participants in the control condition met the moderate-severe depression threshold. There were 202 (90.2%) participants in the intervention condition and 199 (88.4%) participants in the control condition who endorsed ever receiving mental health treatment. In the intervention condition, there was no significant difference in the number of modules completed between those who have (mean 6.80, SD 4.41) and have not (mean 6.77, SD 2.84) ever received mental health treatment (F_1,222<1; P=.98). In the control condition, there was no significant difference in the number of modules completed between those who have (mean 6.84, SD 4.96) and have not (mean 5.81, SD 3.82) ever received mental health treatment (F_1,223=1.04; P=.31).

In the intervention condition, there was a significant positive correlation between baseline loneliness and PHQ-9 scores (r=0.32; P<.001) but not between loneliness and SIDAS scores (r=–0.03; P=.67). In the control condition, significant positive correlations were observed between baseline loneliness and SIDAS scores (r=0.16; P=.02) and loneliness and PHQ-9 scores (r=0.27; P<.001).

For the model with suicidal thinking (SIDAS) as the dependent variable (Table 1), there was a significant main effect of loneliness (P=.03), indicating that higher levels of loneliness were positively associated with higher levels of suicidal thinking overall, irrespective of condition or time point. There was also a significant time 1 × condition interaction (P=.02), as expected given the primary outcomes analyses previously reported [25] reflecting a nonsignificant change in suicidal thinking in the LifeBuoy-C condition (B=–3.23, 95% CI –10.48 to 4.01; z=–0.87; P=.38; d=–0.38) but a significant decrease in suicidal thinking in the LifeBuoy condition (B=–7.84, 95% CI –14.76 to –0.92; z=–2.22; P=.03; d=–1.00), resulting in a between condition effect size (d=0.42) at T1. There were no other significant effects. For the model with depression (PHQ-9) as the dependent variable (Table 1), there was a significant main effect of loneliness (P<.001), indicating that higher levels of loneliness were positively associated with higher levels of depression overall, irrespective of condition or time point. There were no other significant effects.

Among the LifeBuoy condition (n=228), the mean number of app modules completed was 6.84 (SD 4.27), and loneliness scores at baseline were not significantly correlated with module completion (r=0.05; P=.42). In examining whether loneliness moderated the relationship between the number of LifeBuoy modules completed and clinical benefits, there were no significant effects in the model with suicidal thinking (SIDAS) as the dependent variable (Table 2). For the model with depression (PHQ-9) as the dependent variable (Table 2), there was only a significant main effect of loneliness (P<.001), indicating that higher levels of loneliness were positively associated with higher levels of depression overall, irrespective of time point or engagement.

This study investigated whether loneliness played a role in moderating the efficacy of and engagement with a therapeutic smartphone intervention designed to help young people manage suicidal ideation. Baseline loneliness scores were not found to significantly moderate changes in suicidal ideation or depression severity post intervention or at the 3-month follow-up in either condition. The severity or extent of loneliness at baseline also had no effect on how young adults self-engaged with the LifeBuoy app, contrary to prior research that suggests that lonelier individuals may lack self-efficacy to manage mental health issues [21,22].

These results are consistent with the existing literature examining the relationship between loneliness and mental health to the extent that they suggest that loneliness is associated with more complex psychopathology [8,9,11,12,34-36]. In this study, this relationship was evidenced by significant positive correlations between loneliness and suicidal ideation and depression scores. However, these findings go beyond confirming simple relationships between loneliness and mental health. This study shows that despite loneliness being associated with poorer mental health, it does not appear to be a characteristic or condition that moderates how young people engage with, nor benefit from, a digital health intervention. Though prior studies have posited that positive associations between depression and treatment dropout may be moderated by the psychological effort required to sustain treatment adherence [31,43], this study suggests that loneliness is not an independent nor significant factor that exacerbates this nonadherence, at least for brief smartphone interventions.

One potential explanation for the null association is that there was a high prevalence of moderate-severe levels of suicidal ideation and depression across the total sample, which may have obscured or dampened the effects of transdiagnostic conditions such as loneliness. Alternatively, the proportion of participants reporting clinically severe suicidal ideation (SIDAS score ≥21) may be important to consider in efforts to understand why loneliness did not emerge as a unique predictor of engagement nor benefit in this study. The LifeBuoy app was purposefully designed to improve ideation [25], and the trial was marketed to young people for whom ideation was currently or recently a salient concern. As such, LifeBuoy may have been seen as an intervention highly relevant to the needs of the young people who participated in the trial—and need is a strong intrinsic motivator for engagement [44]. Indeed, young people in both conditions engaged similarly well with the smartphone apps over the course of the trial [25]. Actual high levels of motivation or need to engage with the app may have overridden any cognitive barriers typically associated with loneliness. Examining the role of loneliness in a more clinically heterogenous or diverse sample of young adults may help to clarify the relationships between mental ill health, loneliness, and engagement with treatment. As this is the first study to examine the role of loneliness in how young people experiencing suicidal ideation benefit from a digital health intervention, caution must be taken against concluding that loneliness is not an important factor in understanding engagement with, and benefit from, self-guided treatment.

Overall, our findings show that participants engaged with and benefited from LifeBuoy, irrespective of loneliness severity, which shows that LifeBuoy, in its current form, can effectively engage and treat individuals despite how lonely they may be. This finding strengthens the potential for digital health interventions to become an established mental health treatment method for all young people experiencing suicidal ideation. These findings also suggest that a digital therapeutic such as LifeBuoy might be suitable for large-scale delivery during future pandemics or environmental crises when loneliness is expected to increase.

The study findings also add new knowledge about which person-specific characteristics or conditions may be important to target using digital health interventions to enhance adherence and clinical benefits. Loneliness, as an isolated factor, may not be important to uniquely target future digital health intervention development when samples or users present with complex psychopathology.

Although this study found no significant link between baseline loneliness and depression, previous research has shown that greater depression severity can hinder engagement with and benefit from digital health interventions [31,43]. Future studies could explore whether additional social factors such as social connectedness and social anxiety act to moderate the relationship between loneliness, suicidal ideation, and depression.

The study has several strengths. To our knowledge, this study is the first to examine whether loneliness—a known risk factor for suicidal ideation and depression—affects how young people engage with a digital health intervention. Accordingly, this study contributes new knowledge about the person-specific characteristics that may need to be considered when designing and delivering such interventions. The data for this study was derived from a well-designed randomized controlled trial of a therapeutic smartphone intervention involving 455 young adults and thus provided a large robust sample that supports the reliability of the current findings and conclusions.

There are some limitations to consider. As patient information was collected independently via a web-based survey, self-reporting bias could affect the findings via over- or underestimated effects. The TILS was used because it is a brief validated measure that can reduce survey length and minimize attrition; however, the limited number of items and the small score range (3-9) may not appropriately capture nuance or heterogeneity in loneliness. Future studies focusing primarily on loneliness should consider using more comprehensive scales. Given that loneliness was not a primary or major secondary outcome of the main trial examining the efficacy of LifeBuoy, measures that differentiate between social and emotional loneliness, such as the DeJong-Gierveld Loneliness Scale [45], were not considered and would have allowed for more nuanced analyses of the role of loneliness in digital health interventions. Information on current engagement with mental health treatment was not collected, although whether participants ever received mental health treatment was assessed, and app engagement did not differ on this variable in both intervention and control conditions. Future studies similar to this study should assess both current and past mental health treatment of participants and determine the potential impact of these variables on app engagement. Broader limitations relating to the primary randomized controlled trial have been discussed elsewhere [25].

Loneliness did not affect how young people engaged with or benefited from a digital health intervention (LifeBuoy) shown to reduce suicidal ideation in young adults. Despite a significant positive association between loneliness and depression levels, loneliness did not affect young people’s engagement with LifeBuoy nor moderate clinical outcomes. As the first study to explore how loneliness may affect young people’s engagement with and benefits from a targeted digital health intervention, this study has developed new insights into whom these self-guided interventions may work best for. Future studies should analyze the interplay of additional factors such as motivation, social connectedness, and social anxiety, as well as distinguish between social and emotional loneliness using more refined measures.