Smartphone App to Address Loneliness Among College Students: Pilot Randomized Controlled Trial

Background Loneliness is a widespread and significant problem on college campuses. Prolonged loneliness in young adulthood is a risk factor for concurrent and future mental health problems and attrition, making college a critical time for support. Cognitive and behavioral interventions show promise for decreasing loneliness and can be widely disseminated through technology. Objective This pilot randomized controlled trial was conducted to examine the initial efficacy, feasibility, and desirability of a smartphone app, Nod, designed to deliver cognitive and behavioral skill-building exercises to reduce loneliness during the transition to college. Methods First-year college students (N=221, mean age 18.7 years, 59% female) were recruited online during incoming student orientation, and randomized to either receive immediate access to Nod (experimental group, n=100) or access after 4 weeks (control group, n=121). The app delivered skills via fully automated (1) “social challenges,” suggested activities designed to build social connections; (2) reflections, brief cognitive reframing exercises; and (3) student testimonials that encouraged a growth mindset toward social connection building. Main intention-to-treat analyses were used to compare the conditions on self-assessed loneliness, depressive symptoms, and other mental health and college adjustment outcomes at week 4, controlling for baseline values on those variables. Analyses were also performed to test the hypothesis that the treatment benefits would be particularly pronounced for participants with heightened psychological vulnerability at baseline (ie, higher baseline depressive symptoms and loneliness). Results Retention was 97% at week 4, and participants viewed an average 36.7 pages of app content. There were no significant condition differences in loneliness at week 4 (F1, 211=0.05, P=.82; ηp2 <.001). However, there was a significant condition-by-baseline depression interaction to predict week-4 loneliness (F1,209=9.65, P=.002; ηp2 =.04). Simple slope analyses indicated that baseline depression positively predicted week-4 loneliness among control participants (r=0.30, t209=3.81, P<.001), but not among experimental participants (r=–0.09, t209=–0.84, P=.40), suggesting that Nod buffered participants with high baseline depression scores from experiencing heightened midquarter loneliness. Similarly, there were no significant condition differences in other week-4 outcomes. However, moderation by baseline vulnerability was found for week-4 depressive symptoms, sleep quality, and indices of college adjustment (eg, perceived social support and campus belonging). Conclusions Although Nod exposure did not impact outcomes for the full sample, these results provide initial evidence of its benefit for vulnerable students. The results of this trial suggest that cognitive and behavioral skills delivered via a mobile app can buffer psychologically vulnerable college students against heightened loneliness and depressive symptoms, as well as other negative college adjustment outcomes. Future work will aim to improve upon app engagement, and to address loneliness among other key populations. Trial Registration ClinicalTrials.gov NCT04164654; https://clinicaltrials.gov/ct2/show/NCT04164654


Appendix 5. Comparisons between outcomes and engagement in the control group at week 8 to the experimental group at week 4 Data Analytic Strategy
To validate results, we compared loneliness, mental health, and college adjustment outcomes in the control group at week 8 to those of the experimental group at week 4. Specifically, we analyzed whether there were any main effects of condition, or any condition x baseline vulnerability interactions to predict outcomes after each group had been offered Nod for 4 weeks. These analyses allowed us to explore whether the benefits of Nod were similar irrespective of whether Nod was delivered at the start of participants' first year of college, or four weeks into their first year.
Analyses mirrored those described in the 'data analytic strategy' section of the manuscript, with the exception that we controlled for week 4 scores when predicting week 8 scores in the control group, but controlled for week 0 scores when predicting week 4 scores in the experimental group. This was done to ensure that baseline scores for each group were measured four weeks prior to evaluating outcomes.

Results
Descriptive statistics by condition across the three time points are reported in Table 1 below.

Loneliness
Analyses in Step 1 showed no evidence for an overall effect of condition on loneliness at the end of four weeks of Nod use, F1,210=2.47, P=.12. This finding indicates that, on average, students randomly assigned to use Nod four weeks into their first year of college had equivalent outcomes to those randomly assigned to use Nod at the beginning of their first year.
However, participants' depressive symptoms immediately before being assigned to use Nod interacted significantly with condition to predict loneliness at the end of four weeks (F1, 208=6.93, P=.009). Examination of simple slopes suggest that Nod was more effective at buffering students with elevated depression against heightened future loneliness when Nod was delivered at the start of the school year than when delivered four weeks into the school year (Table 2; Figure 1).

Mental Health Indicators
Analyses in Step 1 showed no evidence for an overall effect of condition on depression symptoms, anxiety symptoms, or sleep quality at the end of four weeks of Nod use, all Fs <2.0, Ps >.15. However, adjusting for baseline social anxiety symptoms, the experimental group reported significantly lower social anxiety symptoms after four weeks of Nod use than the control, F1,209=5.84, P=.02. In Step 2, loneliness did not significantly moderate the effect of condition on any of the mental health outcomes (all Fs <2.60, Ps >.11).

College Adjustment Indicators
Analyses in Step 1 showed no evidence for an overall effect of condition on perceived social support, campus belonging, or social adjustment after four weeks of Nod use, all Fs <1.3, Ps >.25. Nor were there condition differences in intention to return to college in the subsequent school year (OR=.86, 95% CI [.38-1.95], z=-0.36, P=.72).
However, participants' self-reported loneliness before being assigned Nod significantly interacted with condition to predict social support (F1, 208=4.50, P=.035), campus belonging (F1, 207=4.49, P=.035), and social adjustment (F1, 208=5.85, P=.016) at the end of four weeks of Nod use. Examination of simple slopes suggested that, in all cases, the effects of Nod on college adjustment indicators were more pronounced for lonelier students when Nod was delivered at the start of the school year (Table 2; Figure 1). Likewise, the significant main effect of condition on intention to return was moderated by a condition x baseline loneliness interaction (OR=1.33, 95% CI [1.11-1.59], z=3.07, P=.002). Probing of this interaction revealed that within the control group, the odds of "definitely" intending to return to campus decreased as baseline loneliness increased, even after receipt of Nod. In contrast, in the experimental group the odds of intending to return significantly increased as baseline loneliness increased (Table 2).

User Experience
There were no significant differences between the experimental and control groups in the percentage of participants that agreed that Nod was easy to understand (X 2 1,210=0.22, P=.64), gave them sound advice (X 2 1,210 =1.07, P=.30), or something new to think about (X 2 1,210=0.17, P=.68). Nor were there differences in desire to continue to use Nod (X 2 1,210=2.21, P=.14), or agreement that they had used what they learned in daily life (X 2 1,210=.32, P=.57) (Figure 2).