Brief Digital Interventions for Psychological Distress: An AI-Enhanced Response-Adaptive Randomized Clinical Trial

Importance A large proportion of college students report experiencing psychological distress. Smartphone app–based interventions may alleviate distress, but their effectiveness across severity levels is unclear. Artificial intelligence (AI)–enhanced response-adaptive randomized clinical trials may offer an efficient method to evaluate competing interventions. Objective To compare the effectiveness of 3 brief, 2-week self-guided smartphone application–based interventions (physical activity, mindfulness, sleep hygiene) or an active control (ecological momentary assessment [EMA]) for reducing psychological distress among college students with mild, moderate, or severe distress. Design, Setting, and Participants This population-based AI-enhanced response-adaptive randomized clinical trial included 1282 participants with distress scores of 20 or more on the 10-item Kessler Psychological Distress Scale in 12 minitrials from November 9, 2021, with final follow-up on February 17, 2023. Participants’ distress was categorized as mild, moderate, or severe based on their normalized 21-item Depression Anxiety Stress Scale (DASS-21) scores at screening. Interventions After a 2-week onboarding period of daily EMA, participants were assigned by a contextual multi-armed bandit algorithm to 1 of 4 two-week self-guided app interventions: physical activity, mindfulness, sleep hygiene, or a control that continued EMA. Main Outcomes and Measures The primary outcome was change in psychological distress (DASS-21 total score) from week 2 (before intervention) to week 4 (after intervention). The primary end point was after the intervention (4 weeks). Secondary outcomes included DASS-21 subscale scores, self-reported physical activity, mindfulness, sleep quality, and app engagement, usability, and satisfaction. Analysis was performed on an intention-to-treat basis. Results A total of 1282 individuals (mean [SD] age, 23.5 [5.2] years; 950 women [74.1%]) participated: physical activity (n = 305), mindfulness (n = 453), sleep hygiene (n = 431), or a control that continued EMA (n = 93). Among 349 participants with severe distress, physical activity (n = 79) and mindfulness (n = 180) were significantly more effective than the EMA control (n = 29) in reducing DASS-21 total scores (physical activity vs control: standardized mean difference [SMD], 0.62 [95% CI, 0.23-1.02]; mindfulness vs control: SMD, 0.53 [95% CI, 0.19-0.87]) and sleep hygiene (n = 61) (physical activity vs sleep hygiene: SMD, 0.50 [95% CI, 0.16-0.84]; mindfulness vs sleep hygiene: SMD, 0.41 [95% CI, 0.13-0.69]). Among 494 participants with mild distress, physical activity (n = 161) and sleep hygiene (n = 224) were significantly more effective than control (n = 37) in reducing DASS-21 total scores (physical activity vs control: SMD, 0.58 [95% CI, 0.30-0.86]; sleep hygiene vs control: SMD, 0.47 [95% CI, 0.20-0.73]). No significant group differences were observed among participants with moderate distress (n = 439). Conclusions and Relevance In this AI-enhanced response-adaptive randomized clinical trial among college students, physical activity and mindfulness were most effective for severe distress, while physical activity and sleep hygiene were most effective for mild distress. These findings can guide personalized mental health interventions for college students. This trial improved efficiency by minimizing control group allocation but had reduced power to detect significant group differences. Trial Registration http://anzctr.org.au Identifier: ACTRN12621001223820