Psychological distress affects over one in five college students, which has an impact on their well-being, academics, and relationships. Scalable mental health support is provided for these students using brief smartphone app-based programs. These programs focus on exercise, sleep hygiene, and mindfulness. Mindfulness improves self-compassion, attention, and physical activity enhances fitness and confidence, and sleep hygiene promotes high-quality sleep. However, few studies directly compare their effectiveness across various distress levels, including mild, moderate, and severe. Artificial intelligence (AI)-enhanced, bandit-based response-adaptive studies provide an efficient alternative to traditional randomized controlled trials (RCTs)Â by dynamically allocating participants to better-performing interventions based on ongoing data.
This was well explained in a recent study published in JAMA Netw Open, which aimed to compare the effectiveness of three app-based interventions and assess an AI-adaptive trial in college students with severe, moderate, and mild distress.
This four-arm AI-enhanced response-adaptive RCT was conducted between 2021 and 2023 at the University of New South Wales, which involved 12 four-week minitrials. Students aged ≥18 years and with a distress score of ≥20 were recruited through traditional and social media platforms, as well as providing electronic consent. Participants with prior trial participation, psychosis, high suicidal ideation, bipolar disorder, and taking concurrent medications were excluded from this study.
An AI-driven multi-armed bandit algorithm dynamically allocated participants based on their distress severity, such as severe, mild, and moderate. All interventions were standardized, accessible, brief, and self-guided. The primary outcome was post-intervention psychological distress, which was estimated using the 21-item depression anxiety stress scale (DASS-21; 0-126 range, higher score means higher distress). The secondary outcomes were sleep (modified Pittsburgh sleep quality index), anxiety, activity (physical activity vital sign [PAVS]), depression (DASS-21), mindfulness (bespoke single item), and stress. All statistical analyses were performed by R, version 2023.09.0+463.
A total of 1282 students (mean age = 23.5±5.2 years, female = 74.1%, male = 21.8%, undergraduate students = 78.3%, Australian = 94.2%) were included in this study. Participants were divided into four subgroups: sleep hygiene (n = 431), mindfulness (n = 453), and physical activity (n = 305), as well as a control or ecological momentary assessment group (EMA, control, n = 93).
Among participants with mild distress (n = 494), sleep hygiene (n = 224), and physical activity (n = 161) significantly reduced DASS-21 total scores compared to control group (n = 37) with standardized mean differences (SMD) of 0.47 (95% CI, 0.20-0.73) and 0.58 (95%CI, 0.30-0.86), respectively. Similar results were observed in participants with severe distress in reducing DASS-21 with SMD of 0.62 (95% CI, 0.23-1.02) for physical activity vs control, 0.53 (95% CI, 0.19-0.87) for mindfulness vs control, 0.41 (95% CI, 0.13-0.69) for mindfulness vs sleep hygiene, and 0.50 (95% CI, 0.16-0.84) for physical activity vs sleep hygiene. Whereas there were no statistically significant differences in DASS-21 scores observed among students with moderate distress.
In mild distress participants, physical activity lowered anxiety more than mindfulness, with an SMD of 0.35 (95% CI, 0.04-0.66). All active interventions reduced the stress scores compared to controls with respect to sleep hygiene (SMD = 0.47 [95% CI, 0.20-0.74]), mindfulness (SMD = 0.57 [95% CI, 0.23-0.90]), and physical activity (SMD = 0.65 [95% CI, 0.36-0.93]). Change in sleep quality did not differ between trial arms in any severity subgroups. Physical activity scores were higher in students with severe distress compared to sleep hygiene (SMD = 0.56 [95% CI, 0.20-0.92]) and mindfulness (SMD = 0.32 [95% CI, 0.07-0.57]).
This study’s limitations include brief interventions, lack of blinding, self-reported short-term outcomes, limited generalizability due to mainly females and English-speaking students, and potential non-applicability of app-based results to in-person interventions.
In conclusion, physical activity most effectively reduced distress in individuals with severe and mild symptoms, sleep hygiene helped mild distress, and mindfulness aided severe distress. This demonstrates potential and feasibility for scalable personalized interventions.
Reference: Newby J, Gupta S, Hoon L, et al. Brief Digital Interventions for Psychological Distress: An AI-Enhanced Response-Adaptive Randomized Clinical Trial. JAMA Netw Open. 2025;8(10):e2540502. doi:10.1001/jamanetworkopen.2025.40502
