Efficacy of a Mobile App-Based Intervention for Young Adults With Anxiety Disorders: A Randomized Clinical Trial

Anxiety disorders among young adults (18-25 years) have risen rapidly, with past-month prevalence near 15%, contributing to reduced quality of life, academic and occupational impairment, and economic burden. Although cognitive behavioral therapy (CBT) is a gold-standard and preferred treatment, access barriers (stigma, cost, geography, limited CBT-trained clinicians) leave many with unmet needs. Digital, self-guided interventions can expand access but existing apps often omit key CBT components (notably exposure) and face low user engagement. Engagement may be enhanced by social support, behavioral economics, and gamification. The study’s primary aim was to test whether adding social support to gain-framed incentives improves anxiety outcomes versus gain- or loss-framed incentives alone when using a comprehensive self-guided CBT app (Maya). Secondary aims evaluated engagement and user satisfaction and whether these varied by incentive condition.

Prior work establishes CBT as an effective first-line treatment for anxiety and related disorders. Self-guided digital CBT has demonstrated efficacy for depression, with fewer studies in anxiety. Typical limitations of mental health apps include incomplete therapeutic coverage (particularly insufficient emphasis on behavioral strategies such as exposure) and poor user engagement. Engagement can be improved via social support and gamification/behavioral economics strategies (gain/loss framing). Meta-analytic evidence suggests digital mental health interventions often have modest effects and high attrition, underscoring the importance of comprehensive content and engagement design. The Maya app was developed to address these gaps by including a broad array of CBT skills with interactive, engaging features and optional incentive systems.

Design: Randomized clinical trial comparing three incentive conditions delivered alongside a self-guided CBT mobile app (Maya). Setting and recruitment: Online advertisements and outpatient psychiatry clinics at Weill Cornell Medicine. Enrollment occurred June 16, 2021–November 11, 2022; analyses December 21, 2022–June 14, 2024. Participants: Young adults aged 18–25 years with a primary anxiety disorder confirmed via ADIS-5 clinical severity rating ≥4; English fluent. Exclusions: Current participation in CBT, psychotropic medication dose change within prior 12 weeks, or suicidal ideation with intent and/or plan. Ethics and registration: IRB-approved; remote e-consent; CONSORT guidelines followed; ClinicalTrials.gov NCT05130281. Intervention: A 6-week program comprising 12 sessions (2/week) covering psychoeducation, quizzes, cognitive restructuring, exposure exercises, mindfulness, and other CBT skills. Each session included homework. The publicly available app mirrors core skills but with streamlined requirements and greater user choice. Incentive conditions (delivered via text outside the app using the Way to Health platform): (1) loss-framed points (start each week with points, lose points if sessions not completed on time); (2) gain-framed points (earn points for timely completion); (3) gain-social support (gain-framed points plus optional designation of a friend/family member to receive progress updates and provide support; if declined, a research assistant served as the support contact). Participants received virtual medals (bronze, silver, gold, platinum) based on points. Randomization and masking: Way to Health generated a simple 1:1:1 nonblocked random allocation. Research assistants enrolling participants were masked to assignment. Due to nonblocked randomization and reduced enrollment related to COVID-19, groups were uneven in size. Procedures: All study visits were remote via HIPAA-compliant Zoom. Baseline included diagnostic and symptom measures and app setup; outcomes were collected at week 3 (midpoint), week 6 (end of intervention), and week 12 (follow-up). Troubleshooting occurred at weeks 3 and 6. Outcomes: Primary—Change in Hamilton Anxiety Rating Scale (HAM-A) from baseline to week 6. Secondary—Anxiety Sensitivity Index (ASI) and Liebowitz Social Anxiety Scale (LSAS). Exploratory—Hamilton Depression Rating Scale (HAM-D). Engagement—total sessions completed; retention at weeks 6 and 12. Satisfaction—User Version of the Mobile Application Rating Scale (uMARS) App Quality score (assessed from week 1). Statistical analysis: Intent-to-treat including all with baseline data. Linear mixed-effects models (restricted maximum likelihood) with random intercept for subjects tested fixed effects of time (baseline, week 3, week 6, week 12), incentive condition (gain-social support, gain-framed, loss-framed), and time × condition interactions for HAM-A, ASI, and LSAS. Session completion across incentive conditions was compared with a Kruskal-Wallis test due to group size imbalance. uMARS analyzed via linear mixed-effects model with fixed effects for time and condition and their interaction. Post hoc paired t tests (Holm correction) examined change from baseline to each time point; effect sizes reported as Cohen d; significance threshold P < .05.

- Participants: 59 randomized (mean age 23.1 years; 78% female). Principal diagnoses: generalized anxiety disorder 56%, social anxiety disorder 41%. Mean baseline HAM-A 15.0 (SD 6.5). - Primary outcome (HAM-A): Significant reductions over time: midpoint vs baseline mean difference −3.20 (95% CI −4.76 to −1.64; d=0.64); end point vs baseline −5.64 (95% CI −7.23 to −4.05; d=0.94); follow-up vs baseline −5.67 (95% CI −7.29 to −4.04; d=1.04). No significant differences between incentive conditions (loss-framed vs gain-social support mean difference −1.40; 95% CI −4.72 to 1.93; gain-framed vs gain-social support 1.38; 95% CI −1.19 to 3.96) and no significant time × condition interactions. - Secondary outcomes: ASI decreased at midpoint −3.69 (95% CI −6.22 to −1.17; d=0.49), end point −9.51 (95% CI −12.08 to −6.94; d=0.93), follow-up −8.90 (95% CI −11.54 to −6.27; d=0.93). LSAS decreased at midpoint −6.84 (95% CI −11.46 to −2.21; d=0.47), end point −12.67 (95% CI −17.38 to −7.96; d=0.70), follow-up −17.61 (95% CI −22.45 to −12.77; d=1.07). No significant main effects of incentive or interactions for ASI or LSAS. - Engagement/retention: 98% completed week 6, 93% completed week 12. Mean sessions completed 10.8 (SD 2.1; 95% CI 10.3–11.4); 64% completed all 12 sessions. Of those with follow-up, 39% reported continued app use beyond 6 weeks. Session completion did not differ by age, gender, race, or incentive condition. - Satisfaction: uMARS App Quality mean scores exceeded acceptability threshold (≥4) at all time points: week 1 mean 4.09 (68% >4); week 3 mean 4.17 (68% >4); end point mean 4.18 (71% >4); follow-up mean 4.22 (68% >4). No differences by incentive condition or over time. - Safety: No serious adverse events reported.

The study did not find differential efficacy across incentive strategies; anxiety improvements were comparable whether participants received loss-framed points, gain-framed points, or gain plus social support. Nonetheless, the app produced clinically meaningful and statistically significant reductions in anxiety that persisted through 12-week follow-up. High retention, strong engagement, and acceptable-to-good app satisfaction were observed across all conditions. Lack of between-group differences may reflect limited power due to smaller, imbalanced groups and/or a ceiling effect from the app’s built-in engagement features combined with effective gain/loss framing across all conditions. The magnitude of anxiety reduction (moderate to large effect sizes) compares favorably with typical digital mental health interventions and approaches effects seen in pharmacologic trials. Reductions in anxiety sensitivity and social anxiety suggest the comprehensive CBT content may target specific symptom clusters. The app’s scalability (free, self-guided, non-monetary incentives) supports potential for broader dissemination to address access barriers in young adult mental health care.

A comprehensive, self-guided mobile CBT app (Maya) was efficacious and feasible for young adults with anxiety disorders, with large improvements in anxiety that did not differ by incentive condition. The app’s design—covering a full suite of CBT skills and incorporating engagement features—addresses common limitations of digital mental health tools. Digital interventions like Maya may help bridge gaps in access to evidence-based care amid clinician shortages and logistical barriers. Future research should include adequately powered trials with balanced allocation, inclusion of control conditions (e.g., app without incentives or waitlist), evaluation in more diverse populations, and investigation of tailored module sets targeting specific symptom profiles to optimize effectiveness and engagement.

- Reduced and imbalanced sample sizes across incentive groups due to pandemic-related disruptions and nonblocked randomization, limiting power to detect between-condition differences. - Absence of a control group (e.g., no-app or app-only without incentives) prevents disentangling intervention effects from regression to the mean or placebo-like effects. - Predominantly female, college-educated, Asian or White, and non-Hispanic sample limits generalizability to more diverse populations. - Pandemic-related modifications could have influenced recruitment and procedures.