Symposium
Artificial Intelligence and Technology-based Interventions
Nur Hani Zainal, M.S., Ph.D. (she/her/hers)
Assistant Professor
National University of Singapore
Singapore, Singapore
Daniel Eisenberg, Ph.D.
Professor
UCLA
Los Angeles, California, United States
Ellen E. Fitzsimmons-Craft, Ph.D.
Associate Professor of Psychological & Brain Sciences and Psychiatry
Washington University in St. Louis
St. Louis, Missouri, United States
Elsa Rojas-Ashe, Ph.D. (she/her/hers)
Clinical Associate Professor
Stanford University
Portland, Oregon, United States
Craig Barr Taylor, M.D.
Professor
Stanford University
Palo Alto, California, United States
Denise E. Wilfley, Ph.D.
Scott Rudolph University Professor of Psychiatry
Washington University School of Medicine
St. Louis, Missouri, United States
Michelle G. Newman, B.S., M.A., Ph.D. (she/her/hers)
Professor
The Pennsylvania State University
State College, PA, United States
Background: Scalable digital-guided cognitive-behavioral therapy (d-GCBT) addresses the access gap in mental health care. Yet economic value depends on cost timing, outcomes prioritized, and decision thresholds. This study evaluated coach-supported d-GCBT (SilverCloud) versus treatment as usual (TAU) from a societal perspective, using randomized trial data from 6,205 U.S. college students.
Method: TAU (n=3,102) and d-GCBT (n=3,103) participants were followed at 6 weeks, 6 months, and 2 years. Analyses incorporated healthcare costs (therapy, medications, emergency care, hospitalizations), intervention costs (platform, coaching, supervision), and work productivity losses (2025 U.S. dollars). Effectiveness was measured as: (a) proportion achieving no probable mental disorder at each time point, yielding incremental cost-effectiveness ratios (ICERs) per percentage-point improvement; and (b) 2-year quality-adjusted life years (QALYs), reduced 3% annually for the time value of money.
Results: At 6 weeks, d-GCBT produced 4.2 percentage-point improvement in disorder-free status with incremental costs of $992, yielding ICER=$23,393 per percentage-point. By 6 months, incremental costs fell to $8 (control group received mental health care) while gains persisted (5.1 percentage points), yielding an ICER of $160 per percentage point. At 2 years, incremental costs were $124 with sustained benefit (3.9 percentage points), yielding ICER=$3,176. In 2-year cost-utility analysis, d-GCBT produced 0.018 additional discounted QALYs per person (1.513 vs. 1.495; 95% CI [0.011,0.025])–one additional month at full health–at incremental cost of $853 (95% CI [$361,$1,359]), yielding ICER=$46,823/QALY. Probability of cost-effectiveness was 53.7% at $50,000/QALY and 98.3% at $100,000/QALY. Participants who received d-GCBT gained 1.00 disorder-free month over 2 years (10.54 vs. 9.54) at incremental cost of $583.
Discussion: d-GCBT demonstrated durable improvements with cost-effectiveness strengthening over time. Early cost increases reflected concentrated intervention delivery; longer-horizon estimates converged favorably. Economic value depended on the chosen willingness-to-pay thresholds and outcome metrics. At standard policy thresholds (≥$50,000/QALY), d-GCBT demonstrated robust cost-effectiveness. Findings support d-GCBT as an efficient treatment, adding value over TAU when sustained symptom-free time and population reach are valued.