Self-Efficacy as a Candidate Mechanism of Change in Hybrid Digital Mental Health Care: An Analysis of Depression and Anxiety Outcomes Using Random-Intercept Cross-Lagged Modeling
Lim, K., Bodner, R., Cong, S., Nostu, H., & Torous, J.
Preprint.
Abstract
Background: Depressive and anxiety disorders are leading causes of global disability, yet workforce shortages and the limited reach of one-to-one therapy sustain a large treatment gap. Hybrid care models integrating clinician sessions, human-supported digital navigation, and smartphone-based monitoring offer a scalable alternative, but the mechanisms through which they reduce symptoms remain unclear.
Objective: To evaluate whether two hypothesized mechanisms, self-efficacy (SE) and emotional self-awareness (ESA), improve within an eight-week hybrid care program and whether their improvements precede and predict reductions in depression (PHQ-9) and anxiety (GAD-7).
Methods: Three cohorts of an eight-week hybrid care program yielded 301 intake-to-completion enrollment pairs from 300 participants. Assessments at intake (Week 0), interim (Week 4), and completion (Week 8) measured PHQ-9, GAD-7, self-efficacy (PROMIS SE-8a), and emotional self-awareness (single 5-point item). Missing data were handled by full information maximum likelihood (FIML). Analyses included paired Wilcoxon signed-rank tests with Cohen dz, repeated-measures analyses of variance, baseline-adjusted regressions, and bivariate random-intercept cross-lagged panel models (RI-CLPM).
Results: Symptoms decreased substantially (PHQ Δ = −3.04, dz = −0.58; GAD Δ = −3.09, dz = −0.64), and both resources rose (SE Δ = +1.38, dz = +0.63; ESA Δ = +0.82, dz = +0.73). ESA rose early and steadily, whereas SE rose only by Week 8. After adjustment for intake severity, ΔSE remained a robust predictor of both ΔPHQ (β = −0.26) and ΔGAD (β = −0.30), whereas ΔESA lost significance. The RI-CLPM provided the strongest temporal evidence: within-person increases in SE prospectively predicted later GAD reductions (β = −0.23, P = .002), with a significant reverse path indicating a bidirectional dynamic; SE → PHQ was directionally consistent but non-significant (β = −0.21, P = .087), and ESA showed no prospective effects.
Conclusions: Self-efficacy is the most credible candidate mechanism in hybrid care, showing temporal precedence for anxiety and a robust baseline-adjusted association with depression. Emotional self-awareness improves during care but does not appear to drive symptom change, supporting explicit self-efficacy enhancement in scalable hybrid interventions.