Toward Optimized Mobile EMI for Managing Mild Depression and Anxiety in University Students

Abstract

This study investigates the efficacy of Ecological Momentary Intervention (EMI) and the determi- nants of user engagement among university students, a population increasingly vulnerable to mental health challenges such as depression and anxiety. Given the limitations of static mobile interventions in addressing the fluctuating psychological states of users, we conducted a six week Micro-Randomized Trial (MRT) involving 215 students with mild to moderate symptoms. Analysis using Generalized Estimating Equations (GEE) revealed that EMI exerted a significant proximal causal effect on reducing stress levels the following day (β = −0.080, p = .020). While im- mediate effects on anxiety were not statistically significant, a cumulative learning effect was observed, where the reduction in anxiety became more pronounced as the number of interventions increased (β =−0.056, p = .008). Furthermore, to understand engagement dynamics, a two-step residual analysis integrating Generalized Additive Mixed Models (GAMM) and LightGBM was employed. The results indicated that even after controlling for stable individual traits, within-person contextual factors, includ- ing specific delivery times (lunch and dinner), previous day engagement (inertia), and recent symptom variability (high fluctuations in stress and anxiety), significantly predicted intervention adherence. These findings provide robust empirical evidence that EMI is an effective tool for immediate stress alleviation and long-term anxiety management. Moreover, the dynamic nature of engagement highlights the necessity of transitioning from traditional, uniform delivery methods to Just-In-Time Adaptive In- terventions (JITAI) that optimize intervention timing and content based on a users real-time state and context.