Understanding Engagement in Mental Health App Trials: Developing Prognostic Multivariate Regression Prediction Models of App Uptake and Attrition in General Adult and Distressed Student Populations Using Baseline Predictors

Digital mental health interventions (online, mobile, game-based) can reduce psychological difficulties, but attrition, especially in mobile apps, remains high. Personalizing treatment using predictors of uptake and dropout may improve engagement. We developed prognostic models of app uptake and study attrition using baseline demographic, psychological, and app usage data from two samples: distressed college students (N = 322) and adults from the general population (N = 450), using two distinct apps, the PsyPills and OCAT apps. Multivariate logistic regressions showed consistent patterns across studies: higher usage frequency predicted lower attrition, while male gender predicted lower uptake. Study-specific effects also emerged: higher depression predicted lower uptake in students, whereas symptoms did not predict uptake in the general population. Overall, our results showed that engagement is shaped by both individual and intervention-level factors, underscoring the need for personalized, ecologically responsive interventions.