Abstract
Introduction: Early identification and intervention for individuals at elevated risk for mental disorders is critical for improving quality of life and reducing social costs. Conventional self-report screening tools, however, are susceptible to social desirability and recall biases. Therefore, this study explored the feasibility and validity of smartphone app-based neuropsychological tasks designed to complement self-report measures and assist in the early screening of individuals with subclinical depression and anxiety. Methods: A subclinical depression/anxiety group (n = 55) and a control group (n = 57), aged 19–50 years (mean age = 36.14 ± 8.34), completed app-based neuropsychological tasks. Criterion-related validity was assessed using Pearson correlations between task scores and self-report scales measuring depression, anxiety, self-esteem, negative rumination, anxiety sensitivity, and distress intolerance. Discriminant validity was evaluated by conducting independent sample t-tests. Finally, discriminant analysis was performed using task variables that significantly differed between groups to evaluate classification accuracy. Results: Several index scores from the app-based tasks were significantly correlated with depression, anxiety, and related self-report measures. Mean differences between the subclinical and control groups were also significant. Discriminant analysis using auditory working memory, abandonment tendency, motivational deficit, and reasoning accuracy scores from the app-based tasks yielded a classification accuracy of 70.5% (leave-one-out cross-validation = 67.0%). When both neuropsychological task scores and depression- and anxiety-related self-report measures were included as independent variables, the classification accuracy increased to 91.1%. Discussion: The findings suggest that app-based neuropsychological tasks may serve as a promising adjunctive tool for early screening of individuals with subclinical depression and anxiety, addressing limitations associated with self-report measures.

This publication has 92 references indexed in Scilit: