Smartphone App Predicts Adolescent Mental Health Risk
Researchers at Imperial College London conducted a 14-day study (n=103) using the Mindcraft smartphone app to collect daily self-reports and continuous passive sensor data from school-going adolescents. A contrastive pretraining deep learning model integrating active and passive streams predicted SDQ-high risk, insomnia, suicidal ideation, and eating disorder with balanced accuracies 0.71, 0.67, 0.77, and 0.70; external validation (n=45) achieved 0.63–0.72. Findings indicate feasibility for scalable, school-based early detection.
Key Points
- 1Achieved multimodal predictions with balanced accuracies 0.67–0.77 across four adolescent mental health outcomes
- 2Demonstrated contrastive pretraining improves representation stability and predictive robustness in noisy behavioral data
- 3Suggests integrating subjective and sensor streams enables scalable early detection and school-based screening strategies
Scoring Rationale
Solid peer-reviewed multimodal study with external validation, but moderate novelty and limited sample constrain generalizability.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
