Construction and effect evaluation of a precise intervention model for student mental health empowered by big data

Abstract Background The prevalence of mental health issues among adolescents continues to rise, significantly impairing academic performance and overall well-being. While big data technologies—characterized by multi-source data integration, precise risk identification, and dynamic tracking—have demonstrated potential in public health and educational management, their application in student mental health lacks systematic quantitative evaluation. To establish a scientific, efficient, and precise intervention framework that enables early risk detection, personalized support, and real-time outcome monitoring, a big data–enabled precision intervention model is proposed. By integrating multi-source data such as psychological questionnaires and academic performance records, the study quantitatively assesses the model’s effectiveness in improving students’ mental health, offering a practical and implementable solution for school-based mental health education. Methods A total of 120 students aged 13–18 from four secondary schools (Symptom Checklist-90 SCL-90 total score ≥ 160; excluding severe psychiatric or physical disorders) were randomly assigned to an intervention group (n = 60) or a control group (n = 60). Groups showed no significant differences in age, gender, or baseline psychological measures (p.05). The intervention group received a big data–enabled precision intervention: multi-source data (psychological assessments, campus smart-card logs, academic platforms) were collected and analyzed via machine learning to build a risk-prediction model, which classified students into risk tiers and triggered tailored interventions—low-risk: online psychoeducation; medium-risk: weekly one-on-one counseling; high-risk: plus group activities. The control group received standard care (two semesterly group lectures and on-demand counseling). Assessments using SCL-90, Self-Rating Anxiety Scale (SAS), Self-Rating Depression Scale (SDS), and Resilience Scale (RS) were conducted pre-intervention, at 3 and 6 months, and at 3-month follow-up. Data were analyzed using repeated-measures ANOVA, independent-samples t-tests, and chi-square tests (SPSS 26.0); p.05 indicated statistical significance. Results Within-group analyses showed that the intervention group exhibited significant reductions in SCL-90, SAS, and SDS scores (p.05) and increased RS scores by month 3; improvements further deepened by month 6 and remained stable at follow-up. The control group showed only a minor SCL-90 reduction at month 6, with no other significant changes. Between-group comparisons at month 6 revealed significantly lower SCL-90 (132.5 ± 18.6 vs. 158.3 ± 20.1; t = 8.76, p.001), SAS (37.2 ± 4.5 vs. 45.9 ± 4.7; t = 9.12, p.001), and SDS scores (38.6 ± 4.8 vs. 47.3 ± 5.1; t = 8.95, p.001), and higher RS scores (65.8 ± 5.3 vs. 56.2 ± 5.6; t = 8.53, p.001) in the intervention group. Clinical response rate was significantly higher (83.3% vs. 46.7%; χ2 = 22.89, p.001). Additionally, the big data–based risk prediction achieved 89.2% accuracy, markedly outperforming traditional manual screening (62.5%). Discussion The big data–enabled precision intervention model demonstrates significant efficacy in improving adolescent mental health, surpassing conventional approaches. Its core advantages lie in accurate risk identification through multi-source data integration, personalized interventions aligned with individual needs, and adaptive strategy adjustment via continuous monitoring—collectively enhancing intervention efficiency. The model integrates seamlessly into existing school management systems and reduces reliance on human resources, providing a scalable foundation for intelligent, evidence-based mental health services in education. Future work may extend the approach to primary and higher education settings and incorporate neurophysiological (e.g., EEG) or behavioral monitoring to deepen mechanistic understanding, facilitating standardization and large-scale deployment.