Application of big data analysis in mental health risk early warning education for nursing students

Abstract Background Nursing students, as a key part of the healthcare system, face multiple pressures including academic demands, clinical practice challenges, and professional adaptation difficulties. The prevalence of mental health problems among them is higher than in general students. Studies show that about 25–35% of nursing students experience anxiety, depression, or burnout, which affects learning outcomes and future career development. Traditional mental health screening relies on periodic surveys and self-reporting, which suffer from poor timeliness, limited coverage, and delayed warnings. With the development of information technology, big data analysis offers a new approach to mental health management. By integrating academic performance, behavioral patterns, and social interaction data, multidimensional predictive models can support dynamic monitoring and early warning. This study aims to develop a big data-based early warning system for nursing students’ mental health risks and evaluate its effectiveness in educational practice. Methods The study included 1200 full-time undergraduate nursing students from the 2020–2022 cohorts and was conducted in two stages: system development and effectiveness evaluation. The system integrated five data sources: academic, behavioral, social, physiological, and self-reported data. Machine learning algorithms built a predictive model that classified students into low-, medium-, and high-risk groups. In the evaluation stage, students were randomly assigned to the intervention group (n = 600) and control group (n = 600). The intervention group received risk-based interventions: low-risk students received mental health education notifications, medium-risk students received one-on-one counseling, and high-risk students underwent crisis intervention including emergency assessment, parental contact, and medical referral. The control group received traditional semester surveys and voluntary support services. Assessments occurred at baseline and at 6, 12, and 24 months, including the Self-Rating Depression Scale (SDS), Self-Rating Anxiety Scale (SAS), and Academic Burnout Scale. Psychological crises, counseling service use, and academic performance were also recorded. Results The predictive model achieved 82.7% accuracy, 85.3% sensitivity, and 81.4% specificity. At 24 months, the intervention group had significantly lower SDS scores than the control group (38.6 ± 8.7 vs. 45.3 ± 10.2, p.001, d = 0.71) and SAS scores (36.2 ± 7.9 vs. 42.8 ± 9.6, p.001, d = 0.75). All three dimensions of academic burnout were significantly improved in the intervention group (p.01, d = 0.58–0.82). The intervention group also showed higher counseling utilization (32.5% vs. 18.7%, p.01). Academic outcomes, including course pass rates and clinical practice scores, were significantly better than the control group (p.05). Early warning analysis indicated that 73.2% of high-risk students were identified 1–2 months before symptom onset, enabling timely intervention. Discussion This study demonstrates the value of big data analysis in early warning education for nursing students’ mental health. Multidimensional predictive models effectively identify mental health risks and, combined with targeted interventions, improve psychological well-being and academic performance. The approach transforms passive periodic screening into proactive, dynamic monitoring, capturing early warning signals that traditional surveys cannot detect, enhancing precision and coverage of mental health services. The model reduces the risk of crisis escalation and optimizes resource allocation. Future research should integrate additional data sources, such as wearable device sleep and activity data, to improve predictive accuracy and develop personalized intervention algorithms matching students’ risk profiles with appropriate services. Funding No. ZQZR202443.