Date
5-20-2026
Degree
Master of Science in Public Health in Epidemiology (MSPH)
Chair
Oswald Attin, Linnaya Graf
Keywords
academic burnout, depression, MBI-GS(S), CES-D, college students, mental health, machine learning, random forest, KMeans clustering, Liberty University
Disciplines
Public Health
Recommended Citation
Song, Hyejoon, "Predictive Models of Student Mental Health: The Correlation Between Academic Burnout and Depression in the Higher Education Environment" (2026). Masters Theses. 1486.
https://digitalcommons.liberty.edu/masters/1486
Abstract
Importance: Over 80% of U.S. college students report significant emotional difficulties, and 58% endorse academic burnout as a persistent concern, yet institution-specific data at faith-integrated universities such as Liberty University remain scarce, limiting evidence-based programmatic response.
Objective: To examine the correlation between academic burnout and depressive symptomatology at Liberty University, identify the most predictive individual burnout item via Random Forest machine learning, and characterize distinct student risk profiles via KMeans clustering. Design, Setting, Participants: Cross-sectional survey (IRB-FY24-25-1491, Exempt) of 236 Liberty University students (mean age 36.9 ± 14.3 years; 84.0% female; 61.0% graduate-level), 2025–2026 academic year.
Main Outcome and Measures: Burnout assessed with the MBI-GS(S) (α = 0.953); depression with the CES-D (clinical threshold ≥ 16; α = 0.934). Analyses: Pearson's r, hierarchical regression, Random Forest feature importance, KMeans clustering. Results: Burnout correlated significantly with depression (r = 0.625, 95% CI [0.541, 0.697], p < 0.001); subscale correlations ranged from r = 0.520 (Cynicism) to r = 0.608 (Emotional Exhaustion). Burnout subscales explained an additional 34.2% of CES-D variance beyond demographics (total R² = 0.422, F(5,232) = 33.19, p < 0.001). Random Forest identified "I feel less confident in my ability to achieve academic goals" as the top predictor (importance ≈ 0.21). Three risk profiles emerged: Low-Risk (n = 99; CES-D M = 10.1), Moderate-Risk (n = 83; 79.5% exceeding clinical threshold), and High-Risk (n = 56; 100% exceeding threshold); overall, 58.0% met the clinical cutoff. Conclusion: Academic burnout—particularly loss of academic self-efficacy—is a significant predictor of depression at Liberty University. The three-cluster architecture supports tiered intervention beginning at moderate burnout, and findings call for proactive institution-specific screening consistent with Liberty University's mission and the Biblical imperative to care for those who are weary and burdened (Matthew 11:28).
