![[ Visit Client Website ]](images/banner.gif)
College student mental health is an urgent social work concern, yet digital mental health research often reduces internet use to screen time or problematic use, overlooking differences in online experiences and contexts. Guided by the I-PACE model and a socio-ecological framework, this study conceptualized online harm, relational use, creative use, and civic engagement as distinct digital experiences associated with depression and anxiety. The study addressed two gaps: limited evidence on how specific online experiences relate to college student mental health, and limited integration of institutional and regional context in de-identified student data. The purpose was to examine whether internet use experiences are associated with depression/anxiety and whether those relationships vary across institutional and regional contexts.
Methods
Data came from the 2023–2024 Healthy Minds Study. The original dataset included 104,729 students across 196 institutions. The primary analytic sample included 43,331 students with complete data on outcomes and core predictors. Institution-context analyses included 45,358 students nested within 195 schools, and weighted regional analyses included 45,385 students across nine U.S. Census divisions. The outcome was a binary indicator of depression or anxiety. Predictors included nonacademic internet time, relational use, online harm, creative use, and civic engagement, with loneliness, demographics, and race/ethnicity controls. Logistic Regression, Random Forest, and LightGBM were compared. Because school coordinates were unavailable, regional context was modeled using institution geography as a spatial proxy.
Results
In the primary analytic sample (n = 43,331), 47.8% screened positive for depression or anxiety. Model performance was similar across approaches, with Logistic Regression (AUC = 0.7927), Random Forest (AUC = 0.7884), and LightGBM (AUC = 0.7908). Loneliness was the strongest correlate (OR = 2.759). Among internet-related factors, online harm was the clearest risk factor (OR = 1.089), while relational use showed a modest protective association (OR = 0.916). Students reporting online harm had higher depression/anxiety prevalence than those without such experiences (68.7% vs. 44.7%; weighted: 69.1% vs. 43.7%). Adding institutional context produced small but consistent improvements (n = 45,358), with Logistic Regression AUC increasing from 0.7874 to 0.7884 and LightGBM from 0.7872 to 0.7899. In weighted regional analyses (n = 45,385), depression/anxiety was highest in Mid-Atlantic (0.4932), East South Central (0.4908), and West South Central (0.4896), and lowest in East North Central (0.4361). The online harm risk gap varied across regions, ranging from 0.2122 in Mid-Atlantic to 0.3166 in New England.
Conclusion
Findings suggest that college student mental health is shaped less by internet exposure alone than by the quality of online experiences, especially harmful interactions and social disconnection. Institutional and regional context contributed modest but meaningful explanatory value, supporting a socio-ecological understanding of digital mental health risk. For social work research and practice, findings highlight the need for interventions addressing online harm, strengthening campus mental health supports, and accounting for place-based inequities in student well-being. In line with the SSWR 2027 theme, this study shows how context-sensitive social work research can inform equitable strategies that advance the common good.