Physical activity type and psychosocial factors as predictors of mental distress in U.S. college students

Mental health disorders, including depression and anxiety, remain among the most pressing challenges f acing U.S. college students, driven by academic pressures, social stressors, and lifestyle disruptions. Physical activity (PA) has been proposed as a preventive measure for poor mental health, yet whether its benefits vary by activity type, frequency, and duration, and whether any such associations persist after adjustment for psychosocial burden, remain unclear. This study tested whether physical activity type was associated with self-reported poor mental health days, and whether psychosocial factors (stress, loneliness, depression history, life satisfaction) accounted for significantly greater variance than physical activity behavior. Using 2023 Behavioral Risk Factor Surveillance System (BRFSS) data, 1,862 U.S. college students aged 18–24 were identified after applying inclusion criteria and excluding three physical activity categories with insufficient sample sizes. The primary outcome was self-reported poor mental health days in the past 30 days (MENTHLTH, 0–30, continuous) and a binary indicator of frequent mental distress (≥ 14 days; CDC definition). Key independent variables included physical activity type, frequency, session duration, weekly minutes of primary activity, strength-training frequency, depression diagnosis, life satisfaction, emotional support, loneliness, stress, and socioeconomic indicators. Welch’s one-way ANOVA, Kruskal–Wallis, and TukeyHSD post-hoc tests examined differences across activity types. Nested multiple linear regression models (Models 0–5) and a parallel binary logistic regression examined predictors of poor mental health. Heteroscedasticity-consistent (HC3) standard errors were used for inference. Survey-weighted sensitivity analyses were conducted using LLCPWT sampling weights (sampling weights only; strata and primary sampling units were not specified). Welch’s ANOVA confirmed significant differences in poor mental health days across PA types (F(7, 148.09) = 8.34, p < 0.0001; η² = 0.032). Despite statistical significance, due to the small effect size, PA type explained little variance in poor mental health days; finding should be interpreted cautiously. TukeyHSD identified three significant pairwise differences, all versus walking: running/jogging (Δ = −3.63 days, p < 0.0001, d = 0.46), weight lifting (Δ = −2.88 days, p < 0.0001, d = 0.35), and other activities (Δ = −2.37 days, p < 0.001, d = 0.28). PA modality variables (frequency, duration, weekly minutes, strength frequency) did not significantly improve model fit beyond PA type and demographics ( p = 0.341). Adding psychosocial factors produced a significant jump in explained variance (R²: 0.079–0.443, LRT p < 0.0001). Socioeconomic indicators did not further improve fit ( p = 0.062). In the full model (Model 5; adjusted R² = 0.433), stress, loneliness, depression diagnosis, and life satisfaction were the dominant predictors. Only running/jogging remained significant among PA types after full adjustment (β = −1.14, p < 0.01; β* = −0.054, a small effect). Binary logistic regression (Nagelkerke R² = 0.434) confirmed that students who never or rarely felt stressed had 94–95% lower odds of frequent distress; running/jogging was the only PA type associated with significantly lower odds (OR = 0.61, 95% CI [0.38, 0.95], p < 0.05). Using data from a large, multi-state U.S. survey with detailed physical activity type data for college students, this study found that psychosocial burden, particularly stress, loneliness, and depression history, accounted for more variance in mental distress than exercise behavior. Running/jogging was the only PA type associated with fewer poor mental health days after full psychosocial and socioeconomic adjustment. This association was small, however, and was significant in the primary unweighted analysis but only borderline significant in the survey-weighted sensitivity analysis; given the cross-sectional design, it should be regarded as hypothesis-generating. These findings suggest that PA may be most effective as one component of campus wellness initiatives that also address stress and social connectedness.