Mental Toughness Subscales Associated With Mental Health Outcomes In NCAA Athletes: Network And Machine Learning Insights

The International Olympic Committee has emphasized the need for structured mental health (MH) screening, introducing the Sports Mental Health Assessment Tool 1 (SMHAT-1). While existing research shows a positive correlation between mental toughness (MT) and MH outcomes, no study has yet confirmed this relationship within NCAA athletes by analyzing MT effects at the subscale level via machine learning and network analyses. PURPOSE: To confirm the MT-MH association and examine relationships between individual MT subscales and MH outcomes in NCAA athletes. METHODS: Retrospective data from standard baseline MH screenings (SMHAT-1) were analyzed for 493 participants (Mage = 20.4 years, SD = 3.1; 57% female), who completed the MT Index assessing all eight subscales. MH outcomes included anxiety, depression, sleep disturbance, alcohol misuse, and disordered eating. Logistic regression models estimated odds ratios (ORs) between MT total scores and MH outcomes, using established clinical thresholds. Network analysis assessed centrality among MT subscales and MH outcomes. Machine learning models evaluated each MT subscale’s predictive value for MH outcomes, with feature importance scores identifying key predictors. RESULTS: Higher MT total scores were associated with lower odds of MH challenges, with significant ORs observed for anxiety (OR = 0.69, 95% CI [0.57, 0.82], p < .001), depression (OR = 0.72, 95% CI [0.61, 0.84], p < .001), sleep disturbance (OR = 0.68, 95% CI [0.57, 0.81], p < .001), alcohol misuse (OR = 0.84, 95% CI [0.74, 0.95], p < .05), and disordered eating (OR = 0.76, 95% CI [0.65, 0.88], p < .001). Network analysis identified Attention Regulation (AR) (degree centrality = .85, betweenness centrality = .26) and Generalized Self-efficacy (GSE) (dc = .69, bc = .08) as central nodes in the MT-MH network. Machine learning models reinforced these findings, with AR (feature importance = .18 in Random Forest, .16 in Gradient Boosting) and GSE (fi = .14 in RF, .12 in GB) as the strongest predictors across MH outcomes. CONCLUSIONS: Targeting AR and GSE in MT interventions may enhance MH outcomes, providing a focused, profile-based approach to MH support within NCAA athletes. Future studies should incorporate longitudinal designs.