Athletic identity and depressive symptom risk in college students: a machine learning approach to identify at-risk profiles and protective correlates.

The relationship between athletic participation and depressive symptom risk among college students remains complex and contested, with prior research yielding inconsistent findings regarding whether sports involvement serves as a protective correlate or a correlate of psychological vulnerability. Using cross-sectional survey data from 795 undergraduate students at a large public university in the northeastern United States, this study employed machine learning methods to examine how athletic identity profiles, sports participation patterns, and related psychosocial correlates were associated with depressive symptoms. We specifically aimed to identify distinct athletic identity profiles, evaluate whether these profiles improved depression-risk prediction, and examine whether risky health behaviors and masculine norm variables statistically accounted for part of the observed associations. Using unsupervised clustering algorithms, we identified four distinct athletic identity profiles characterized by varying combinations of jock identity strength, athlete identity strength, and sport goal orientation. Gradient boosting classification models were then developed to predict elevated depressive symptoms, achieving robust predictive performance validated through five-fold cross-validation. Feature importance analysis revealed that the association between athletic participation and depressive symptom risk is highly context-dependent, moderated by factors including gender conformity norms, primary sport type, competitive level, and the balance between task and ego orientation in sports. Notably, strong athlete identity characterized by intrinsic motivation and task mastery orientation was associated with lower depression risk, whereas strong jock identity characterized by status-seeking and conformity to traditional masculine norms was associated with elevated risk, particularly among male students. Models of indirect association further suggested that risky health behaviors statistically accounted for part of the association between jock identity and depressive symptoms. Because all variables were measured cross-sectionally, these indirect effects should not be interpreted as evidence of causal mediation. These findings challenge the simplistic notion that sports participation uniformly benefits mental health and highlight the importance of considering identity formation processes and motivational orientations in understanding athlete wellbeing. The machine learning framework developed in this study offers practical utility for identifying students who may benefit from targeted depression-prevention screening and support and for informing evidence-based approaches to promoting psychological wellness within collegiate athletic programs.