Purpose <p>Sport-related concussion (SRC) is one of the most complex injuries in sports medicine, and many athletes experience one or multiple SRCs over their collegiate career. However, from the start to the end of an athlete’s collegiate career, it remains unclear which methods can accurately predict the change, if any, in an athlete’s reported symptom burden, cognitive status, and risk of psychological distress. It is also unclear which factors (e.g., number of concussions) have the greatest influence on this change.</p> Methods <p>We consider 3201 (1668 male, 1533 female) collegiate varsity athletes from the Concussion Assessment, Research, and Education (CARE) Consortium. Using five machine learning methods, we predict the change in athletes’ reported symptom burden (i.e., Sport Concussion Assessment Tool [SCAT]), cognitive status (i.e., Standardized Assessment of Concussion [SAC]), and risk of psychological distress (i.e., Brief Symptom Inventory 18 [BSI-18]) over their collegiate careers.</p> Results <p>All machine learning methods outperform a simple model that assumes no change with respect to mean squared error (i.e., machine learning helps avoid large prediction errors). We find that the initial baseline evaluation score is of greatest importance across all metrics of interest (e.g., SCAT/SAC/BSI-18). We also find that when an athlete has a poorer performance on the initial baseline evaluation, the prediction for the change in score is often larger and change for the better.</p> Conclusion <p>This research provides insights on an athlete’s change in critical areas of functioning over their collegiate career, and how machine learning can help improve the prediction of this change.</p>