Can machine learning predict sport-related injuries among student-athletes? A longitudinal study

In this longitudinal study, we examined whether machine learning could be used to predict the role of psychological factors in the development of sport-related injuries among adolescent student-athletes aged 13–14 years. A total of 207 student-athletes (54% males, 46% females) from seven Norwegian lower secondary sport schools completed weekly questionnaires on sport-related injuries, time use in sport, sports burnout, and sports engagement across Grades 8–10, yielding a total of 2,131 person-weeks of data. Using a random forest approach, the results showed that previous injury history was the most important predictor of future injuries, alongside cumulative training load and changes in training volume. Symptoms of sport-related exhaustion and cynicism increased injury risk when combined with concurrent changes in training load. High sport engagement was associated with an increased risk of reinjury following return to sport, whereas chronically low sport engagement was linked to a higher injury risk over time. Overall, the findings suggest that injury and health monitoring among student-athletes in this age group should begin early. Moreover, rather than focusing solely on cross-sectional correlations at single time points, machine-learning approaches that capture how risk factors accumulate and interact over time may offer particularly valuable insights into injury development.