Associations between training behaviors and injuries in performance-oriented parkrunners using traditional statistical methods and machine learning.

BACKGROUND: parkrun is the most popular recreational running event in the UK, yet running-related injuries (RRIs) pose a threat to participation. Training behaviors are likely to influence the likelihood of RRIs with strength and conditioning (S&C) activities widely believed to reduce RRI risk. This study aimed to examine associations between S&C training habits plus broader training behaviors and RRIs using traditional statistical methods and machine learning (ML). METHODS: parkrun participants completed an online survey including questions about personal characteristics, training behaviors, and RRIs during the past 12 months. Chi square and Mann-Whitney U tests were used to investigate associations between training behavior and RRIs. ML classifiers were employed to establish multi-factorial prediction of RRIs. Shapley Additive Explanations were used to explain how frequently selected features affect RRI probability. RESULTS: 1,203 responses were obtained and 570 performance-oriented parkrun participants were used for data analysis. Resistance training (Φ = 0.099, p = 0.018) and stretching/yoga (Φ = 0.095, p = 0.023) were positively associated with RRIs. No other S&C or running training behaviors were associated with RRIs. ML classifiers achieved moderate predictive capacities and found non-linear associations between RRIs and number of interval training sessions per week, a positive association with stretching/yoga, and a negative association with coach-guided training. CONCLUSION: Participation in S&C activities is not associated with a lower likelihood of RRIs in parkrunners and resistance training and stretching/yoga was linked to greater likelihood of RRIs. ML classifiers suggest there is benefit to performing a high frequency of interval training sessions, excluding stretching/yoga, and having qualified coaches prescribe training.