Predictive modeling of lower extremity injury risk in male elite youth soccer players using least absolute shrinkage and selection operator regression

Abstract

Purpose: To (1) identify neuromuscular and biomechanical injury risk factors in elite youth soccer players and (2) assess the predictive ability of a machine learning approach. Material and Methods: Fifty-six elite male youth soccer players (age: 17.2 ± 1.1 years; height: 179 ± 8 cm; mass: 70.4 ± 9.2 kg) performed a 3D motion analysis, postural control testing, and strength testing. Non-contact lower extremities injuries were documented throughout 10 months. A least absolute shrinkage and selection operator (LASSO) regression model was used to identify the most important injury predictors. Predictive performance of the LASSO model was determined in a leave-one-out (LOO) prediction competition. Results: Twenty-three non-contact injuries were registered. The LASSO model identified concentric knee extensor peak torque, hip transversal plane moment in the single-leg drop landing task and center of pressure sway in the single-leg stance test as the three most important predictors for injury in that order. The LASSO model was able to predict injury outcomes with a likelihood of 58% and an area under the ROC curve of 0.63 (sensitivity = 35%; specificity = 79%). Conclusion: The three most important variables for predicting the injury outcome suggest the importance of neuromuscular and biomechanical performance measures in elite youth soccer. These preliminary results may have practical implications for future directions in injury risk screening and planning, as well as for the development of customized training programs to counteract intrinsic injury risk factors. However, the poor predictive performance of the final model confirms the challenge of predicting sports injuries, and the model must therefore be evaluated in larger samples.