476 MEP033 – Predicting groin injury in male soccer players using machine learning and multivariate analysis. a prospective cohort study

AbstractBackground Groin injury remains one of the most prevalent injuries in soccer. Injury prediction challenges sports medicine researchers to provide high injury risk accuracy to mitigate injury incident rates.Objective To identify soccer players with increased risk for groin injury through the predictive ability of a Machine Learning (ML) algorithm based on a large set of pre-season data. Study Design Case-control studyMethods One hundred and twenty (n=120) male amateur soccer players (mean age: 20.0 +/- 6.96 years) were prospectively followed during the 2018/19 competitive season. Pre-season Maximal Isometric Strength of the hip adductors, abductors, flexors, and the knee flexors was measured using a handheld dynamometer and ratios were calculated. Previous injury history and anthropometric data were additionally registered. Injury data were collected throughout season. The K- Nearest Neighbour (k-NN), a supervised machine learning algorithm was used to predict injury based on the total pre-season data, while the significant variables from the k-NN algorithm was fitted into a multivariate logistic regression model (LR) to analyze the likelihood of players sustaining a groin injury.Results Twenty-two (n=22) players experienced 25 groin injuries. The k-NN model achieved a mean accuracy score of 55%, indicating a somewhat reasonable prediction of the player's injury. The AUC was 0.43, which signifies a reasonable modelling in predicting the groin injury. The LR model revealed two variables as significant predictors of groin injury. Players were less likely to sustain groin injury by 76% for each decrease of Adductors/Abductors isometric strength ratio in the non-dominant limb (OR = 0.238, CI95%= [(0.098–0.572]). Players with history of previous injury had 67% more risk of sustaining an injury (OR = 0.333, CI95%= [(0.1068–1.038]). Conclusion Isometric hip adductor and abductor strength imbalances of the non-dominant lower limb and history of previous injury were risk factors for groin injury in amateur soccer players.