74 (8A) Quantifying full-season head acceleration exposure in professional men’s rugby league players: exploring imputation methods with instrumented mouthguards

Purpose Head acceleration events (HAEs) are a concern in sports due to potential long-term neurological effects.Instrumented mouthguards (iMG) quantify HAEs, however missing data limits their ability to provide full-season exposure.This study utilises imputation methods to estimate HAE exposure in professional men's rugby league.Methods Video analysis was conducted across 167 rugby league matches of the 2024 Super League season, 254 male players recorded iMG data in at least one match.These players participated in 4,600 player matches and iMG data were collected in 68.8% (n=3,163), meaning imputation was necessary in the remaining 31.2%.Player matches with iMG data were split into training (n=1,704) and testing (n=424) datasets to train XGBoost, random forest, and generalised linear mixed models (GLMM).Estimated HAE count was compared against the actual via root mean square error (RMSE), mean absolute error (MAE), and R 2 values.Results The median (interquartile range [IQR]) HAE count per player match was 24 (16-33).The GLMM had the highest predictive performance with RMSE, MAE and R 2 of 6.47, 4.86, and 0.70, respectively.XGBoost and Random Forest models performed worse with RMSE of 7.98 and 8.09, MAE of 6.06 and 6.13, and R 2 of 0.57 and 0.54.The median (IQR) full-season HAE count was 390 (218-601) using GLMM imputation, the highest was 1,606 (figure 1).Conclusion Imputation is necessary for estimating full HAE exposures given missing iMG data.GLMMs may be beneficial for imputation due to their ability to account for individuals.Predicted HAE exposures revealed considerable variation in full-season HAE exposure across players.