Predictors of Brain Injury: The Performance of Biomechanical Head Acceleration Severity Metrics for Concussion Prediction in Men's and Women's Rugby League Players

<jats:p>Objective Describe and compare the biomechanical severity of head acceleration events (HAEs) associated with diagnosed concussion in elite men's and women's rugby league using instrumented mouthguards (iMGs) and evaluate the diagnostic accuracy and screening performance of severity metrics within the current Head Injury Assessment (HIA) process. Methods A prospective cohort of 398 men and 252 women from Super League teams wore iMGs across 515 matches. Following a matching and data quality screening procedure, 123 HIAs (106 men, 17 women) were matched to HAEs, 52 of which were diagnosed concussions (44 men, 8 women). High-magnitude asymptomatic control HAEs (22,676 men, 3,200 women) were sampled proportionally to the number of HIAs. A range of biomechanical severity metrics were calculated for HAEs. Statistical comparisons between outcomes were made. Receiver operating characteristic (ROC) analysis evaluated diagnostic accuracy (ability to predict diagnosed concussions within the HIA). Precision-recall analysis evaluated screening performance (ability to discriminate observable concussion signs from asymptomatic events). Results Diagnosed concussions had greater severity than asymptomatic HAEs across all metrics in both sexes. For diagnostic accuracy, area under the ROC curve ranged from 0.61 to 0.72 in men and 0.53 to 0.86 in women. For screening performance, optimal thresholds in several metrics provided theoretical improvements to precision over current thresholds used in rugby, but recall remained &lt;0.20. Conclusion These findings support integrating iMG-derived severity metrics into a multimodal, clinician-led HIA pathway as objective adjuncts for diagnosis and screening, while reinforcing that they cannot replace clinical judgement or other assessment modalities.</jats:p>