An examination of computerized neurocognitive test scores as predictors of risk of concussion in collegiate athletes

Sport-related concussion affects between 1.6 and 3.8 million people annually. Incidence of concussion in the population of collegiate athletes is a public health issue because there are nearly half a million student-athletes at risk every year and physiological effects have the potential to be lifelong and severe in nature. To combat incidence and severity of concussion in college athletes, the CDC and NCAA have partnered to promote education about concussion and implementation of a concussion protocol. At the University of Pittsburgh, this protocol includes a baseline ImPACT test for each incoming student-athlete for the purpose of tracking recovery and facilitating “return to play” decisions following incidence of a concussion. This study explores the utility of the baseline ImPACT test scores in determining risk for concussion. Computerized neurocognitive assessments are developed for post-injury purposes. The analysis in this study uses the ImPACT test scores in a novel way, pre-injury, and examines novel risk factors for concussion, e.g., the composite scores, which were hypothesized to affect incidence through an effect on athletic performance. Two independent survival analyses were performed: case-wise deletion and multiple imputation by chained equations. Because the data were determined to have MAR missingness, multiple imputation models were adopted in lieu of case-wise deletion models to ensure unbiased parameter estimates. Two independent models were created: male and female. The model for female athletes included the history indicator and contact expected by sport variables. The model for male athletes included the history indicator, contact expected, processing speed composite score, and height variables. Although these results cannot be generalized with certainty, the results of this study provide an exploratory analysis of the utility of computerized neurocognitive assessments in determining risk for concussion. Four of the six composite scores (visual memory, processing speed, reaction time, and total symptom) were univariably significant (a=0.20) for male athletes. The multivariable model for male athletes included the processing speed composite score. Subsequent studies and analyses, with a focus on securing a representative sample, are required to corroborate the findings of this study.