Abstract
A quick, portable and reliable tool for predicting ACL injury could be an invaluable instrument for athletes, coaches, and clinicians. The gold standard, Vicon motion analysis, despite having a high sensitivity and risk specificity, is not practical for coaches or clinicians to use on a routine basis for assessing athletes. The present study validated the Kinect device to the currently used method of chart review in predicting athletes at high risk.
A total of 114 participants were recruited from both the men and women McGill Varsity Sports Program. 69 males and 45 female athletes were evaluated to assess the specificity and sensitivity of the Kinect device in predicting athletes at high risk of injury. Each athlete performed three-drop vertical jumps off of a 31cm box and the data was recorded and risk score was generated. Generation of this data is done by our uniquely programmed software that measures landing angles at different time frames and compares live results to previously known data of injured athletes. A chart review was then performed by a clinician, blinded to these risk scores, to risk stratify the same athletes as high or low risk of ACL injury based on their medical charts. Data reviewed incorporated pre-season physical exams along with documented known risk factors for ACL injury, including previous knee injuries, family history of ACL injury, gender, sport, and BMI. Positive risk factors were assigned one point while negative risk factors assigned zero points.
The Kinect device, powered by our software, identified 40 athletes as having a high-risk score (> 55%), and subsequently, five (4.39%) sustained an ACL injury by the end of their respective sport seasons. Two male and two female basketball players along with one male soccer player sustained non-contact ACL injuries. Given that all five of the injured athletes were in the cohort of 40 identified as high risk by the Kinect, this yielded a sensitivity of 100% for the device. As for the specificity, the Kinect computed 35 false positives, yielding a specificity of 68% for the duration of the study. The medical chart review identified 36 athletes as high risk and 60 as being low risk of ACL injury. Four of the athletes that sustained an ACL injury were in the group of 36 identified as high risk by the clinician. However, one of the five participants who sustained an ACL injury was not captured by the medical chart assessment, yielding a sensitivity of 80% and a specificity of 65% for the clinician.
When it comes to injury prediction, it is preferred to have a high sensitivity even if the specificity is slightly lower as this ensures that all athletes who are at risk will be captured. Our data demonstrated that the chart analysis provided one false negative and led to missing one high-risk athlete who ended up sustaining an ACL injury. Based on the comparison of sensitivity and specificity, the Kinect system provides a slightly better predictive analysis for predicting ACL injury compared to chart review.
