Background: Markerless motion capture (MMC) systems used to screen for musculoskeletal injury (MSKI) risk have become popular in military and collegiate athletic settings. However, little is known regarding the test-retest reliability or, more importantly, the ability of these systems to accurately identify individuals at risk for MSKI. Purpose: To determine the association between scores from a proprietary MMC movement screen test and the likelihood of suffering a subsequent MSKI and establish the test-retest reliability of the MMC system used. Study Design: Cohort study; Level of evidence, 3. Methods: Trainees for the Air Force Special Warfare program underwent MMC screenings immediately before entering the 8-week training course. MSKI data were extracted from a database for the surveillance period for each trainee. Logistic regression analyses were performed to identify associations between baseline MMC scores and the likelihood of suffering any MSKI or, specifically, a lower extremity MSKI. The test-retest portion of the study collected MMC scores from 10 separate participants performing 4 trials of the standard test procedures. Reliability was assessed using intraclass correlation coefficients by a single rater. Results: Overall, 1570 trainees, of whom 800 (51%) suffered an MSKI, were included in the analysis. MMC scores poorly predicted the likelihood of any or a lower extremity MSKI (odds ratio, 1.01-1.02). Further, receiver operating characteristic curve analyses demonstrated poor sensitivity and specificity for prediction of MSKI with MMC scores (area under the curve = 0.53). Finally, intraclass correlation coefficients from the test-retest analysis of MMC scores ranged from 0.157 to 0.602. Conclusion: This MMC system displayed poor to moderate test-retest reliability and did not demonstrate the ability to discriminate between individuals who were and were not likely to suffer an MSKI.
Machine Learning to Predict Lower Extremity Musculoskeletal Injury Risk in Student Athletes