A Machine Learning Model for Post-Concussion Musculoskeletal Injury Risk in Collegiate Athletes.

Area of Science:

• Sports Medicine
• Biostatistics
• Machine Learning in Healthcare

Background:

• Collegiate athletes face an increased risk of post-concussion musculoskeletal (MSK) injuries.
• Identifying specific athletes at high risk for these injuries is crucial but challenging.

Purpose of the Study:

• To develop a predictive model for post-concussion MSK injury risk in collegiate athletes.
• Utilize machine learning to integrate diverse variables for accurate risk assessment.

Main Methods:

• A machine learning model was trained on 155 athletes and tested on 39, using 135 variables including health, athletic history, and concussion assessment data.
• Techniques included Weight of Evidence transformation, L1-penalized logistic regression for variable selection, Akaike Information Criterion for model selection, and L2-regularized logistic regression fitting.

Main Results:

• A 48-variable model demonstrated significant predictive performance (AUC=0.82) for subsequent MSK injuries.
• Key predictors included cognitive, balance, and reaction assessments at baseline and acute post-concussion stages.
• The model achieved 79% sensitivity and 95% precision at a 6.67% false positive rate for identifying at-risk athletes.

Conclusions:

• The developed model offers a sensitive and specific approach to identify collegiate athletes at high risk for post-concussion MSK injuries.
• Integrating machine learning with comprehensive data allows for targeted injury risk reduction strategies.
• This approach can improve clinical targeting of student-athletes most susceptible to post-concussion MSK injuries.