Decision Tree Learning Algorithm for Classifying Knee Injury Status Using Return-to-Activity Criteria.

Area of Science:

• Orthopedics
• Sports Medicine
• Biomechanical Analysis

Background:

• Increasing rates of anterior cruciate ligament (ACL) injuries in adolescent females.
• High incidence of secondary ACL injuries post-return to activity (RTA) due to lack of evidence-informed guidelines.
• Need for reliable functional assessment measures to determine RTA readiness.

Purpose of the Study:

• To evaluate functional capacity assessment tests for adolescent females after ACL surgery.
• To identify key variables for determining readiness for return to activity.
• To utilize a Reduced Error Pruning Tree (REPT) model for variable selection.

Main Methods:

• Thirty-six healthy and forty-two ACL-injured adolescent females participated.
• Functional tasks were assessed using motion analysis and spatiotemporal measures.
• A REPT model was employed to reduce thirty variables to key predictors.

Main Results:

• The REPT model identified maximum anterior hop and maximum lateral hop limb symmetry measures as significant.
• These two measures achieved 69% sensitivity and 78% specificity in classifying injury status.
• A moderate confidence (kappa statistic of 0.464) in differentiating between ACL-injured and healthy participants was observed.

Conclusions:

• The REPT model effectively evaluated functional capacity related to ACL injury status in adolescent females.
• Maximum anterior and lateral hop tests are recommended for developing RTA assessments and guidelines.
• These spatiotemporal measures show potential for clinical evaluation of functional capacity and RTA readiness.