Comprehensive Summary
This study explores the use of a machine learning regression analysis model to predict anterior cruciate ligament reconstruction (ACLR) revision risk. Data was gathered from over 15,000 patients in a Danish National Registry between 2005 and 2023 and the enhanced ML-Cox regression analysis was performed. Key variables were identified from a 12-month follow up survey to serve as inputs for shaping the model. The model identified four main factors in predicting ACLR revision risk. This included one independent factor (age at time of primary ACLR), and 3 items from the Knee injury and Osteoarthritis Outcome Score (KOOS) – Pain P1, Quality of Life Q2 and Q3. The model was found to have good prediction accuracy at 1, 2, and 5 years (0.73, 0.73 and 0.74 respectively) based on an assessment performed using the C-index. Compared to previous models, this proposed model is able to predict ACLR revision risk using post-operative factors. Various follow up studies are suggested by the authors, including predicting ACLR revision risks within 1 year post-op, that aim to increase the applicability of machine learning models in predicting ACLR revision risk.
Outcomes and Implications
ACL injuries are debilitating with rising incidence rates in adolescent populations. The ability to predict revision risk in ACLR patients using emerging machine learning models decreases the risks associated with ACLR surgery and improves patient outcomes. The applications of this model are significant; however, more studies may be needed before integration into clinical practice. Future research suggested by the authors might focus on model accuracy during a 12 month period or using external data sources.