Search Filters

  • Presentation Format
  • Media Type
  • Diagnosis / Condition
  • Diagnosis Method
  • Patient Populations
  • Treatment / Technique

Predicting Subsequent Revision ACL Reconstruction: A Machine Learning Analysis of the Norwegian Knee Ligament Register

Predicting Subsequent Revision ACL Reconstruction: A Machine Learning Analysis of the Norwegian Knee Ligament Register

R. Kyle Martin, MD, FRCSC, UNITED STATES Solvejg Wastvedt, BA, UNITED STATES Ayoosh Pareek, MD, UNITED STATES Andreas Persson, MD, PhD, NORWAY Havard Visnes, MD, PT, PhD, NORWAY Anne Marie Fenstad, MSc, NORWAY Gilbert Moatshe, MD, PhD, NORWAY Julian Wolfson, PhD, UNITED STATES Lars Engebretsen, MD, PhD, NORWAY

University of Minnesota, Minneapolis, MN, UNITED STATES


2021 Congress   Abstract Presentation   5 minutes   Not yet rated

 

Anatomic Location

Anatomic Structure

Diagnosis / Condition

Ligaments

ACL

This media is available to current ISAKOS Members, Global Link All-Access Subscribers and Webinar/Course Registrants only.

Summary: This machine learning analysis of a national knee ligament register can predict a patient’s risk of primary ACL reconstruction failure (defined as a subsequent revision surgery). The resulting algorithm supports the creation of an easy-to-use calculator for point-of-care risk stratification which can be used to guide surgical discussions with patients and quantify their specific risk of failure.


Background

Several factors are associated with an increased risk of anterior cruciate ligament (ACL) reconstruction failure. However, due to the multiple patient, surgical, and rehabilitation factors that influence outcome, the ability to accurately translate these factors into a quantifiable risk of failure at a patient-specific level has remained elusive. Our hypothesis was that machine learning analysis of existing large national knee ligament registers has the potential to improve our predictive capability. The purpose of this study was to determine if machine learning analysis of primary ACL reconstructions in the Norwegian Knee Ligament Register (NKLR) can: (1) identify the most important risk factors associated with undergoing a subsequent revision ACL reconstruction, and (2) develop a clinically meaningful calculator for predicting the risk of requiring a revision operation.

Methods

Machine learning analysis was performed on the NKLR dataset. The primary outcome was probability of revision ACL reconstruction within 1, 2, and/or 5 years. Data was split randomly into training (75%) and test (25%) sets. Four machine learning models were tested: Cox Lasso, survival random forest, generalized additive model, and gradient boosted regression. Concordance and calibration were calculated for all four models.

Results

The dataset included 24,935 patients, and 4.9% underwent revision surgery during an average follow-up of 8.1 years (SD 4.1). All four models were well-calibrated, with moderate concordance (0.67-0.69). The Cox Lasso model required only five variables for outcome prediction: graft choice, femoral fixation device, KOOS QOL score at surgery, years from injury to surgery, and age at surgery. The other models either used more variables without an appreciable improvement in accuracy or had slightly lower accuracy overall. An in-clinic calculator was developed which can estimate the risk of graft failure (https://swastvedt.shinyapps.io/calculator_rev/). Whereas the overall risk of revision in the registry was 4.9%, this calculator can quantify risk at a patient-specific level.

Conclusions

Machine learning analysis of a national knee ligament register can predict the risk of a patient undergoing a subsequent revision ACL reconstruction after primary surgery with moderate accuracy. This algorithm supports the creation of an in-clinic calculator for point-of-care risk stratification prior to primary surgery based on the input of only five variables. Similar analysis using larger or more comprehensive data may improve the accuracy of risk prediction.


More ISAKOS 2021: Global Content