2023 ISAKOS Biennial Congress ePoster
Machine Learning Clustering Analysis to Identify High Achievers after Anterior Cruciate Ligament Reconstruction: Meniscal Debridement Strongly Predicts Negative Clinically Significant Outcomes
Yining Lu, MD, Rochester, Minnesota UNITED STATES
Elyse Berlinberg, BS, New York, NY UNITED STATES
Parker Maxwell Rea, BS, Morton Grove, IL UNITED STATES
Harsh Patel, BS, New Brunswick, New Jersey UNITED STATES
Matthew Cohn, MD, Philadelphia, PA UNITED STATES
James Baker, MD, Chicago, IL UNITED STATES
Adam B. Yanke, MD, Chicago, IL UNITED STATES
Brian J. Cole, MD, MBA, Chicago, IL UNITED STATES
Brian Forsythe, MD, Chicago, IL UNITED STATES
Rush University Medical Center, Chicago, IL, UNITED STATES
FDA Status Not Applicable
Summary
Machine learning identifies lower preoperative demand, greater preoperative symptomatology, and concurrent irreparable meniscal injury are associated with postoperative CSO achievement in patients undergoing ACL reconstruction.
ePosters will be available shortly before Congress
Abstract
Background
Anterior cruciate ligament reconstruction (ACLR) is a clinically successful procedure, although there are variations in patient reported outcome measurements (PROMs) not fully explained by traditional predictive models. The aim of this study was to determine predictors of achieving clinically significant outcome (CSO) thresholds on PROM instruments, utilizing unsupervised machine learning (UML).
Methods
A retrospective analysis was performed to identify patients with ACLR, 2015-2018. Minimum clinically important difference (MCID), substantial clinical benefit (SCB), and patient acceptable symptomatic state (PASS) achievement on the Knee Injury and Osteoarthritis Outcome Score (KOOS) and International Knee Documentation Committee (IKDC) subscales at 1-year were determined. Clusters were established based on patterns in PROM performance using UML. Predictors of cluster membership were assessed via stepwise logistic regression.
Results
179 patients were included in the final two clusters: 60 patients in a relatively high CSO-achievement group, and 119 in an average-achievement group. Stepwise multivariable logistic regression identified lower preoperative scores on KOOS sport subscales (OR:0.98, 95% CI:0.96-1), KOOS symptom subscale (OR:0.98, 95%CI:0.96-1.01), and older age (OR:1.05, 95% CI:1.01-1.1) as associated with higher likelihood of CSO achievement. Concurrent meniscectomy predicted reduced likelihood of achievement (OR: 0.23, 95% CI: 0.07-0.76), but meniscal repair was not significantly associated with CSOs (OR:1.22, 95% CI:0.33-2.55, P= 0.76).
Conclusion
Lower preoperative demand, greater preoperative symptomatology, and concurrent irreparable meniscal injury are associated with postoperative CSO achievement. Patients undergoing any concurrent meniscectomy are 4-times less likely to achieve CSOs. This information can be utilized to counsel patient expectations during surgical decision-making.