Application of Machine Learning for Predicting Clinically Meaningful Outcome After Arthroscopic Femoroacetabular Impingement Surgery

BACKGROUND: Hip arthroscopy has become an important tool for surgical treatment of intra-articular hip pathology. Predictive models for clinically meaningful outcomes in patients undergoing hip arthroscopy for femoroacetabular impingement syndrome (FAIS) are unknown. PURPOSE: To apply a machine learning model to determine preoperative variables predictive for achieving the minimal clinically important difference (MCID) at 2 years after hip arthroscopy for FAIS. STUDY DESIGN: Case-control study; Level of evidence, 3. METHODS: Data were analyzed for patients who underwent hip arthroscopy for FAIS by a high-volume fellowship-trained surgeon between January 2012 and July 2016. The MCID cutoffs for the Hip Outcome Score-Activities of Daily Living (HOS-ADL), HOS-Sport Specific (HOS-SS), and modified Harris Hip Score (mHHS) were 9.8, 14.4, and 9.14, respectively. Predictive models for achieving the MCID with respect to each were built with the LASSO algorithm (least absolute shrinkage and selection operator) for feature selection, followed by logistic regression on the selected features. Study data were analyzed with PatientIQ, a cloud-based research and analytics platform for health care. RESULTS: 2 years, preoperative intra-articular injection, and high preoperative outcome scores are most consistently predictive of inability to achieve clinically meaningful outcome. These findings have important implications for shared decision-making algorithms and management of preoperative expectations after hip arthroscopy for FAI.