BACKGROUND: Hip arthroscopy outcomes for femoroacetabular impingement remain heterogeneous. The Hip Outcome Score (HOS) comprises 2 distinct subscales measuring activities of daily living and sports performance, yet their dissociation after surgery remains poorly characterised.
PURPOSE: To identify distinct recovery phenotypes after hip arthroscopy using unsupervised machine learning and determine preoperative predictors of phenotype membership.
METHODS: We analysed 326 consecutive patients with complete 1-year HOS-ADL and HOS-Sports data. K-means clustering was performed on 5 outcome variables with Gaussian Mixture Model validation. Multinomial logistic regression identified preoperative predictors.
RESULTS: 4 phenotypes emerged: Complete Responders (41.7%), ADL-Dominant Responders (27.3%), Slow Progressors (19.0%), and Non-Responders (12.0%). ADL-Dominant phenotype was predicted by age >40 years (OR 3.42, p < 0.001), Tegner <5 (OR 2.87, p = 0.001), and Outerbridge III-IV (OR 2.41, p = 0.005).
CONCLUSIONS: 4 robust recovery phenotypes exist, with 27% demonstrating ADL-sports dissociation predicted by older age, lower activity, and greater chondral damage. Recognition enables personalised preoperative counselling.
Read the article: PubMed · Publisher (DOI)