A Predictive Model for Achieving the Minimal Clinically Important Difference Following Hip Arthroscopy: An Analysis of 2,511 Patients

Abstract

Objectives: To build a statistical model to predict two year-post operative Minimal Clinically Important Differences (MCID) for patient reported outcome scores using only preoperative patient data. Methods: Prospectively collected data of all consecutive hip arthroscopy cases for femoroacetabular impingement (FAI) between January 2012-July 2016 were retrospectively identified from a high-volume, single fellowship trained surgeon. Exclusion criteria included dysplasia, patients without a diagnosis of FAI, and follow-up less than two-years. Predictive models for achieving MCID with respect to HOS-ADL, HOS-SS, and mHHS were each built in the following fashion. To reduce the dataset to its most meaningful features and reduce overfitting, the LASSO algorithm was used. This algorithm fits a model on the full dataset and returns non-zero coefficients for features that are determined to be most descriptive while being generalizable. A receiver operating characteristic (ROC) analysis was then performed on each model. In total, 57 independent features were used for modeling. The selected features are then retained for performing a binary logistic regression analysis. Study data were analyzed using PatientIQ (PatientIQ, Chicago IL 60607), a cloud-based research and analytics platform for healthcare. Results: 2,511 patients were identified within the dataset. After application of inclusion and exclusion criteria 1,194 patients were entered into the modeling algorithm. The following MCID cutoffs were used: HOS-ADL = 9.8, HOS-SS = 14.4, mHHS = 9.14. Of patients entered into the algorithm, 71.7% met the HOS-ADL, 73.7% met the HOS-SS, and 77.7% met mHHS MCIDs cutoffs. All ROC curves showed an area under the curve of greater than 0.84. Predictors of not achieving the HOS-ADL MCID included psychiatric history, symptom duration greater than 2 years, age 30-45 years, obesity, low baseline HOS-ADL, and preoperative injection. Predictors of not achieving the HOS-SS MCID included anxiety and depression, proximal hamstring pain with palpation, symptom duration greater than 2 years, low preoperative HOS-SS, and preoperative injection. Predictors of not achieving the mHHS MCID included presence of asthma, prior narcotic use, preoperative physical therapy, presence of snapping IT band, low preoperative mHHS, and preoperative injection. Predictors of achieving the HOS-ADL MCID included a trochanteric pain sign, and higher preoperative mHHS. Predictors of achieving the HOS-SS MCID included a history of running. Predictors of achieving the mHHS MCID included female gender and digestive health history. Conclusion: Several preoperative factors, including mental health, symptom duration length, non intra-articular hip related pain, and preoperative intra articular injections predict failure to achieve the MCID for several patient reported outcome scores. These findings have implications for managing preoperative expectations and anticipated outcomes following hip arthroscopy for FAI.