Risk of Conversion to Arthroplasty After Hip Arthroscopy: Validation of a Published Risk Score Using an Independent, Prospectively Collected Database

Background: Hip arthroscopy is rapidly advancing and increasingly commonly performed. The most common surgery after arthroscopy is total hip arthroplasty (THA), which unfortunately occurs within 2 years of arthroscopy in up to 10% of patients. Predictive models for conversion to THA, such as that proposed by Redmond et al, have potentially substantial value in perioperative counseling and decreasing early arthroscopy failures; however, these models need to be externally validated to demonstrate broad applicability.

Purpose: To utilize an independent, prospectively collected database to externally validate a previously published risk calculator by determining its accuracy in predicting conversion of hip arthroscopy to THA at a minimum 2-year follow-up.

Study design: Cohort study (diagnosis); Level of evidence, 1.

Methods: Hip arthroscopies performed at a single center between November 2015 and March 2017 were reviewed. Patients were assessed pre- and intraoperatively for components of the THA risk score studied-namely, age, modified Harris Hip Score, lateral center-edge angle, revision procedure, femoral version, and femoral and acetabular Outerbridge scores-and followed for a minimum of 2 years. Conversion to THA was determined along with the risk score’s receiver operating characteristic (ROC) curve and Brier score calibration characteristics.

Results: A total of 187 patients (43 men, 144 women, mean age, 36.0 ± 12.4 years) underwent hip arthroscopy and were followed for a mean of 2.9 ± 0.85 years (range, 2.0-5.5 years), with 13 patients (7%) converting to THA at a mean of 1.6 ± 0.9 years. Patients who converted to THA had a mean predicted arthroplasty risk of 22.6% ± 12.0%, compared with patients who remained arthroplasty-free with a predicted risk of 4.6% ± 5.3% (P < .01). The Brier score for the calculator was 0.04 (P = .53), which was not statistically different from ideal calibration, and the calculator demonstrated a satisfactory area under the curve of 0.894 (P < .001).

Conclusion: This external validation study supported our hypothesis in that the THA risk score described by Redmond et al was found to accurately predict which patients undergoing hip arthroscopy were at risk for converting to subsequent arthroplasty, with satisfactory discriminatory, ROC curve, and Brier score calibration characteristics. These findings are important in that they provide surgeons with validated tools to identify the patients at greatest risk for failure after hip arthroscopy and assist in perioperative counseling and decision making.

Keywords: arthroplasty risk; external validation; hip arthroscopy; predictive models; total hip arthroplasty.