Estimating an Individual’s Probability of Revision Surgery After Knee Replacement: A Comparison of Modeling Approaches Using a National Data Set

Abstract

Tools that provide personalized risk prediction of outcomes after surgical procedures help patients make preference-based decisions among the available treatment options. However, it is unclear which modeling approach provides the most accurate risk estimation. We constructed and compared several parametric and nonparametric models for predicting prosthesis survivorship after knee replacement surgery for osteoarthritis. We used 430,455 patient-procedure episodes between April 2003 and September 2015 from the National Joint Registry for England, Wales, Northern Ireland, and the Isle of Man. The flexible parametric survival and random survival forest models most accurately captured the observed probability of remaining event-free. The concordance index for the flexible parametric model was the highest (0.705, 95% confidence interval (CI): 0.702, 0.707) for total knee replacement and was 0.639 (95% CI: 0.634, 0.643) for unicondylar knee replacement and 0.589 (95% CI: 0.586, 0.592) for patellofemoral replacement. The observed-to-predicted ratios for both the flexible parametric and the random survival forest approaches indicated that models tended to underestimate the risks for most risk groups. Our results show that the flexible parametric model has a better overall performance compared with other tested parametric methods and has better discrimination compared with the random survival forest approach.

Personalized estimation of one-year mortality risk after elective hip or knee arthroplasty for osteoarthritis

“… 13 , 17 However, these variables do influence other clinical outcomes of interest to patients in the decision-making process that justifies their inclusion in the model. 7 …”

“…4 - 6 Personalized instruments are more complex to construct than generic ones and require data gathered from a population of sufficient size, and with sufficient diversity of input and output variables to provide a statistically meaningful estimation of the outcomes. 7 , 8 …”

“…We recently reported the development of the first comprehensive web-based personalized decision support tool for patients considering hip arthroplasty ( https://jointcalc.shef.ac.uk ). 7 , 18 , 19 This tool was developed in partnership with patients to enable them to come to an informed decision about the risks and benefits of surgery. Here, we describe the development and validation of the tool’s underlying personalized models that generate absolute predicted risk of death over the first year after hip or knee arthroplasty using demographic, prosthesis, and procedure-related data from the National Joint Registry for England and Wales (NJR).…”

Hot spots and trends in knee revision research since the 21st century: a bibliometric analysis

“…Nationwide and worldwide statistical data has been collected to analyze the rate and trends of revision (2,(54)(55)(56). Recently, a research group established several parametric and non-parametric models to estimate prostheses' survivorship more accurately (57). Another research group analyzed the predictive factors of revision, prosthetic infection and mortality in rheumatoid arthritis patients based on Danish healthcare registers (58).…”

Dynamic Survival Prediction Combining Landmarking with a Machine Learning Ensemble: Methodology and Empirical Comparison

“…In this paper, we will include these ensemble techniques in the definition of machine learning even though some of the component algorithms are parametric statistical models. Recent applications of machine learning methods for clinical prediction include: assessment of delirium risk (Wong et al., 2018), 3‐year survival for CF patients (Alaa & van der Schaar, 2018), mortality in coronary artery disease (Steele et al., 2018) and time to revision surgery after knee replacement (Aram et al., 2018). These studies and others have compared machine learning methods to traditional methods and found mixed results with some finding superior performance for parametric statistical models and others finding improved results with various machine learning algorithms.…”