When compared to the first THA, revision total hip arthroplasty (THA) is associated with greater morbidity, mortality, and healthcare expenditures due to a technically more difficult surgical process. As a result, a better knowledge of risk factors for early revision is required. THA is required to develop techniques to reduce the probability of patients having early revision. For a study, researchers sought to create and verify new machine learning (ML) models for predicting early revision after primary THA.
A total of 7,397 patients who underwent primary THA were assessed, with 566 patients (6.6%) having confirmed early revision THA (<2 years after index THA). Electronic patient data carefully evaluated medical demographics, implant characteristics, and surgical factors related to early revision THA. About 6 machine learning methods were constructed to predict early revision THA, and their performance was evaluated using discrimination, calibration, and decision curve analysis.
The Charlson Comorbidity Index, body mass index of more than >35 kg/m2, and depression were the best predictors of early revision after initial THA. In addition, the six ML models all performed well in discrimination (area under the curve >0.80), calibration, and decision curve analysis. The study used ML models to predict early revision surgery for individuals with original THA. The study findings revealed that all six candidate models perform well in discrimination, calibration, and decision curve analysis, underlining the possibility of these models to aid in clinical practice patient-specific preoperative estimation of greater risk of early revision THA.
Reference:journals.lww.com/jaaos/Abstract/2022/06010/The_Utility_of_Machine_Learning_Algorithms_for_the.4.aspx
