While two-stage revision is the gold standard for treatment of knee prosthetic joint infection (PJI), it is not without risk. The purpose of this study was to develop a tool to preoperatively predict the probability that a two-stage revision would fail to eradicate knee PJI. 3,809 surgical cases were retrospectively reviewed and data were collected from 314 charts. Overall, 105 (33.4%) cases failed to eradicate PJI using this procedure. Univariate analysis identified multiple variables independently associated with reinfection. Logistic regression was used to generate a model (bootstrap-corrected concordance index of 0.773) predicting failure of infection eradication. Preoperative knowledge of a high probability of failure may improve risk assessment, lead to more aggressive management, and allow for time to consider alternative therapies.
Keywords: infection; nomogram; predictive modeling; revision; total knee arthroplasty.
Copyright © 2014 Elsevier Inc. All rights reserved.