Ålund, Oskar; Unwin, Robert; Challis, Benjamin; Kalra, Philip A; Taal, Maarten W; Wheeler, David C; Fraser, Simon DS; ... Söderberg, Magnus; + view all Ålund, Oskar; Unwin, Robert; Challis, Benjamin; Kalra, Philip A; Taal, Maarten W; Wheeler, David C; Fraser, Simon DS; Cockwell, Paul; Söderberg, Magnus; - view fewer (2024) A note on performance metrics for the kidney failure risk equation. Nephrology Dialysis Transplantation , Article gfae098. 10.1093/ndt/gfae098. (In press).

Text gfae098.pdf - Accepted Version Access restricted to UCL open access staff until 28 April 2025. Download (1MB)

Abstract

The kidney failure risk equation[1] (KFRE) is based on a Cox model first published in 2011 to quantify the risk of initiating kidney replacement therapy (KRT) over 2 and 5 years in patients with stage 3-5 CKD. It is commonly described as an accurate model[1-4] achieving excellent discrimination[5]. These descriptions are grounded in several evaluation metrics, most prominently the receiver operating characteristic (ROC) curve, the area under the ROC curve (AUROC), and Harrell’s C-index. However, these metrics exhibit limitations in conveying model performance in the context of highly censored populations with low event rate. In this letter we discuss the shortcomings of the aforementioned evaluation metrics and argue for a greater emphasis on more clinically relevant metrics such as the positive and negative predictive values (PPV and NPV).

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as CSVJSONXML

Archive Staff Only

View Item