Quantifying and Estimating the Predictive Accuracy for Censored Time-to-Event Data with Competing Risks

TL;DR

This paper develops a nonparametric framework to accurately estimate the predictive performance of models for censored time-to-event data with competing risks, improving bias and robustness over existing methods.

Contribution

It introduces a unified, nonparametric estimation method for discrimination and calibration metrics in competing risks, adaptable to various loss functions and robust to model misspecification.

Findings

Proposed estimator is unbiased, efficient, and robust.

Method effectively accounts for censoring in predictive accuracy.

Applied to kidney disease data, demonstrating practical utility.

Abstract

This paper focuses on quantifying and estimating the predictive accuracy of prognostic models for time-to-event outcomes with competing events. We consider the time-dependent discrimination and calibration metrics, including the receiver operating characteristics curve and the Brier score, in the context of competing risks. To address censoring, we propose a unified nonparametric estimation framework for both discrimination and calibration measures, by weighting the censored subjects with the conditional probability of the event of interest given the observed data. We demonstrate through simulations that the proposed estimator is unbiased, efficient and robust against model misspecification in comparison to other methods published in the literature. In addition, the proposed method can be extended to time-dependent predictive accuracy metrics constructed from a general class of loss functions. We apply the methodology to a data set from the African American Study of Kidney Disease and Hypertension to evaluate the predictive accuracy of a prognostic risk score in predicting end-stage renal disease (ESRD), accounting for the competing risk of pre-ESRD death.