Overcoming Dependent Censoring in the Evaluation of Survival Models

TL;DR

This paper introduces three new survival analysis metrics based on Archimedean copulas that accurately evaluate models under dependent censoring, addressing biases in traditional metrics.

Contribution

The paper proposes novel evaluation metrics and a framework for semi-synthetic data generation to handle dependent censoring in survival model assessment.

Findings

Proposed metrics outperform traditional ones under dependent censoring.

Framework enables realistic semi-synthetic data generation.

Metrics provide more accurate error estimates in experiments.

Abstract

Conventional survival metrics, such as Harrell's concordance index (CI) and the Brier Score, rely on the independent censoring assumption for valid inference with right-censored data. However, in the presence of so-called dependent censoring, where the probability of censoring is related to the event of interest, these metrics can give biased estimates of the underlying model error. In this paper, we introduce three new evaluation metrics for survival analysis based on Archimedean copulas that can account for dependent censoring. We also develop a framework to generate realistic, semi-synthetic datasets with dependent censoring to facilitate the evaluation of the metrics. Our experiments in synthetic and semi-synthetic data demonstrate that the proposed metrics can provide more accurate estimates of the model error than conventional metrics under dependent censoring.