On Ranking in Survival Analysis: Bounds on the Concordance Index
Abstract
In this paper, we show that classical survival analysis involving censored data can naturally be cast as a ranking problem. The concordance index (CI), which quantifies the quality of rankings, is the standard performance measure for model \emph{assessment} in survival analysis. In contrast, the standard approach to \emph{learning} the popular proportional hazard (PH) model is based on Cox's partial likelihood. In this paper we devise two bounds on CI--one of which emerges directly from the properties of PH models--and optimize them \emph{directly}. Our experimental results suggest that both methods perform about equally well, with our new approach giving slightly better results than the Cox's method. We also explain why a method designed to maximize the Cox's partial likelihood also ends up (approximately) maximizing the CI.
Cite
Text
Steck et al. "On Ranking in Survival Analysis: Bounds on the Concordance Index." Neural Information Processing Systems, 2007.Markdown
[Steck et al. "On Ranking in Survival Analysis: Bounds on the Concordance Index." Neural Information Processing Systems, 2007.](https://mlanthology.org/neurips/2007/steck2007neurips-ranking/)BibTeX
@inproceedings{steck2007neurips-ranking,
title = {{On Ranking in Survival Analysis: Bounds on the Concordance Index}},
author = {Steck, Harald and Krishnapuram, Balaji and Dehing-oberije, Cary and Lambin, Philippe and Raykar, Vikas C.},
booktitle = {Neural Information Processing Systems},
year = {2007},
pages = {1209-1216},
url = {https://mlanthology.org/neurips/2007/steck2007neurips-ranking/}
}