On the calibration of survival models with competing risks

Julie Alberge, Tristan Haugomat, Gaël Varoquaux, Judith Abécassis

TL;DR: New calibration metrics for models that can handle competing risks.

Abstract: Survival analysis deals with modeling the time until an event occurs, and accurate probability estimates are crucial for decision-making, particularly in the competing-risks setting where multiple events are possible. While recent work has addressed calibration in standard survival analysis, the competing-risks setting remains under-explored as it is harder (the calibration applies to both probabilities across classes and time horizon). We show that existing calibration measures are not suited to the competing-risk setting and that recent models do not give well-behaved probabilities. To address this, we introduce a dedicated framework with two novel calibration measures that are minimized for oracle estimators (*i.e.*, both measures are proper). We also introduce some methods to estimate, test, and correct the calibration. Our recalibration methods yield better probabilities while preserving discrimination.

Code Dataset Promise: Yes

Code Dataset Url: https://github.com/soda-inria/hazardous/

Format Confirmation: I agree that I have read and followed the formatting instructions for the camera ready version.

Submission Number: 782