Interpretable machine learning for time-to-event prediction in medicine and healthcare

TL;DR

This paper introduces methods for interpreting time-to-event machine learning models in medicine, enabling bias detection, biomarker discovery, and model debugging through novel explanations and comprehensive datasets.

Contribution

It presents new interpretability techniques for survival models, including time-dependent feature effects and global importance explanations, supported by open datasets and code.

Findings

Post-hoc interpretation reveals biases in hospital stay predictions.

Multi-omics feature importance aids in understanding cancer survival.

Open datasets and tools facilitate future explainable survival analysis.

Abstract

Time-to-event prediction, e.g. cancer survival analysis or hospital length of stay, is a highly prominent machine learning task in medical and healthcare applications. However, only a few interpretable machine learning methods comply with its challenges. To facilitate a comprehensive explanatory analysis of survival models, we formally introduce time-dependent feature effects and global feature importance explanations. We show how post-hoc interpretation methods allow for finding biases in AI systems predicting length of stay using a novel multi-modal dataset created from 1235 X-ray images with textual radiology reports annotated by human experts. Moreover, we evaluate cancer survival models beyond predictive performance to include the importance of multi-omics feature groups based on a large-scale benchmark comprising 11 datasets from The Cancer Genome Atlas (TCGA). Model developers can use the proposed methods to debug and improve machine learning algorithms, while physicians can discover disease biomarkers and assess their significance. We hope the contributed open data and code resources facilitate future work in the emerging research direction of explainable survival analysis.