DICE: Deep Significance Clustering for Outcome-Aware Stratification

TL;DR

DICE is a novel deep learning framework that jointly learns representations and clusters populations into outcome-aware groups, with statistical significance and optimized architecture, demonstrated on medical datasets for patient risk stratification.

Contribution

The paper introduces DICE, a new method combining representation learning, clustering, and neural architecture search for outcome-aware stratification in healthcare.

Findings

DICE outperforms baseline methods in clustering quality metrics.

DICE achieves higher AUC in outcome classification tasks.

DICE effectively stratifies patients by risk in real-world datasets.

Abstract

We present deep significance clustering (DICE), a framework for jointly performing representation learning and clustering for "outcome-aware" stratification. DICE is intended to generate cluster membership that may be used to categorize a population by individual risk level for a targeted outcome. Following the representation learning and clustering steps, we embed the objective function in DICE with a constraint which requires a statistically significant association between the outcome and cluster membership of learned representations. DICE further includes a neural architecture search step to maximize both the likelihood of representation learning and outcome classification accuracy with cluster membership as the predictor. To demonstrate its utility in medicine for patient risk-stratification, the performance of DICE was evaluated using two datasets with different outcome ratios extracted from real-world electronic health records. Outcomes are defined as acute kidney injury (30.4\%) among a cohort of COVID-19 patients, and discharge disposition (36.8\%) among a cohort of heart failure patients, respectively. Extensive results demonstrate that DICE has superior performance as measured by the difference in outcome distribution across clusters, Silhouette score, Calinski-Harabasz index, and Davies-Bouldin index for clustering, and Area under the ROC Curve (AUC) for outcome classification compared to several baseline approaches.