Cluster-weighted modeling of lifetime hierarchical data for profiling COVID-19 heart failure patients

2025-07-16

AI Summary

This study addresses survival heterogeneity among hospitalized heart failure (HF) patients with COVID-19 in Lombardy, Italy. We propose a novel hierarchical mixture survival model that innovatively incorporates a shared frailty term to jointly disentangle patient-level latent subgroups—defined by clinical characteristics—and hospital-level random effects. By integrating local distributional modeling with random effects, the framework enables cluster-weighted inference. Parameter estimation under right-censoring is performed via two EM variants: the Classification EM (CEM) and Stochastic EM (SEM). Results identify clinically distinct patient subgroups with significantly divergent survival patterns, quantify respiratory status as a key prognostic factor, and—uniquely—systematically assess inter-hospital variation in clinical management efficacy for HF-COVID-19 patients. The findings provide empirical support for risk-stratified clinical decision-making and evidence-based allocation of healthcare resources.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Mixed Discrete/Continuous Search
• Reasoning under Uncertainty: Probabilistic Programming

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Systems and Infrastructure for Web, Mobile and WoT: Web applications in cross-disciplinary domains and verticals such as mixed reality, smart cities, and digital health
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

This study investigates the heterogeneity in survival times among COVID-19 patients with Heart Failure (HF) hospitalized in the Lombardy region of Italy during the pandemic. To address this, we propose a novel mixture model for right-censored lifetime data that incorporates random effects and allows for local distributions of the explanatory variables. Our approach identifies latent clusters of patients while estimating component-specific covariate effects on survival, taking into account the hierarchical structure induced by the healthcare facility. Specifically, a shared frailty term, unique to each cluster, captures hospital-level variability enabling a twofold decoupling of survival heterogeneity across both clusters and hierarchies. Two EM-based algorithms, namely a Classification EM (CEM) and a Stochastic EM (SEM), are proposed for parameter estimation. The devised methodology effectively uncovers latent patient profiles, evaluates within-cluster hospital effects, and quantifies the impact of respiratory conditions on survival. Our findings provide new information on the complex interplay between the impacts of HF, COVID-19, and healthcare facilities on public health, highlighting the importance of personalized and context-sensitive clinical strategies.

Problem

• Model survival heterogeneity in COVID-19 heart failure patients
• Identify latent patient clusters with hierarchical healthcare effects
• Quantify impact of respiratory conditions on patient survival

Innovation

• Cluster-weighted modeling with random effects
• EM-based algorithms for parameter estimation
• Shared frailty term for hospital variability