Currently submitted to: JMIR Medical Informatics
Date Submitted: Mar 27, 2026
Open Peer Review Period: Apr 17, 2026 - Jun 12, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
A Risk-Aware Pseudotime Approach for Modeling Heart Failure Progression: A Geometric Analysis of Longitudinal Electronic Health Record Data
ABSTRACT
Background:
Risk stratification using longitudinal electronic health records (EHRs) remains challenging due to the complex nature of disease trajectories. While deep sequential models have shown promising performance, the inherent complexity of these classifiers often limits their interpretability and clinical trustworthinessin in high-stakes medical decision-making.
Objective:
This study proposes a representation learning framework to capture clinically meaningful disease progression within a geometric trajectory space, thereby improving model transparency and reducing reliance on complex, non-interpretable classifiers.
Methods:
We developed a framework that integrates comorbidity structures and clinical deviation vectors to map clinically meaningful disease progression at the representation level. The learned representations were evaluated using both linear (LR) and non-linear (XGBoost) classifiers under rigorous out-of-fold validation. Systematic ablation analyses were performed to disentangle the contributions of the geometric representation design from classifier choice.
Results:
Across multiple experimental settings, simple linear classifiers achieved performance comparable to gradient-boosted tree models, yielding a peak AUPRC of .651 (overall) and .825 (high-risk subgroup). These findings indicate that the proposed representations effectively internalize non-linear disease interactions, resulting in a linearly separable risk space. Visualization of disease trajectories further demonstrates clear clinical face validity, with high-risk and low-risk patient groups exhibiting distinct geometric patterns consistent with domain knowledge.
Conclusions:
Our results suggest that clinically meaningful disease trajectories can be effectively captured through geometric representation design, significantly reducing the dependence on complex classifiers for risk stratification. This representation-driven approach offers a transparent, stable, and interpretable framework for longitudinal EHR analysis, with high potential for deployment in clinical risk stratification systems. Clinical Trial: N/A
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.