Deep Kernel Learning for Stratifying Glaucoma Trajectories

TL;DR

This paper introduces a deep kernel learning model using transformer-based features and Gaussian Processes to stratify glaucoma patients by progression risk from electronic health records, aiding targeted clinical interventions.

Contribution

The novel DKL architecture effectively decouples disease progression from current severity, identifying high-risk patients with worsening trajectories despite better current health status.

Findings

Successfully identified three distinct patient subgroups.

The model distinguishes progression risk from disease severity.

High-risk group shows worsening despite better visual acuity.

Abstract

Effectively stratifying patient risk in chronic diseases like glaucoma is a major clinical challenge. Clinicians need tools to identify patients at high risk of progression from sparse and irregularly-sampled electronic health records (EHRs). We propose a novel deep kernel learning (DKL) architecture that leverages a Gaussian Process (GP) backend. The GP's kernel is defined by a transformer-based feature extractor applied to clinical-BERT embeddings to model glaucoma patient trajectories from multimodal EHR data. Our method successfully identifies three clinically distinct patient subgroups. Crucially, the model learns to decouple disease progression from current severity, identifying a high-risk group with a worsening trajectory despite having better average visual acuity than a second, stably poor group. This reveals that the model learns to identify progression risk rather than just the current disease state. This ability to stratify patients based on their risk trajectory progression offers a powerful tool for clinical decision support, enabling targeted interventions for high-risk individuals and improving the management of glaucoma care.