Problem
Research questions and friction points this paper is trying to address.
patient care trajectories
temporal irregularity
class imbalance
healthcare utilization
event prediction
Modeling Patient Care Trajectories with Transformer Hawkes Processes
🤖 AI Summary
This work addresses the challenges of irregular timestamps and severe class imbalance in patient medical event trajectories by proposing a continuous-time modeling framework that integrates Transformer architecture with a Hawkes process to jointly predict future event types and their occurrence times. To enhance sensitivity to rare yet clinically critical events without altering the original data distribution, the model incorporates an inverse square root class weighting strategy. Coupled with attention mechanisms and a tailored loss function, the approach demonstrates significant performance gains over baseline models on real-world clinical data. Beyond improved predictive accuracy, the method yields interpretable insights of direct clinical relevance, facilitating the identification of high-risk patients.
📝 Abstract
Patient healthcare utilization consists of irregularly time-stamped events, such as outpatient visits, inpatient admissions, and emergency encounters, forming individualized care trajectories. Modeling these trajectories is crucial for understanding utilization patterns and predicting future care needs, but is challenging due to temporal irregularity and severe class imbalance. In this work, we build on the Transformer Hawkes Process framework to model patient trajectories in continuous time. By combining Transformer-based history encoding with Hawkes process dynamics, the model captures event dependencies and jointly predicts event type and time-to-event. To address extreme imbalance, we introduce an imbalance-aware training strategy using inverse square-root class weighting. This improves sensitivity to rare but clinically important events without altering the data distribution. Experiments on real-world data demonstrate improved performance and provide clinically meaningful insights for identifying high-risk patient populations.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Transformer Hawkes Process
continuous-time modeling
class imbalance
patient care trajectories
inverse square-root weighting
🔎 Similar Papers
No similar papers found.
