Patient Trajectory Prediction: Integrating Clinical Notes with Transformers

TL;DR

This paper introduces a transformer-based method that combines clinical notes with structured EHR data to improve disease trajectory predictions, addressing the limitations of models that use only structured data.

Contribution

The paper presents a novel approach that integrates unstructured clinical notes into transformer models for enhanced disease prediction from EHRs.

Findings

Outperforms traditional models using only structured data.

Improves accuracy of disease trajectory predictions.

Demonstrates effectiveness on MIMIC-IV dataset.

Abstract

Predicting disease trajectories from electronic health records (EHRs) is a complex task due to major challenges such as data non-stationarity, high granularity of medical codes, and integration of multimodal data. EHRs contain both structured data, such as diagnostic codes, and unstructured data, such as clinical notes, which hold essential information often overlooked. Current models, primarily based on structured data, struggle to capture the complete medical context of patients, resulting in a loss of valuable information. To address this issue, we propose an approach that integrates unstructured clinical notes into transformer-based deep learning models for sequential disease prediction. This integration enriches the representation of patients' medical histories, thereby improving the accuracy of diagnosis predictions. Experiments on MIMIC-IV datasets demonstrate that the proposed approach outperforms traditional models relying solely on structured data.