Emergency Department Decision Support using Clinical Pseudo-notes

TL;DR

This paper introduces MEME, a novel method that converts multimodal EHR data into pseudo-notes to leverage pretrained models for improved emergency department decision support.

Contribution

The paper presents MEME, a new approach that serializes EHR data into text, enabling better representations and utilization of foundation models for clinical decision support.

Findings

MEME outperforms traditional ML and existing foundation models.

MEME improves decision support accuracy across multiple hospital systems.

The approach effectively encodes multimodal EHR data for clinical tasks.

Abstract

In this work, we introduce the Multiple Embedding Model for EHR (MEME), an approach that serializes multimodal EHR tabular data into text using pseudo-notes, mimicking clinical text generation. This conversion not only preserves better representations of categorical data and learns contexts but also enables the effective employment of pretrained foundation models for rich feature representation. To address potential issues with context length, our framework encodes embeddings for each EHR modality separately. We demonstrate the effectiveness of MEME by applying it to several decision support tasks within the Emergency Department across multiple hospital systems. Our findings indicate that MEME outperforms traditional machine learning, EHR-specific foundation models, and general LLMs, highlighting its potential as a general and extendible EHR representation strategy.