Representation Learning of EHR Data via Graph-Based Medical Entity Embedding

TL;DR

This paper introduces ME2Vec, a graph-based embedding framework that learns low-dimensional representations of medical entities in EHR data, improving disease diagnosis prediction from real-world clinical data.

Contribution

The paper presents a novel graph embedding framework, ME2Vec, tailored for different EHR entities, enhancing the quality of medical data representations.

Findings

ME2Vec outperforms baseline methods in disease diagnosis prediction

Effective embedding of medical services, doctors, and patients

Demonstrated on real-world clinical data

Abstract

Automatic representation learning of key entities in electronic health record (EHR) data is a critical step for healthcare informatics that turns heterogeneous medical records into structured and actionable information. Here we propose ME2Vec, an algorithmic framework for learning low-dimensional vectors of the most common entities in EHR: medical services, doctors, and patients. ME2Vec leverages diverse graph embedding techniques to cater for the unique characteristic of each medical entity. Using real-world clinical data, we demonstrate the efficacy of ME2Vec over competitive baselines on disease diagnosis prediction.