Neural Document Embeddings for Intensive Care Patient Mortality Prediction

TL;DR

This paper introduces a convolutional document embedding method for predicting ICU patient mortality from clinical notes, demonstrating significant improvements over previous text-based approaches, especially for post-discharge mortality prediction.

Contribution

The paper proposes a novel convolutional document embedding technique that outperforms existing methods in ICU mortality prediction from clinical notes.

Findings

Significant performance gains over latent topic and doc2vec methods

Notable improvement in post-discharge mortality prediction

Effective use of unstructured clinical text for mortality prediction

Abstract

We present an automatic mortality prediction scheme based on the unstructured textual content of clinical notes. Proposing a convolutional document embedding approach, our empirical investigation using the MIMIC-III intensive care database shows significant performance gains compared to previously employed methods such as latent topic distributions or generic doc2vec embeddings. These improvements are especially pronounced for the difficult problem of post-discharge mortality prediction.