Improving clinical outcome predictions using convolution over medical entities with multimodal learning

Highlights：

• We propose a multimodal method for the use of clinical notes, in addition to other features, to predict clinical outcomes.

• The proposed method consistently outperforms the baselines.

• We also report the performance of multiple clinical note embedding methods for outcome prediction.

• Besides mortality and length-of-stay, the proposed method can also be extended to other outcomes as well.