Reproduce “MIMICause: Representation and automatic extraction of causal relation types from clinical notes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Vivek Khetan, Md Imbesat Hassan Rizvi, Jessica Huber, Paige Bartusiak, Bogdan Sacaleanu, Andrew Fano
MIMICause, a new dataset of 2714 de-identified clinical notes, was created and meticulously annotated with nine fine-grained causal relation types to enable automatic extraction of nuanced causal relationships, achieving a baseline Macro F1 score of 0.56 with a Clinical-BERT model augmented with entity context.
Key insights
- Distinguishing between exclusive (Cause, Prevent) and contributory (Enable, Hinder) causal relationships is crucial for accurately representing the multi-factorial nature of clinical events.
- Explicitly providing models with the context of interacting entities significantly improves performance in causal relation extraction, as demonstrated by the jump in F1 score from 0.27 to 0.56 with entity context.
- Domain-specific pre-trained language models like Clinical-BERT offer an advantage over general BERT models for understanding the complex language and specialized terminology of clinical notes in this task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.