CACHE

This repo contains our code for paper Counterfactual and Factual Reasoning over Hypergraphs for Interpretable Clinical Predictions on EHR, in Proceedings of 2nd Machine Learning for Health symposium 2022 (ML4H 2022).

Model Framework

Data

In order to facilitate the reproducibility, we provide two toy datasets in data. Note that although they're named as mimic3 and cradle (the two datasets we mentioned in our paper), they're in fact randomly generated due to the privacy issue. We include them in this repo only to show the format of the two datasts we used. Thus, their experimental results should not reflect the performance we report in the paper.

Package

• PyTorch 1.4
• python 3.7
• tqdm
• torch-scatter 2.0.4
• torch-sparse 0.6.0
• torch-cluster 1.5.2
• torch-geometric 1.6.3
• sklearn

Run the Code

Please use run.sh in src to run the code for the two toy datasets in data. It runs four experiments:

• CACHE for MIMIC-III dataset
• CACHE for CRADLE dataset
• vanilla backbone model for MIMIC-III dataset
• vanilla backbone model for CRADLE dataset

Citation

If you find this paper useful for your research, please cite the following in your publication. Thanks!

@inproceedings{xu2022counterfactual,
  title={Counterfactual and Factual Reasoning over Hypergraphs for Interpretable Clinical Predictions on EHR},
  author={Xu, Ran and Yu, Yue and Zhang, Chao and Ali, Mohammed K and Ho, Joyce C and Yang, Carl},
  booktitle={Machine Learning for Health},
  pages={259--278},
  year={2022},
  organization={PMLR}
}

Acknowledgement

We would like to thank the authors from AllSet for their open-source efforts.