CICD-Coder: Chinese EMRs Based ICD Coding With Multi-axial Supported Clinical Evidence

心心 尤

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Automatic ICD coding, Chinese EHR, Evidence-based medicine, Knowledge of multi-axes, Prompt-tuning

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Abstract:

Although automatic ICD coding has achieved some success in English, there still exist significant challenges for the Chinese electronic medical records(EMRs) based ICD coding task. The first problem is the difficulty of extracting disease code-related information from Chinese EMRs due to the concise writing style and specific internal structure content of EMRs. The second problem is that previous methods have not exploited the disease-based multi-axial knowledge and are neither associated with the corresponding clinical evidence, resulting in inaccuracy in disease coding and lack of interpretability.
In this paper, we develop a novel automatic ICD coding framework CICD-Coder for the Chinese EMRs based ICD coding task. In the presented framework, we first investigate the multi-axes knowledge (crucial for the ICD coding) of the given disease and then retrieve corresponding clinical evidence for the disease-based multi-axes knowledge from the whole content of EMRs. Finally, we present an evaluation module based on the masked language modeling strategy to ensure each knowledge under the axis of the recommended ICD code is supported by reliable evidence. The experiments are conducted on a large-scale Chinese EMRs dataset collected from varying hospitals and the results verify the effectiveness, reliability, and interpretability of our proposed ICD coding method.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 4545