Automatic ICD Coding via Interactive Shared Representation Networks with Self-distillation Mechanism

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 5948–5957 August 1–6, 2021. ©2021 Association for Computational Linguistics 5948 Automatic ICD Coding via Interactive Shared Representation Networks with Self-distillation Mechanism Tong Zhou1,3,∗, Pengfei Cao1,2, Yubo Chen1,2, Kang Liu1,2, Jun Zhao1,2, Kun Niu3 Weifeng Chong4, Shengping Liu4 1 National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences 2 School of Artificial Intelligence University of Chinese Academy of Sciences 3 Beijing University of Posts and Telecommunications 4 Beijing Unisound Information Technology Co., Ltd {tongzhou21, niukun}@bupt.edu.cn {pengfei.cao,yubo.chen,kliu,jzhao}@nlpr.ia.ac.cn {chongweifeng,liushengping}@unisound.com Abstract The ICD coding task aims at assigning codes of the International Classification of Diseases in clinical notes. Since manual coding is very laborious and prone to errors, many methods have been proposed for the automatic ICD cod- ing task. However, existing works either ig- nore the long-tail of code frequency or the noisy clinical notes. To address the above is- sues, we propose an Interactive Shared Repre- sentation Network with Self-Distillation mech- anism. Specifically, an interactive shared rep- resentation network targets building connec- tions among codes while modeling the co- occurrence, consequently alleviating the long- tail problem. Moreover, to cope with the noisy text issue, we encourage the model to focus on the clinical note’s noteworthy part and extract valuable information through a self-distillation learning mechanism. Experimental results on two MIMIC datasets demonstrate the effective- ness of our method. 1 Introduction The International Classification of Diseases (ICD) is a healthcare classification system launched by the World Health Organization. It contains a unique code for each disease, symptom, sign and so on. Analyzing clinical data and monitoring health is- sues would become more convenient with the pro- motion of ICD codes (Shull, 2019) (Choi et al., 2016) (Avati et al., 2018). The ICD coding task aims at assigns proper ICD codes to a clinical note. It has drawn much attention due to the importance of ICD codes. This task is usually undertaken by experienced coders manually. However, the man- ually process is inclined to be labor-intensive and *Work was done during an internship at National Labora- tory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences. Figure 1: An example of automatic ICD coding task. error-prone (Adams et al., 2002). A knowledge- able coder with medical experience has to read the whole clinical note with thousands of words in medical terms and assigning multiple codes from a large number of candidate codes, such as 15,000 and 60,000 codes in the ninth version (ICD-9) and the tenth version (ICD-10) of ICD taxonomies. On the one hand, medical expert with specialized ICD coding skills is hard to train. On the other hand, it is a challenge task even for professional coders, due to the large candidate code set and tedious clin- ical notes. As statistics, the cost incurred by coding errors and the financial investment spent on improv- ing coding quality are estimated to be $25 billion per year …

Content truncated to 500 words