EMR Coding with Semi-Parametric Multi-Head Matching Networks

Proceedings of NAACL-HLT 2018, pages 2081–2091 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics EMR Coding with Semi-Parametric Multi-Head Matching Networks Anthony Rios Department of Computer Science University of Kentucky Lexington, KY anthony.rios1@uky.edu Ramakanth Kavuluru Division of Biomedical Informatics University of Kentucky Lexington, KY ramakanth.kavuluru@uky.edu Abstract Coding EMRs with diagnosis and procedure codes is an indispensable task for billing, sec- ondary data analyses, and monitoring health trends. Both speed and accuracy of coding are critical. While coding errors could lead to more patient-side financial burden and mis- interpretation of a patient’s well-being, timely coding is also needed to avoid backlogs and additional costs for the healthcare facility. In this paper, we present a new neural network ar- chitecture that combines ideas from few-shot learning matching networks, multi-label loss functions, and convolutional neural networks for text classification to significantly outper- form other state-of-the-art models. Our eval- uations are conducted using a well known de- identified EMR dataset (MIMIC) with a vari- ety of multi-label performance measures. 1 Introduction Electronic medical record (EMR) coding is the process of extracting diagnosis and procedure codes from the digital record (the EMR) pertain- ing to a patient’s visit. The digital record is mostly composed of multiple textual narratives (e.g., dis- charge summaries, pathology reports, progress notes) authored by healthcare professionals, typ- ically doctors, nurses, and lab technicians. Hos- pitals heavily invest in training and retaining pro- fessional EMR coders to manually annotate all pa- tient visits by reviewing EMRs. Proprietary com- mercial software tools often termed as computer- assisted coding (CAC) systems are already in use in many healthcare facilities and were found to be helpful in increasing medical coder productiv- ity (Dougherty et al., 2013). Thus progress in au- tomated EMR coding methods is expected to di- rectly impact real world operations. In the US, the diagnosis and procedure codes used in EMR coding are from the Interna- tional Classification of Diseases (ICD) terminol- ogy (specifically the ICD-10-CM variant) as re- quired by the Health Insurance Portability and Accountability Act (HIPPA). ICD codes facili- tate billing activities, retrospective epidemiologi- cal studies, and also enable researchers to aggre- gate health statistics and monitor health trends. To code EMRs effectively, medical coders are ex- pected to have thorough knowledge of ICD-10- CM and follow a complex set of guidelines to code EMRs. For example, if a coder accidentally uses the code “heart failure” (ICD-10-CM code I50) in- stead of “acute systolic (congestive) heart failure” (ICD-10-CM code I50.21), then the patient may be charged substantially more1 causing significant unfair burden. Therefore, it is important for coders to have better tools at their disposal to find the most appropriate codes. Additionally, if coders become more efficient, hospitals may hire fewer coders to reduce their operating costs. Thus auto- mated coding methods are expected to help with expedited coding, cost savings, and error control. In this paper, we treat medical coding of EMR narratives as a multi-label text classification prob- lem. Multi-label classification (MLC) …

Content truncated to 500 words