Leveraging Medical Literature for Section Prediction in Electronic Health Records

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 4864–4873, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 4864 Leveraging Medical Literature for Section Prediction in Electronic Health Records Sara Rosenthal IBM Research Yorktown Heights, NY, USA sjrosenthal@us.ibm.com Ken Barker IBM Research Yorktown Heights, NY, USA kjbarker@us.ibm.com Jason Zhicheng Liang∗ Rensselaer Polytechnic Institute Troy, New York, USA liangz4@rpi.edu Abstract Electronic Health Records (EHRs) contain both structured content and unstructured (text) content about a patient’s medical history. In the unstructured text parts, there are common sections such as Assessment and Plan, So- cial History, and Medications. These sec- tions help physicians find information easily and can be used by an information retrieval system to return specific information sought by a user. However, it is common that the exact format of sections in a particular EHR does not adhere to known patterns. There- fore, being able to predict sections and headers in EHRs automatically is beneficial to physi- cians. Prior approaches in EHR section pre- diction have only used text data from EHRs and have required significant manual annota- tion. We propose using sections from med- ical literature (e.g., textbooks, journals, web content) that contain content similar to that found in EHR sections. Our approach uses data from a different kind of source where la- bels are provided without the need of a time- consuming annotation effort. We use this data to train two models: an RNN and a BERT- based model. We apply the learned models along with source data via transfer learning to predict sections in EHRs. Our results show that medical literature can provide helpful su- pervision signal for this classification task. 1 Introduction Electronic Health Records (EHRs) are an impor- tant tool used by physicians that contain an abun- dance of information about each patient. In addi- tion to aiding the physician in providing patient care, EHRs can be used for downstream tasks ∗Work completed during internship at IBM Research such as predicting patient outcome, representa- tion learning, and information extraction (Shickel et al., 2017). All of these tasks can be applied to the unstructured text parts of an EHR. Further- more, they would benefit from the section struc- ture to pinpoint the likely location where the infor- mation should be found (e.g. drug codes are more likely to be in the Medication section). However, when physicians edit the unstructured text parts of an EHR, there are no set rules for indicating sections and format is not strictly enforced, nor is there any markup (e.g. XML). An added difficulty is that the formatting is not uniform across EHRs. This is even more common when the EHRs come from different hospitals. For example, one EHR may have the section heading “Assessment and Plan” where another may use “A&P”. As an added challenge there may also be nested sections, such as information about a specific medication. Being able to predict sections and headers …

Content truncated to 500 words