Development of drug adverse event detection method from electronic medical record unstructured data

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K09171
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Medical and hospital managemen
• Research Institution: Osaka University
• Principal Investigator: Takeda Toshihiro 大阪大学, 医学部附属病院, 准教授 (70506493)
• Co-Investigator(Kenkyū-buntansha): 三原 直樹 国立研究開発法人国立がん研究センター, 中央病院, 部長 (20379192) ; 松村 泰志 大阪大学, 医学系研究科, 教授 (90252642)
• Research Collaborator: Shimai Yoshie ; Sugimoto Kento
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: 薬剤性有害事象 / 間質性肺炎 / テキストマイニング / 機械学習

Outline of Final Research Achievements

This study aims to detect adverse drug events using electronic medical record data.
Keywords were extracted by text mining from the diagnostic imaging report described in free text, and interstitial pneumonia could be detected with high accuracy from the frequency of keyword appearance. From The incidence rate of drug-induced interstitial pneumonia could be calculated from the chronological relationship between drug administration period and the onset of interstitial pneumonia.
Detection of lung cancer, breast cancer and esophagus cancer reports from imaging diagnostic reports was performed by the same keyword method. However, the accuracy was not high. This is thought to be due to the fact that interstitial pneumonia has characteristic keywords, whereas cancer has less characteristic keywords. To solve this problem, extraction of the findings by machine learning was needed.

Free Research Field

医療情報学

Academic Significance and Societal Importance of the Research Achievements

薬剤性有害事象は治験、市販後調査において多額の費用を用いて実施されている。
電子カルテデータを用いて薬剤性有害事象を検出できれば、安価に有害事象の発生率が算出できる。また、リアルワールドデータを解析することで、治験や市販後調査で把握できなかった、新たな副作用を発見できる可能性がある。さらに、継続的に有害事象を検出する仕組みを開発すれば、新薬の有害事象をより早期に発見できる可能性がある。