Contextualized Medication Information Extraction Using Transformer-based Deep Learning Architectures

Details

• Alternative Title:
  J Biomed Inform
• Personal Author:
  Chen, Aokun ; Yu, Zehao ; Yang, Xi ; Guo, Yi ; Bian, Jiang ; Wu, Yonghui
• Description:
  Objective
  
  To develop a natural language processing (NLP) system to extract medications and contextual information that help understand drug changes. This project is part of the 2022 n2c2 challenge.
  
  Materials and methods
  
  We developed NLP systems for medication mention extraction, event classification (indicating medication changes discussed or not), and context classification to classify medication changes context into 5 orthogonal dimensions related to drug changes. We explored 6 state-of-the-art pretrained transformer models for the three subtasks, including GatorTron, a large language model pretrained using >90 billion words of text (including >80 billion words from >290 million clinical notes identified at the University of Florida Health). We evaluated our NLP systems using annotated data and evaluation scripts provided by the 2022 n2c2 organizers.
  
  Results
  
  Our GatorTron models achieved the best F1-scores of 0.9828 for medication extraction (ranked 3rd), 0.9379 for event classification (ranked 2nd), and the best micro-average accuracy of 0.9126 for context classification. GatorTron outperformed existing transformer models pretrained using smaller general English text and clinical text corpora, indicating the advantage of large language models.
  
  Conclusion
  
  This study demonstrated the advantage of using large transformer models for contextual medication information extraction from clinical narratives.
  
  
• Subjects:
  Deep Learning Information Storage And Retrieval Natural Language Processing
• Keywords:
  Clinical Natural Language Processing Deep Learning Medication Information Extraction Named Entity Recognition Text Classification
• Source:
  J Biomed Inform. 142:104370
• Pubmed ID:
  37100106
• Pubmed Central ID:
  PMC10980542
• Document Type:
  Journal Article
• Funding:
  R56 AG069880/AG/NIA NIH HHSUnited States/ ; U18 DP006512/DP/NCCDPHP CDC HHSUnited States/ ; R01 MH121907/MH/NIMH NIH HHSUnited States/ ; R01 CA246418/CA/NCI NIH HHSUnited States/ ; U18DP006512/ACL/ACL HHSUnited States/ ; R21 CA253394/CA/NCI NIH HHSUnited States/ ; R21 CA245858/CA/NCI NIH HHSUnited States/ ; R01 DA050676/DA/NIDA NIH HHSUnited States/ ; R21 AG068717/AG/NIA NIH HHSUnited States/
• Volume:
  142
• Collection(s):
  CDC Public Access
• Main Document Checksum:
  urn:sha256:25c2e4d7e342d5a981efed5fe79128208908ce44f7ad354e79ab04e875e10360
• Download URL:
  https://stacks.cdc.gov/view/cdc/153621/cdc_153621_DS1.pdf
• File Type:
  [PDF - 102.77 KB ]