Understanding the Clinical Context of Medication Change Events in Clinical Narratives using Pre-trained Clinical Language Models

The ability to understand medication events in clinical narratives is crucial to gaining a comprehensive picture of the patient's medication history. There has been some prior research on identifying medication changes in clinical notes. However, because clinical documentation is longitudinal and narrative, capturing medication changes without the necessary clinical context is not sufficient in real-world applications, such as medication reconciliation and medication timeline generation. In this research, we propose a framework to use multiple clinical-based Bidirectional Encoder Representations from Transformers (BERT) for Contextualized Medication Event Extraction, which is a task to capture the multi-dimensional context of medication changes documented in clinical notes. In addition, the BERT models in the proposed framework infused clinical context-sensitive features into the method to learn the text information around the descriptions of medication. The experiments are conducted by using Contextualized Medication Event Dataset, and the results demonstrate that the proposed method outperforms ClinicalBERT, the state-of-the-art BERT model in the previous study.