Semantic-Level New Information Identification in Electronic Health Records Using Text-Mining Techniques

Abstract

Electronic health records (EHRs) are widely used in healthcare systems to store and transmit patients’ health records. They have many advantages, such as saving space, increasing efficiency, and facilitating communication. However, they also have a major drawback: information redundancy. Healthcare professionals often use copy and paste to write clinical notes, which leads to excessive similarity and low diversity in EHRs. This impairs the readability and quality of EHRs and hinders decision making. To address this problem, this study proposes a text-mining approach to identify new information at semantic-level in EHRs. Unlike previous studies that focused on word-level identification, we use concept occurrence and concept similarity score methods to annotate new information at semantic-level and evaluate them with gold standards. The experimental evaluation demonstrates that the method proposed in this study achieves an F1-score ranging from 78.57 to 80.31 under various parameter combinations. The proposed method enables healthcare professionals to read EHRs more efficiently and make more informed decisions.