Unstructured Electronic Health Record

Unstructured electronic health records (EHRs), such as clinical notes and reports, contain rich clinical information not readily accessible through structured data alone. Current research focuses on leveraging large language models (LLMs), particularly transformer-based architectures, to extract meaningful insights from this unstructured data, improving tasks like cohort retrieval, risk prediction, and diagnostic support. This work aims to enhance the accuracy and interpretability of clinical prediction models and improve efficiency in various healthcare applications by integrating information from both structured and unstructured EHR data sources. The ultimate goal is to unlock the wealth of knowledge hidden within unstructured EHRs for improved patient care and medical research.