Extracting Angina Symptoms from Clinical Notes Using Pre-Trained Transformer Architectures

Anginal symptoms can connote increased cardiac risk and a need for change in management. This study evaluated the potential to extract these from physician notes using the Bidirectional Encoder from Transformers model fine-tuned on a domain-specific corpus. The history of present section of 459 expert annotated primary care physician notes from patients referred for cardiac testing without known atherosclerotic disease were included. Notes were annotated for positive and mentions of chest pain and shortness of breath characterization. The demonstrate high sensitivity and specificity for the detection of chest or discomfort, substernal chest pain, shortness of breath, and dyspnea on. Small sample size limited extracting factors related to provocation palliation of chest pain. This study provides a promising starting point the natural language processing of physician notes to characterize actionable anginal symptoms.