Clinical Note Section Identification Using Transfer Learning.

In the healthcare industry, clinical records written by doctors, during the consultation with patients, are often in an unstructured fashion. The standard clinical practice recommends the notes to be written in a structured format such as SOAP. Through NLP Transfer Learning, clinical note-taking application can be trained to intelligently recognize sections and subsections within the clinical notes. In this study, we propose several Transfer Learning models based on clinical contextual embeddings for the classification of clinical notes into several major SOAP sections. We validate our models using the clinical notes from the i2b2 2010 dataset. We show that intelligent note-taking applications can be easily developed using the proposed Transfer Learning models, and their integration into Hospital Information Systems will pave the way for effective clinical note generation and alleviation of physician burnout.