[Un]supervised Patient Characterization using Electronic Health Records
Data from Electronic Health Records (EHR) — also known as Electronic Medical Records (EMR) — offer huge potentials for patient characterization using advanced learning algorithms. The reason is EHRs uniquely contain labeled data that were labeled with highest precision — as the labels (e.g., diagnoses codes) are created by very well trained professionals (i.e., physicians). At the MGH Laboratory of Computer Science we apply supervised and unsupervised classification algorithms to EHR data to characterize patients with similar health outcomes. In this talk, I will report our experience using Deep Learning for patient characterization and how it compares with some of the other classification methods we have used.
Hossein Estiri, PhD, is a research fellow with the Massachusetts General Hospital Laboratory of Computer Science at Harvard Medical School and an informatics training fellow of the National Library of Medicine. Dr. Estiri’s research involves designing data-driven systems for clinical decision making and healthcare policy. His recent work has focused on applying Data Science methodologies to data from Electronic Health Records in order to design systems that characterize patients and evaluate data quality.