Unsupervised Approaches for Phenotyping Using Electronic Health Record Data

• An important goal of the Sentinel Innovation Center is to find ways to leverage machine learning approaches – including natural language processing and unsupervised learning – to identify health outcomes more efficiently. Computable phenotypes developed from electronic health record data could reduce both the cost and time needed to define and validate a gold standard. This webinar will focus on approaches for phenotyping using EHR data with a focus on annotation-free unsupervised classification methods such as PheNorm for phenotyping of medical conditions using electronic health record data. The discussion will consider the ways in which these approaches may be applicable to the Sentinel System.

Target Audience

• Medical informaticists, medical product safety and real world evidence researchers and regulators

Materials

• Unsupervised Approaches for Phenotyping Using Electronic Health Record Data