2015 Data Science Workshop

Wed, April 29, 2015
Location: Fisher Conference Center, Arrillaga Alumni Center

"Use of Electronic Phenotyping and Machine Learning Algorithms to Identify Familial Hypercholesterolemia Patients in Electronic Health Records"

Abstract:

FIND FH (Flag, Identify, Network and Deliver for Familial Hypercholesterolemia) aims to pioneer new techniques for the identification of individuals with Familial Hypercholesterolemia (FH) within electronic health records (EHRs). FH is a common but vastly underdiagnosed, inherited form of high cholesterol and coronary heart disease that is potentially devastating if undiagnosed but can be ameliorated with early identification and proactive treatment. Traditionally, patients with a phenotype (such as FH) are identified through rule-based definitions whose creation and validation are time consuming. Machine learning approaches to phenotyping are limited by the paucity of labeled training datasets. In this project, demonstrate the feasibility of utilizing noisy labeled training sets to learn phenotype models from the patient's clinical record. We will search both structured and unstructured data with in EHRs to identify possible FH patients. Individuals with possible FH will be “flagged” and contacted in a HIPAA compliant manner to encourage guideline-based screening and therapy.Algorithms developed will be broadly applicable to several different EHR platforms and the principles can be applied to multiple conditions thereby extending the utility of this approach.The project is in partnership with the FH Foundation (www.thefhfoundation.org), a non-profit organization founded and led by FH patients that is dedicated to improving the awareness and treatment of FH.