Description
Although phenotyping methods have been used to characterize cohorts for epidemiological studies, these methods are largely unused for driving clinical care processes. The objective of this project is to stratify patient populations for targeted clinical care through use of advanced phenotyping techniques. The approach is to integrate libraries for machine learning (ML) with workflows for creating clinical annotations to train ML. The outcome is to identify patients for selected therapeutic pathways, which will facilitate efficiency and scalability of care delivery to the patient population. We have developed and implemented a patient stratification process, and applied it in the domain of cardiovascular medicine at Partner Healthcare. Moreover we are working on implement the stratification process by extending i2b2 into a patient stratification platform (i2b2-PSP). Specifically we have developed extensions to implement key steps the patient stratification process including creation of ontologies to model and compute concepts needed at point-of-care, and a feedback process to allow the clinical staff to provide inputs for improving the phenotyping algorithms. In this session, we will demonstrate the process for a real world use case but using simulated patient data, and demonstrate the developed i2b2 plugins.
Learning Objective: After participating in this session, the learner should be better able to:
1. understand the challenges posed by bias and noise in electronic health record for constructing patient cohorts.
2. how use of standards and phenotyping can address the above challenges.
Authors:
Kavishwar Wagholikar (Presenter)
Harvard Medical School
Vishal Vernekar, Persistent Systems
Akshay Zagade, Persistent Systems
Yuri Ostrovsky, Persistent Systems
Shek-Wayne Chan, Partners Healthcare
Alyssa Goodson, Partners Healthcare
Ameet Pathak, Persistent Systems
corey Glynn, Partners Healthcare
Christopher Herrick, Partners Healthcare
Shawn Murphy, Partners Healthcare
Presentation Materials: