- This event has passed.
Webcast: Medical Device Surveillance Using “Big Data” Methods
January 5, 2016 @ 15:00 - 17:00
Listen to a recording of the webinar: Play recording (1 hour 27 minutes)
Associated Document: Safety Signal Detection Project 2016Jan5 16-01-05 FINAL
Joseph Ross, MD, MHS – Yale Center for Outcomes Research and Evaluation (CORE)
Craig Parzynski, MS – Yale Center for Outcomes Research and Evaluation (CORE)
Jonathan Bates, PhD – Yale Center for Outcomes Research and Evaluation (CORE)
Shu-Xia Li, PhD – Yale Center for Outcomes Research and Evaluation (CORE)
The surveillance of medical devices is intended to provide critical information to all relevant stakeholders about device safety, long-term product performance, and effectiveness in improving patient outcomes. However, we are far from this ideal system. Yale has been working in collaboration with the FDA to better understand whether use of big data analytic methods, including the use of unstructured data and machine learning techniques, enhances post-market surveillance of medical devices. This project will increase our knowledge of signal detection methodologies that are capable of utilizing large, high dimensional health care data sets. This information is intended to contribute to the ongoing development of the National Medical Device Surveillance System infrastructure.
The specific aim of this project is to employ big data methods for signal detection of device-related complications after implantable-cardioverter defibrillator (ICD) therapy, accounting for patient and procedural characteristics at implantation. Working with the MDEpiNet Methodology Center and others within the MDEpiNet community, Yale has partnered with the American College of Cardiology (ACC) and is utilizing their National Cardiovascular Data Registry ICD Registry linked to Medicare administrative claims as the dataset for this project.
In this forum, we will:
- Provide study methods and analytic approaches, including explanation of the endpoints identified and decision on appropriate ‘comparators’
- Share preliminary results from the novel and traditional analytic methods
- Describe options for evaluating the quality of propensity score matching
- Invite the MDEpiNet community to provide feedback and suggest alternate approaches