Prospective, active safety surveillance of medical devices offers significant advantages over traditional passive event reporting strategies in identifying previously unexpected safety risks arising from device failure risks or damage/sabotage from chemical, biologic, radiologic or nuclear (CBRN) events. However, methodological, data availability, and access of data owners to validated tools with which to monitor safety all challenge the development of a coordinated national approach to active medical device surveillance. This proposal seeks to extend our prior work in the development and validation of an active medical device safety surveillance system, by developing methodologies to address learning effects and to implement simplify risk adjustment model development. These efforts will permit generalized application of the active surveillance tools to a very broad range of medical product safety monitoring efforts for traditional post-market needs or for Medical Countermeasure efforts. The proposal centers on the multi-stakeholder collaboration among MDEpiNet methodology center in cooperation with three large and representative cardiovascular device registry owners: The American College of Cardiology National Cardiovascular Device Registry (NCDR), the Massachusetts Data Analysis Center (Mass-DAC), and the Department of Veterans Affairs. We propose to study the safety of 15 new cardiovascular devices, recently introduced, or anticipated to become available for clinical use over the next 5 years. These monitoring studies will employ prospective risk- adjusted cohort and sequential analytic methods, adjusted for impacts of device learning. In addition, the studies will evaluate the performance of advanced automated variable selection methods, as compared with domain-expert informed model development. Each of the three cardiovascular registries collaborating on this proposal have unique attributes regarding geographic representation, data quality, integration with electronic health record systems, and ability to link to longitudinal administrative and clinical health records. These differences in clinical data scope, scale and depth, will permit unique active device safety surveillance, through the establishment of pre-specified signal-verification analyses that will be triggered when a safety alert is identified in one dataset, through rapid assessment of the signal in the other registries. Finally, the complete safety surveillance system will be released as open-source software through an existing National Center for Biomedical Computing to foster collaboration in this field. The software will be available for free, to be used for any public health or academic application toward active medical product safety surveillance. The availability of these validated tools will permit data owners to monitor their clinical medical device datasets independently, or through programs coordinated through the FDA, the MDEpiNet program or other public and private stakeholders, to support the emerging national medical device safety surveillance network.
We will develop advanced methodologies to support dynamic, active safety surveillance of medical devices and share this operational monitoring system as an open-source toolkit for use with any medical device. The software will be tested using multiple large cardiovascular registries on 15 new medical devices.