Abstract 13128: Development and Initial Utilization of a Multi-institutional Distributed Data Network Designed to Evaluate Real-world Clinical Outcomes in Patients Presenting for Percutaneous Coronary Intervention (PCI): Results From the Building the Unique Device Identifier (UDI) Into Longitudinal Data for Medical Device Evaluation (BUILD) Network

Introduction: Generating real-world evidence for regulatory and research purposes is a fundamental transformation of healthcare. To develop and demonstrate the operational components, we created the BUILD network, a framework for the longitudinal follow-up of all patients undergoing PCI at three large health systems, Mercy, Geisinger, and Intermountain. It leverages consistent data through a standardized common data model using distributed analyses without site-specific programming or external transfer of clinical data. To demonstrate BUILD’s potential, we evaluated clinical outcomes between two commonly used drug-eluting coronary stents. Methods: We included all patients receiving zotarolimus, or everolimus stents from 2012-19. Data were compiled from registries and electronic health/billing records systems without explicit chart abstraction, and retained at each site in the BUILD analysis database. A common query examined 32 baseline covariates, created propensity-matched cohorts, assessed for balance, and compared 1-year major adverse cardiac event (MACE) rates between groups, with weighted data combined to create event-free survival plots. Results: We identified 11,030 patients (Mercy=4,906, Geisinger=4,109, Intermountain=127; mean age 65±12, 69% males, 23% zotarolimus). Figure 1 shows hazard ratios and survival curves for the composite endpoints in matched cohorts. Differences in subsequent PCI, mortality, and composite MACE were not significantly different between stent types. Conclusions: This large real-world, multicenter demonstration project showed no statistical differences in 1-year clinical outcomes of patients receiving zotarolimus versus everolimus stents—results similar to other studies. More importantly, it demonstrated that multi-institutional data networks can be successfully developed to provide clinically relevant evidence while maintaining the source data locally on-premise at each institution.