AI Health Virtual Seminar: From Silos to Insights: a Data Strategy for Integrating Clinical Registries

Real world data, such as that from medical claims, the electronic health record, or clinical registries, are becoming increasingly important in quality improvement and research. Of these, clinical registries are unique in containing clinical outcomes data that are manually abstracted from the unstructured portions of the EHR by trained registrars or clinicians. This makes them a rich source of high-quality, clinical outcomes data with broad secondary use cases including comparative effectiveness research, cohort generation, and model development. However, use of clinical registry data is limited by idiosyncrasies in data storage, access, and governance - resulting in siloed datasets that are cut off from the broader health data landscape. Our work aims to unlock the potential of patient registries by creating a pilot data mart that links several of Duke's clinical registries with existing EHR data. In this session, I will give an overview of Duke's strategy for creating this data mart. I will highlight unique challenges we faced, and design considerations employed to address them.