Data and databases in cardiovascular medicine and surgery
Vast amounts of complex information are created in the processes of healthcare. This chapter delineates three categories of computational readiness of that information: native data that is analytics (and AI) ready, discrete data that must be translated to be computationally ready, and unstructured documentation. Unfortunately, 80%–90% of healthcare information is unstructured, locked in the vagaries and imprecision of verbose text. To understand the journey to good data, a high-level data framework is reviewed, specifically the concepts of common data elements, analysis-ready data structures, and common data models. Initiatives such as the federal USCDI and the HL7 Fast Healthcare Interoperability Resources (FHIR) application programming interfaces are acknowledged as accelerators of data liquidity that will enable AI. Finally, the need for transformation of healthcare itself is recognized—specifically, a call for the reengineering of healthcare where the capture of good data occurs at the point of care via a computationally intensive environment that actually reduces clinician burden.
