Abstract
Background: The virtual data warehouse (VDW) was created as a mechanism for streamlining the process of proposing and conducting multi-site research by eliminating the need for custom data extraction programming at each site. At the core of the VDW are a series of standard file definitions. Content areas and data elements commonly required for research are identified, and data dictionaries created for each one. The data dictionaries specify a common format for each element —variable name, label, extended definitions, code values, and value labels. Local site programmers then developed programs to aggregate and transform their local legacy data into the standard VDW files. The data at the sites is thus made uniform, making it possible for multiple sites to run the same program against their local VDW data, and produce comparable results. The VDW is ‘virtual’ in the sense that the actual data remain at the local sites. It is not a centralized database held at a coordinating center and accessible only to a privileged few. Rather, it is a mechanism for sharing and reusing SAS code, and accelerating the research process.
Methods: This poster presents data from 3 HMOs that shows how the VDW is used at these sites.
Results: Programmers at Group Health, Kaiser Northwest, and Kaiser Colorado all use the VDW for more than multi-site projects. Using both automated file auditing and surveys of staff we present information demonstrating how the VDW is used at the 3 HMOs.
Conclusions: The original intent of the VDW was to increase efficiency in the use of data on multi-site studies. However, single site studies can also gain efficiencies from using the VDW and at some HMOs the VDW has become the first data source to use for almost all research. The VDW advantages of standard data structures and standard methods are also useful for single site projects.
- Received September 11, 2008.
