Scaling an expert system data mart: more facilities in real-time

Clinical Data Repositories are being rapidly adopted by large healthcare organizations as a method of centralizing and unifying clinical data currently stored in diverse and isolated information systems. Once stored in a clinical data repository, healthcare organizations seek to use this centralized data to store, analyze, interpret, and influence clinical care, quality and outcomes. A recent trend in the repository field has been the adoption of data marts--specialized subsets of enterprise-wide data taken from a larger repository designed specifically to answer highly focused questions. A data mart exploits the data stored in the repository, but can use unique structures or summary statistics generated specifically for an area of study. Thus, data marts benefit from the existence of a repository, are less general than a repository, but provide more effective and efficient support for an enterprise-wide data analysis task. In previous work, we described the use of batch processing for populating data marts directly from legacy systems. In this paper, we describe an architecture that uses both primary data sources and an evolving enterprise-wide clinical data repository to create real-time data sources for a clinical data mart to support highly specialized clinical expert systems.