Bridging the Representational Heterogeneity of Clinical Databases

Artificial Intelligence in Medicine: Interpreting Clinical Data

Volume

Papers from the 1994 AAAI Spring Symposium

Contents

Clinical decision-support ( CDS) applications are difficult share among health care institutions because such applications require patient-specific data, and the electronic representation of clinical data varies widely across institutions. Our technology facilitates the sharing of CDS applications by provMing a standard abstract database schema of clinical information and a means to automatically translate queries that are formulated with respect to the standard schema to equivalent queries that are valid with respect to arbitrary relational representations. We briefly describe our technology, discuss the inadequacies of extant methods that address the same problem, and present a set of formal evaluation criteria by which our technology can be shown superior to these methods. The evaluation criteria are applied in a formal controlled experiment we have designed based on the clinical ICU data set.