Data Extraction and Ad Hoc Query of an Entity– Attribute–Value Database

n J Am Med Inform Assoc. 1998;5:511–527. A clinical patient record needs to store several thousand possible facts for patients across all specialties. For any given patient, however, the number of actual findings may not exceed a few dozen. A conventional (‘‘orthodox’’) database table design (one fact per column) is unsuitable for such data, because of database vendor limitations on the number of columns per table and the need to continually add new tables or columns whenever new facts need incorporation. Most mainstream electronic patient record systems (EPRSs) deal with this problem through the entity–attribute– value (EAV) representation. In this methodology (also referred to as row modeling), the fact descriptors (attributes) are treated as data, so that the addition of new facts does not make database restructuring necAffiliation of the authors: Yale University, New Haven, Con-