Introduction The Medical Entities Dictionary is a large repository of medical concepts that are drawn from a variety of sources either developed or used at  the New York Presbyterian Hospital, including the UMLS, ICD9-CM and LOINC. Currently numbering over 100,000, these concepts correspond to coded terms used in systems and  applications throughout both medical centers (Columbia-Presbyterian and New York-Cornell). It continues to grow at about 6,000 terms per year, although accelerated growth is anticipated as additional network hospitals are integrated into the NYPH system.   Structure   The terms are brought together in the MED and represented as frames, arranged in a semantic network. Each frame includes information specific to the term, such as its name, its code or codes in various systems, and related textual information (e.g., units of measure for tests, synonyms, etc.). The frames also contain pointers to related terms in the MED's semantic net. Some of these pointers form the multiple hierarchy of the MED, while others provide name-attribute information (for example, the concept "Serum Sodium Test" is linked to the concepts "Serum Specimen" and "Sodium Ion" through the relations "has-specimen" and "substance-measured", respectively).   Integration   The MED has proven to be a powerful tool in two respects. The first is the ability to support multiple applications across the institution. Programs which, for example, display laboratory results do not need to stay synchronized with the various laboratory systems with respect to  terminology. Instead, they can obtain test names and units from the the MED for any test term encountered in the clinical repository.   Construction   The second powerful aspect of the MED is to support the knowledge-based editing of itself so that, for example, new test terms can be placed in the appropriate classes for later aggregation (it does this through the use of the semantic links "has-specimen" and "substance-measured", which map individual tests to test classes). A number of knowledge-based tools have been created to support knowledge-based term classification and other maintenance functions, including programs which check for validity and internal consistency.