Many endocrinologic diagnoses rest on results of functional tests. Such tests yield large and complex data sets, whose normal values depend on gender, age, developmental stage, and the laboratory in which the tests are performed. The ratios of hormonal values are also computed and compared to accepted normal ratios, which may also vary among institutions. In addition to the intrinsic complexity of this analysis task, automated interpretation of these tests is made substantially more difficult because of the heterogeneity of the databases in which these tests are stored. Functional adrenal hormone testing serves as an archetypal illustration of these points. To facilitate the retrieval and analysis of these data from an institution's electronic medical record system (EMRS), we have developed a computer-based decision-support system called ATRAS (Adrenal Test Retrieval and Analysis System).

ATRAS is written in the Java computer language, and will be integrated into the W3-EMRS. The W3-EMRS is an Internet-based multi-institutional architecture, allowing for uniform user access to heterogeneous medical databases. This will enable ATRAS's use in virtually any medical environment, regardless of EMRS.

A particular challenge in ATRAS' development has been the determination of which normal values to apply to a particular patient. For example, should age or Tanner stage take precedence in specifying normal values? Which values are used for patients whose age falls outside the specified ranges? User-selectable switches and analysis heuristics were developed to allow for these factors.

ATRAS automatically retrieves all relevant adrenal lab tests performed on a patient from the EMRS, and calculates relevant hormonal ratios. It displays these data, including normal values, in tabular format within a web browser(e.g. Netscape Navigator). Abnormal values are highlighted in red. Users may select conventional or SI units for display. ATRAS retrieves, calculates, and displays all relevant adrenal data in < 2 seconds; most endocrinologists take several minutes to perform this task. A randomized prospective trial assessing performance of ATRAS versus conventional methods is planned.