ABSTRACT
Laboratory tests can appear similar from the test names but may be vastly different in the way a result is achieved. Currently, for example, cervical cancer evaluation is moving from the traditional Papanicolaou smear to new smear preparation technologies and testing for human papillomavirus. Monitoring the quality of these three tests, and of all tests, requires that computers "understand" how these tests are similar and different. The National Inventory of Clinical Laboratory Testing Services (NICLTS) found that the approximately 20,000 most commonly performed tests used combinations of 635 analytes and 1,699 methods. These analytes and methods provide the base data for a semantic model that makes the requisite similarities and differences explicit. The semantic relationships, e.g. the method principle enabling a test and the nature of the substance tested, were evaluated against empirically derived, uni-dimensional relations. The resulting multi-dimensional semantic model expands our ability to monitor the quality of laboratory testing in the face of rapid change. Use of common terminology tools and representations enable the creation, expansion and reuse of this model beyond the needs of NICLTS.
