Evidence-based laboratory interpretation system built on a large collection of case records with well-defined...

A computer-based system for laboratory diagnosis was created to promote evidence-based practice of laboratory medicine. It runs on a database consisting of 1992 well-defined untreated cases from 38 common diagnostic categories. Clinical symptoms, signs, and severity were recorded together with laboratory test results both general and specific to the clinical diagnosis. The system has two modes in which to view the database: either single or multiple diagnostic categories at a time. In the single mode, it allows flexible filtering and quick subgrouping of cases within the diagnostic category by specifying parameters of interest. It is also capable of computing a similarity index of a case at hand to those in the database. The index is defined as a weighted sum of log-likelihood calculated dynamically from parameters chosen for the query. In the multi-mode, only the parameters commonly recorded in selected categories are retrieved from the database. The system offers a between-category comparison view of any parameter. The similarity indices can be also computed among categories to see how well a set of parameters differentiates a reference category from the others.