Typen begrifflicher ordnungssysteme in der medizin

In medicine, systematic documentation of data and knowledge is a prerequisite for nearly all aspects of patient care and clinical research. For various purposes data and knowledge have to be represented in order to enable interpretation, analysis, aggregation, and presentation. As medical concepts are quantitatively and qualitatively complex, and as there are problems involved in the utilization of both natural and medical language (e.g., synonymy, ambiguity), the importance of standardization of medical terminology has become increasingly important in this field, especially for computer-supported processing of medical data like information retrieval, data-based decision support, communication of data, and statistical evaluations. This paper deals with classifications of diagnoses as basic entities in medicine. Essentially, they make use of hierarchical concept relations such as "Diabetic Retinopathy is an Eye Disease". The occurrence of this relation can be deduced from left to right on the basis of appropriate conceptual knowledge (e.g., retina is part of eye). It can, however, also be established by inference from right to left on the basis of the most relevant and frequent eye diseases statistically evaluated within a disjunct and exhaustive case classification. This leads to two types of conceptual ordering: On the one hand, a classification of concept sets based on intensions which permit the formal definition of concepts and the inference of conceptual knowledge on the basis of appropriate logical calculi; on the other hand, a statistically evaluated classification of object sets based on a disjunct and exhaustive subdivision of cases on several hierarchical levels. Since it is possible to aggregate intensionally represented concepts or at least to map them to statistically evaluated object classifications, an interesting question arises, which is being discussed controversially in medical informatics: Can statistically evaluated classifications, which are characterized by a number of disadvantages, be replaced without loss in efficiency by formal concept systems?