Abstract:
Expert systems in medicine are frequently restricted to assisting the physician to
derive a patient-specific diagnosis and therapy proposal. In many cases, however,
there is a clinical need to use these patient data for other purposes as well. The
intention of this paper is to show how and to what extent patient data in expert systems
can additionally be used to create clinical registries and for statistical data analysis.
At first, the pitfalls of goal-oriented mechanisms for the multiple usability of data
are shown by means of an example. Then a data acquisition and inference mechanism
is proposed, which includes a procedure for controlling selection bias, the so-called
knowledge-based attribute selection. The functional view and the architectural view
of expert systems suitable for the multiple usability of patient data is outlined
in general and then by means of an application example. Finally, the ideas presented
are discussed and compared with related approaches.
Key-Words:
Decision Support - Diagnosis - Therapy - Expert Systems - Clinical Registries - Attribute
Selection