The GAMES European Community project deals with the general architecture of medical expert systems. Within the framework of this project, a methodology has been proposed, based on an epistemological model of reasoning of the hypothetico-deductive type. We applied this methodological approach to the design and the building of a system for diagnosing and predicting hospital-acquired infections, in the environment of DIOGENE, the Hospital Information System (HIS) of Geneva University.
Diagnostic rules were applied to DIOGENE databases, allowing the establishment of a daily list of selected patients with suspected infection: the process was applied to the data of 15 laboratories which are transferred everyday from the dedicated laboratory system to Archimed, an integrated archiving system.
Our goal was to test a sensitive diagnostic process, in a particularly well-suited HIS architecture. We tested this method for the diagnosis of pneumonia in the surgical intensive care unit: a list of patients with suspected infection was established daily on the basis of this process. Validation of the system was performed by applying the standard rules published by the Centers for Disease Control (CDC) of Atlanta for the diagnosis of nosocomial infections to all the patients of the care unit.
This highly sensitive system has a practical interest for the infection control staff for their reporting of nosocomial infections. In a high-risk unit, where all the patients' charts should be checked, the system allows to control only a small proportion of the patients, approximately one fifth with suspected infection, and therefore to save work time.
This strategy is made possible with the Geneva University HIS, firstly by a new open distributed architecture, and secondly, by the development of Archimed, a system for archiving distributed databases.
A computerized system for surveillance of nosocomial infections will play a major role for the quality of patient care and reduction of hospital costs, and furthermore, represents a commercial potential within the field of hospital information systems.