This paper presents a decision support system for nosocomial infections and its integration in the large HIS of the University Hospital of Giessen. The model comprises five different engines and a data dictionary. It is designed to detect hospital aquired infections even in a situation where only a restricted amount of clinical data is available (the data is split up in different information systems). Furthermore the model prevents time consuming manual data entry.

The five engines split the main task into 1) a preselection, which sort out patients who definitely do not have a nosocomial infection; 2) a rule based reasoning process which detects patients likely to have such an infection; 3) an alarm process which is responsible for the presentation of the alert; 4) an explanation process to follow up the reasoning and 5) statistic tools to answer specific hygienic questions.

A data dictionary supplies the controlled vocabulary, but it is also required to understand datastructures used in the different clinical subsystems.