[Nosocomial infection monitoring system featuring detection of local clustering].

We have developed a nosocomial infection surveillance system making use of data from laboratory information system. The system makes cross-reference table of detected bacteria according to either the site of occurrence(hospital wards) or antibiotic sensitivity. It is capable of listing all the patients or serial changes in frequency for any specified bacterium. Furthermore, we have developed an algorism to detect local clustering. For each ward, the system calculates all combinations of distance between beds of patients with specified bacteria. We named the statistics as DC(degree of cluster) and its significance was judged by a confidence interval of DC obtained by a bootstrap method by randomly assigning the same number of patients to the beds in the same wards. Retrospective analysis of the distribution of 4 major bacteria in the wards proved that DC is a sensitive indicator of local clustering.