Knowledge discovery from microbiology data: Many-sided analysis of antibiotic resistance in nosocomial infections

Abstract

Nosocomial infections and antimicrobial resistance (AR) are highly important problems that impact the morbidity and mortality of hospitalized patients as well as their cost of care. The goal of this paper is to demonstrate our analysis of AR by applying a number of various data mining (DM) techniques to real hospital data. The data for the analysis includes instances of sensitivity of nosocomial infections to antibiotics collected in a hospital over three years 2002-2004. The results of our study show that DM makes it easy for experts to inspect patterns that might otherwise be missed by usual (manual) infection control. However, the clinical relevance and utility of these findings await the results of prospective studies. We see our main contribution in this paper in introducing and applying a many-sided analysis approach to real-world data. The application of diversified DM techniques, which are not necessarily accurate and do not best suit to the present problem in the usual sense, still offers a possibility to analyze and understand the problem from different perspectives. © Springer-Verlag Berlin Heidelberg 2005.