Evaluation of Data Correction Methods for Positive Blood Culture Results

Background The outcome of patients with bacteremia is influenced by the initial selection of adequate antimicrobial therapy. However, at the early stage of a bloodstream infection, microbiological identification and susceptibility of the causative organism are not available. Therefore, a calculated (empirical) antimicrobial therapy has to be started, based on the knowledge of the most likely microorganism and its susceptibility pattern. In hospitals, analysis of blood culture results is widely used. However, because of the difficulty of distinguishing true episodes of infection from specimen contamination and the generation of repeated results from one patient in the course of treatment, crude data has to be corrected for common skin contaminants and duplicate results before further analysis is performed. In this context, the objective of our study was to evaluate the influence of different data correction methods on ranking of pathogens and the cumulative antibiotic susceptibility pattern of blood culture isolates from surgical intensive care units (ICUs) to obtain a data correction preprocessing method for implementation in a knowledge-based surveillance system for bloodstream infections.