Background: In preparing hospital antibiograms for individual organisms and antibiotics, laboratories often combine susceptibility data for isolates from a variety of sources and patient types. If results from patients with known resistance patterns that vary from normal are included, the overall susceptibility value for the institution could be misleadingly skewed.
Objective: To assess the degree of bias introduced into a hospital antibiogram by combining cystic fibrosis (CF) and non-CF isolates of Pseudomonas aeruginosa to produce one hospital-wide percent susceptible figure for each tested antibiotic.
Methods: A retrospective analysis was conducted of an academic, tertiary care medical center's microbiology database. We examined quarterly and annual susceptibility data from 2004, comparing non-CF data with combined susceptibility data for 10 antibiotics within each quarter, as well as those reported in the annual antibiogram. Differences were assessed for statistical significance using chi(2) testing with Bonferroni correction.
Results: Large differences were observed between non-CF and combined percent susceptible data in the 4 quarters (aminoglycosides 3% vs 20%, fluoroquinolones 2% vs 18%, respectively) and when comparing annual non-CF (n = 191) with annual combined (n = 266) data. With the annual figures, these differences were frequently statistically significant (70% vs 58%, 91% vs 83%, 85% vs 70%, and 72% vs 60% for gentamicin, tobramycin, amikacin, and gatifloxacin/levofloxacin, respectively; all p< or =0.01).
Conclusions: Combining CF and non-CF P. aeruginosa susceptibility into one percent susceptibility value for all isolates may produce figures that underestimate the activity of some antibiotic classes against non-CF isolates. Clinicians may make less than optimal empiric antibiotic selection choices based on such data.