Use of different thresholds of prior antimicrobial use in defining exposure: Impact on the association between antimicrobial use and antimicrobial resistance

Objectives: Although many studies have explored the association between antimicrobial use and antimicrobial resistance, definitions of exposure (i.e., prior antimicrobial use) differ across such studies. Specifically, it has been noted that some studies define exposure as any antimicrobial use, while others require the administration of at least 24 or 48 h of an antimicrobial to constitute exposure. The impact of different definitions of exposure on final study results is unknown. We conducted the current study to determine the impact of varying the minimum threshold of prior antimicrobial use to define exposure status in studies of antimicrobial resistance. Methods: We used a dataset from a prior study of risk factors for fluoroquinolone (FQ) resistant Pseudomonas aeruginosa (FQRPA) to address the study aim. Four separate multivariable models of risk factors for FQRPA were built. Each model defined a different threshold for the duration of antimicrobial administration to determine exposure to that antimicrobial: (1) no threshold (i.e., a subject is considered exposed if any use of the antimicrobial was documented); (2) > 24 h of use of the antimicrobial is necessary to be considered exposed; (3) > 48 h of use is required; and (4) > 72 h of use is required. Except for these definitions, the four multivariable models were built in exactly the same way, each using prior FQ use as the primary risk factor of interest. Results: Among 872 P. aeruginosa isolates included in the original dataset, 332 (38.2%) were FQ-resistant. Each of the four multivariable models identified prior FQ use as an independent risk factor for FQRPA. However, as increasingly strict thresholds were used, the association between FQ use and FQRPA increased. Furthermore, prior use of an agent with activity against anaerobic bacteria was associated with FQRPA in models 2, 3 and 4, but not in model 1. Conclusions: The use of different thresholds to define prior exposure to antimicrobials altered the associations between antimicrobial use and resistance. Stricter thresholds resulted in a higher estimate of the association between antimicrobial use and resistance. Furthermore, different risk factors may be identified depending on how exposure is defined. Greater attention to this issue is necessary to optimize identification of modifiable risk factors for resistance as well as accurately compare results across studies. (c) 2007 Published by Elsevier Ltd on behalf of The British Infection Society.