Potential confounding in evaluating infection-control interventions in hospital settings: Changing antibiotic prescription

The colonization dynamics of antibiotic-resistant pathogens in hospital settings are complex, with multiple and continuously interacting variables (e.g., introduction of resistance, infection-control practices, antibiotic use). Quantification of these variables is indispensable in the evaluation of intervention studies, because these variables represent potential confounders. In this article, the complexity of colonization dynamics is described. Through a systematic review, we identified studies that evaluated the modification of antibiotic prescription to reduce antibiotic resistance in intensive care units (n = 19), and the extent of confounding-control was determined. Most studies evaluated antimicrobial restriction/substitution (n = 12) or antibiotic rotation (n = 4). Sixteen studies had a prospective cohort design (before-after), of which 12 were without a control group. Introduction of antibiotic resistance was determined in 10 studies. The relative importance of colonization routes and adherence to infection-control measures were not determined in any study. Therefore, it remains uncertain whether observed changes in the prevalence of antibiotic resistance after intervention were causally related to the intervention. Appropriate choices of study design, primary end point (colonization rates rather than infection rates) and statistical tests, determination of colonization routes, and control of potential confounders are needed to increase validity of intervention studies. © 2006 by the Infectious Diseases Society of America. All rights reserved.