Case-control studies reporting on risk factors for emergence of antimicrobial resistance: bias associated with the selection of the control group

The optimal control group for case-control studies examining antibiotics as risk factors for the emergence of antimicrobial resistance is patients selected randomly from the total hospital population, while the selection of patients with a susceptible bacterium is deemed suboptimal. We sought to theoretically elaborate on potential parameters that introduce bias associated with the use of randomly selected control subjects, based on personal experience and data from the literature. In addition, we considered parameters that introduce potential bias associated with the definition of case patients. Parameters that may introduce potential bias associated with the randomly selected control subjects are use of antibiotics in the community (background exposure), availability of an antibiotic in a country, ability to purchase specific antibiotics or health care, the bacterial resistance pattern in the country, in vitro evaluation issues, source of admitting patients (nursing home or community), type of hospital to which patients are admitted (general or disease specific), and ward of hospital to which patients are admitted. Parameters that may introduce potential bias associated with the case definition are multidrug resistance versus resistance to only one antibiotic, resistance phenotype of the microbe, multistep versus one-step development of resistance, appropriateness versus adequacy of antibiotic treatment, antibiotic synergy, details regarding the daily dose and duration of administration of the specific antibiotic, and use of other antibiotics. In conclusion, selection of control subjects from the hospital population is also associated with bias. The most acceptable solutions to evaluate the risk factors for antimicrobial resistance are probably the case-control-control study design and the case-case-control study design.