To the Editor: Staphylococcus aureus and Enterococcus faecium commonly cause healthcare-associated bloodstream infections (BSI) in the intensive care unit (ICU). Antimicrobial resistance is increasing in both organisms. The impact of antimicrobial resistance on dying of BSI has been studied extensively (1,2). Many studies have concluded that BSI caused by an antimicrobial-resistant organism results in higher death rates (1,3–8). However, as discussed in a recent report by Kaye et al., "outcome studies of antimicrobial drug resistance are notoriously hard to perform because of confounding variables related to coexisting conditions" (9). Indeed, almost all studies have shown that infections with antimicrobial-resistant organisms occur later in hospitalization than infections with antimicrobial-susceptible organisms, which suggests that differences in death rates may be, at least in part, caused by a difference in the patients' underlying illnesses and protracted hospital course. We report 2 additional methodologic issues that can affect estimates of the impact of antimicrobial resistance: combining different organisms and combining populations from different types of ICUs.

The original objective of our multicenter observational study was to quantify the clinical impact of antimicrobial resistance in S. aureus and E. faecium infections when these bacteria cause a specific type of infection: a monomicrobial, ICU-attributable, central vascular catheter–associated bloodstream infection (CVC-BSI). We studied 187 adult ICU patients with BSI caused by S. aureus and E. faecium at 3 tertiary care institutions from 1994 to 1999. The institutional review boards of each institution and the Centers for Disease Control and Prevention approved this study. Severity of illness was measured with an APACHE II score at ICU admission and on day 7 in the ICU (if applicable). The score would indicate the patient's risk of dying in the hospital before a BSI developed by using a measure validated for predicting in-hospital deaths in ICU patients (10).

The study population stratified by organism is shown in the Table. Fifty-eight percent of patients had CVC-BSI with S. aureus, and 42% had CVC-BSI with E. faecium. Overall, 58% of the organisms causing CVC-BSI were resistant to oxacillin if S. aureus or to vancomycin if E. faecium. However, patients with E. faecium CVC-BSI were more likely to be infected with antimicrobial-resistant bacteria (69% versus 50%, p<0.01), and had a higher mortality rate (54% versus 34%, p<0.01) than patients with S. aureus CVC-BSI. This finding indicates that the type of organism (E. faecium versus S. aureus) confounds the association between resistance and death. In addition, the distribution of ICU type by organism varies, which suggests that patient populations infected with these 2 different organisms were different in other ways. Thus, confounding factors for the association between resistance and death may differ for E. faecium and S. aureus, and analysis of the 2 organisms should be conducted separately. This is consistent with the results of Kaye et al. who showed that the effect of resistance was higher for S. aureus (odds ratio [OR] 3.4) than for E. faecium (OR 2.5) by using separate analyses to show death rates (9). Furthermore, these researchers found different confounding factors in the adjusted analysis of S. aureus than in the adjusted analysis of E. faecium. Because of the need to conduct separate analyses, which reduced our statistical power, our study was ultimately unable to show a difference in death rates if it existed.

In summary, future studies measuring the impact of antimicrobial resistance on death rates should be restricted to a specific type of infection cause by a single organism in a uniform setting using a validated system to predict mortality in that setting. As such, future studies should involve multiple study sites.