Clinical Correlates of Enterococcal Bacteremia: Attributable or Associated?

Sir—In their analysis of clinical outcomes that correlate with enterococcal bacteremia, Caballero-Granado et al. [1] correctly state that, in observational studies, we must account for potential confounding variables to protect against making false inferences of causation. This principle was observed in their analysis of mortality, which incorporated multivariate logistic regression analysis. Unfortunately, this principle was not observed in their analysis of duration of hospital stay, which provided the only (putatively) significant association between outcome and enterococcal bacteremia [1]. Adjustment or stratification for potential confounders is essential for this analysis as well, because, compared with the control group, the patients with bacteremia exhibited a significantly higher prevalence of several baseline characteristics (i.e., cirrhosis, neutropenia, history of organ transplantation, prolonged prior hospital stay, urinary catheter use, nasogastric tube use, receipt of parenteral nutrition, and prior antibiotic therapy) [1]. Although these characteristics possibly predispose patients to develop enterococcal bacteremia, they also would be expected to be associated with prolonged hospital stays, independent of bacteremia. Furthermore, even if analysis were to show an independent association between enterococcal bacteremia and duration of stay, the possibility that unmeasured confounding variables were responsible for the association could not be excluded (as inevitably is the case in observational studies). Thus, the title of the article (“Attributable...Duration of Hospital Stay...” [1], p. 587) is potentially misleading, and the concluding sentence of the abstract (“Enterococcal bacteremia...extends the duration of hospital stay of patients who develop it” [1], p. 587) is inadequately supported by the data presented. Perhaps the authors could provide a multivariate analysis that includes the other clinical variables (i.e., in addition to enterococcal bacteremia) as predictors of duration of hospital stay, to convince skeptical readers that bacteremia remains a significant predictor after adjustment for other known host factors.