Answers are grounded in this paper and cite their sources.
Sir—Cosgrove et al. [1] estimated the impact of methicillin resistance on mortality associated with Staphylococcus aureus bacteremia. Using a meta-analytic method to aggregate results across 31 studies, they demonstrated a significant difference between the mortality rate among patients with methicillin-resistant Staphylococcus aureus (MRSA) infection and that among patients with methicillin-susceptible S. aureus (MSSA) infection. However, the type of bacteria responsible for infection is not the only characteristic which differentiates the groups. The length of hospital stay (LOS) before the onset of bacteremia is significantly longer for the patients with MRSA infection, according to 14 publications in which this data is available (table 1). The median LOS before bacteremia for MSSA groups is 9.2 days, and the median LOS for MRSA groups is 22.2 days (P < .001; Wilcoxon rank sum test). Length of hospital stay before the onset of bacteremia for patients with methicillin-susceptible or methicillin-resistant Staphylococcus aureus bacteremia, as reported in 13 studies. Moreover, the possible confounding of the patient's underlying disease and treatment deserves further analysis. Cosgrove et al. [1] found a similar OR for the subgroup of 11 studies that adjusted for confounding factors. It should be noted that, among these studies, the OR after the adjustment increased in 2, decreased in 2, and remained the same in the others. Although some investigators [2] have used APACHE II scores to adjust for the severity of illness, this is an unproven method of adjustment for patients who have been hospitalized for different lengths of time. What, then, is the best method to adjust for possible confounding factors? With these questions in mind, I offer an ecological analysis of the data that uses a regression analysis. I have used LOS before the onset of bacteremia if this information has been stated in the original articles, and I have also included information from a more recently published study [3]. LOS information was not available from 18 of the other studies, including the 4 studies of patients with endocarditis. Figure 1 shows the group mortality rates compared with the median LOS before bacteremia for each group, as reported in the 14 studies in which this data is available. The relationship between length of hospital stay before onset of bacteremia and mortality rates for groups of patients with either methicillin-resistant Staphylococcus aureus (closed symbols) or methicillin-susceptible S. aureus (open symbols) bacteremia, as reported in 14 studies for which data on length of stay was available. The size of each group is represented by the corresponding size of the symbols (⩽30, 31–50, 51–109, 110–299, and ⩾300 patients, respectively, from smallest to largest; modified from Hurley [3] with permission). An unweighted regression analysis reveals a significant association between the LOS and group mortality rate (P < .005). However, the addition of group MRSA or MSSA status is not significant in this regression, and LOS remains a significant predictor of group mortality risk. This suggests that, with S. aureus bacteremia of either type, the mortality rate among members of a group increases with LOS. Patients hospitalized for longer periods of time not only have more-severe illness than others, but they are also more likely to have been exposed to multiple courses of antibiotics and are more likely to acquire an S . aureus infection that is drug-resistant. The mortality risk is no different among patients with MRSA bacteremia than it is among patients with MSSA bacteremia if LOS, a surrogate marker for severity of patient illness, is taken into account. The analysis that I present is ecological; that is, it is not based on patient-specific data and, hence, should not be regarded as conclusive evidence of a cause-and-effect relationship. However, it does suggest a systematic bias between the 2 groups, a bias that a meta-anlysis of observational data would not be able to control for. The difference that Cosgrove et al. [1] describe could be confounded by the difference between the LOS before bacteremia of groups of patients with MRSA bacteremia and that of groups of patients with MSSA bacteremia.
No takes yet. Share an insight, caveat, or question.
James C. Hurley (2003) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: