The Importance of Continuing Local Bacterial Surveillance

To the Editor—We appreciate the insightful comments by Vandijck et al regarding our article on the cost of methicillinresistant Staphylococcus aureus (MRSA) bloodstream infections. We included variables in the propensity score that are clinically important even if they were not statistically significant, as recommended in statistics textbooks. We decided to use only hospital data because prehospitalization data may be less reliable. However, data on residence in a nursing home is an exception, and we agree with Vandijck et al that it should have been included. When we repeated the calculations with the inclusion of nursing home residence as a covariate in the propensity score regression, the separation between patients with methicillinsusceptible S. aureus (MSSA) BSI and patients with MRSA BSI increased. According to the original calculations, there were 41 patients with a propensity score greater than 0.8; 2 of these patients developed a MSSA BSI, and 39 developed a MRSA BSI. When we applied the new regression, there were 47 patients with a propensity score greater than 0.8; 2 of these patients developed a MSSA BSI, and 45 developed a MRSA BSI. The effect of MRSA BSI on length of stay, cost after infection, and mortality became even less significant for the group of patients who had a propensity score greater than 0.1 and less than 0.8, while the effect of MRSA BSI on intensive care unit patients remained significant over the entire range of propensity scores. Previous antimicrobial therapy may also have an effect on clinical outcome and on hospital cost; however, these data were not available.