The letter by Dr. Diekema raises several items for discussion. We regret that he found our title misleading. As indicated in our article, the program's critical and enforced aspect was universal admission surveillance. Decolonization was at the attending physician's discretion. We routinely culture all positive polymerase chain reaction samples for MRSA and test isolates for mupirocin resistance, which occurred in 5.6% of strains before universal surveillance began and 4.1% at the end of the program's first year (1). In addition, to minimize the potential for resistance, we removed mupirocin for clinical use other than for nasal decolonization.

The overall impact of isolation on patient care is an important question. Some reports, none of which are randomized trials, have suggested a possible negative impact on the isolated patient. The cited review by Diekema and colleagues (2) references the report by Stelfox and colleagues, which seems to demonstrate more adverse events in isolated patients as well as fewer acceptable process-of-care measures (3). However, there were unacceptably high lapses in care of patients in the control (unisolated) group as well, with 21% to 41% having days with no vital signs recorded as ordered, 9% to 11% having days with no nursing notes, and 2% to 24% having days with no physician notes. This suggests that considerable improvement in many care practices were needed at the 2 study sites. Of note, all statistically significant increases in adverse events were related to supportive care failures, which is consistent with the unacceptably low process-of-care documentation that could be readily monitored and targeted for an improvement intervention to lower supportive care failures.

As expected, patients are placed into isolation precautions for reasons other than MRSA infection (for example, undiagnosed pneumonia, other multidrug-resistant organisms, or Clostridium difficile–associated diarrhea). We found that universal MRSA surveillance only modestly increased the overall rate of patient isolation (4). A recent 1-day prevalence survey at Evanston Hospital, Evanston, Illinois (29 April 2008), found that 10 of 30 patients were in isolation because of the universal MRSA surveillance program. As a monitor of satisfaction or dissatisfaction with our program, after the study period, we asked our service excellence department to review any patient complaints related to our MRSA control program; there were none. On the contrary, there were spontaneous comments thanking Evanston Northwestern Healthcare for our aggressive approach to preventing health care–associated infections.

Another critical issue raised was that of focus on MRSA alone. We fully concur. Although it was not part of our report, we managed an outbreak of Acinetobacter baumanii in 1 of our intensive care units and an increase in Clostridium difficile–associated diarrhea across our 3-hospital organization during the time frame of our MRSA program. We also developed a new program to minimize our surgical site infections from S. aureus. All of these required separate investigation and intervention initiatives that were successfully completed (5–7), illustrating that MRSA need not be the only focus of an infection control program. This is not a zero-sum game, in which control of an infection necessarily leads to fewer funds for the control of another. In our case, success with MRSA and the cost savings that resulted have, in fact, led to a greater willingness of our hospital administration to support other infection control efforts.

We do believe, as our data suggest, that preventing transmission of MRSA infection so that fewer patients become colonized with this pathogen is a worthwhile goal for infection control programs. Another example of this is the recent publication by Harbarth and colleagues (8). This group demonstrated that targeted surveillance for MRSA infection in their surgical patients showed no benefit in lowering overall surgical site infections. However, they documented the important risk associated with MRSA colonization by showing that 5% to 13% of MRSA-colonized patients developed infection compared with 0% to 0.5% of noncarriers.

Dr. Nelson raises the issue of independent review board approval and asks questions about our statistical methods. Independent review board approval was obtained for this project and disclosed to Annals; unfortunately, mention of it was unintentionally omitted in the final publication. Although decolonization was recommended by infection control, it was not enforced and was at the discretion of the attending physician, requiring a specific order for this treatment. No adverse event related to decolonization was reported during the 21-month intervention.

Regarding the statistical methods, 37 MRSA and 37 MSSA bacteremias occurred in the baseline period, accounting for the identical rates and 95% CIs. In view of the comments regarding our data on MRSA and MSSA infection rates, we were perhaps insufficiently clear in outlining the aim of this analysis. We did not set out to directly compare the change in MRSA bacteremia rates with that in MSSA bacteremia rates; because of the small number of bacteremic infections, statistical power for detecting such differences was limited. However, we have reanalyzed our data in this way, and even with the limited available power, a test comparing the period-3 MRSA and MSSA bacteremia rates on the basis of a Poisson regression results in a P value of 0.085. That this difference should approach statistical significance has "face validity" because we targeted MRSA and not MSSA, thus the expectation is that only MRSA bacteremia would decline if no other programs effective against hospital-acquired bacteremia were in place. Although MRSA and MSSA bacteremia rates were identical at baseline, MSSA rates were 50% higher than MRSA rates during period 3. We believe Dr. Nelson arrived at a different result because the "back-of-the-envelope" estimate did not account for the differing lengths of time in periods 1 and 3.

Dr. Nelson's next set of calculations (in which a P value of 0.08 is presented for the change in MSSA infections) erroneously treats the 37 infections at baseline as a fixed value, effectively ignoring the random error. He also uses 1-sample tests to compare the universal surveillance results with this value. Our calculations used 2-group comparisons, accounting for random variation in measurement for both periods, and the result was larger P values.