SIR—In his letter , Dr. Diekema asks more questions than we are allowed space to answer, including why most US hospitals have not implemented surveillance culture (SC) programs and contact precautions (CPs) to control nosocomial vancomycin-resistant Enterococcus (VRE) and methicillin-resistant Staphylococcus aureus (MRSA) infections (NVMIs). Since 1983, the Centers for Disease Control and Prevention (CDC) has recommended CPs for patients colonized with epidemiologically important antibiotic-resistant pathogens but has not explicitly said that detection of colonized patients was necessary. Many US hospitals do the minimum required for infection control, assuming that this saves money. Northern European countries explicitly recommending SC/CP, however, have controlled endemic MRSA infection to <1% of nosocomial S. aureus infections for years to decades (evidence of sustainability) [2, 3], with very low estimated per capita infection costs  and rare adverse effects of (rarely needed) isolation. Diekema  also neglected to mention the positive results of all published cost-effectiveness analyses; for example, one analysis found that the costs of SC/CP were 19- to 27-fold lower than those of MRSA bacteremia in a comparable intensive care unit where endemicity was allowed to persist for 51 months (with 75 cases of bacteremia and 14 associated deaths) .
Diekema  says SCs and CPs are used in his hospital but will not be widely implemented until there is a “well-designed” randomized, controlled trial (RCT). We disagree with his emphasis on the results of a single study and his self-contradictory implication that there are insufficient data to begin control efforts. Recent meta-analyses of RCTs showed that individual RCTs provide results different from the mean results of all RCTs of the same question as often as did unrandomized epidemiologic studies of that same question [6, 7], which suggests that Hill  was right to emphasize the need for consistent results from multiple studies by different investigators in different populations before it is concluded that an association is causal. Consistent NVMI control with SC/CP has been reported in numerous studies , as have high strength of association, reversibility, and specificity, several more of Hill's causal criteria . We encourage interested readers to read the Society for Healthcare Epidemiology of America (SHEA) guideline on this topic , accessible at http://www.shea-online.org/PositionPapers.html.
RCTs examining the usefulness of active surveillance for VRE and MRSA likely have not been conducted because (1) they are expensive, (2) neither the National Institutes of Health nor the CDC wished to support RCTs of controlling NVMI during the last 3 decades of the 20th century, (3) many unrandomized studies have shown control of NVMI with SC/CP, and (4) some consider it unethical to randomize patients to suboptimal protection against potentially lethal infection . The occurrence of intra- and interhospital spread means that the optimal unit of randomization should not be individual patients, wards, or even hospitals. A recent study showed spread of 2 MRSA strains to all 12 study hospitals throughout 7 states, accounting for three-fourths of MRSA infections in the study hospitals . Without randomization of large clusters (i.e., each cluster large enough to render spread from contiguous nonparticipating wards/hospitals negligible), transmission from surrounding areas will bias such an RCT toward the null hypothesis. The proposed RCT mentioned by Diekema , however, randomizes an individual ward or two inside otherwise nonparticipating hospitals where NVMI spread is known to be frequent and where compliance with experimental SC/CP may be suboptimal. It also proposes comparison of SC/CP with hand hygiene more aggressive than most hospitals have been able to consistently achieve, potentially limiting durability and reproducibility.