JSTOR

Methods

We conducted a prospective cohort study involving patients admitted to the Iowa City VAMC during the first year of a 4‐year, quasi‐experimental study of the impact of rapid MRSA and VRE detection. From January 1, 2006, through December 31, 2006, all patients who were admitted to acute care units at the Iowa City VAMC during the first week of each month had nares swab samples collected for MRSA testing and perirectal swab samples collected for VRE testing. Screening was performed only at hospital admission, which was defined as within the first 24 hours after hospitalization.

Perirectal swab samples were plated on bile‐esculin agar supplemented with vancomycin (6 μg/mL). Enterococci were identified using conventional methods and by Vitek GPI cards (bioMérieux). Nasal swab samples were plated on blood agar, and conventional methods were used to identify S. aureus. Bacterial identification and drug susceptibility profiles were confirmed using the Vitek automated platform (bioMérieux). In addition, we performed polymerase chain reaction (PCR) to detect MRSA directly from nasal swab samples using the GeneOhm‐MRSA assay (BD Becton Dickinson) on the Cepheid SmartCycler (Cepheid). VRE detection by PCR was performed using a locally developed assay for vanA/vanB detection with the ABI 7500 FAST Real Time PCR System (Applied Biosystems). Conventional methods and RT‐PCR were performed in parallel as part of our validation of the PCR assays. The performance of PCR was excellent for both MRSA (sensitivity and specificity, 98%) and VRE (sensitivity, 90%; specificity, 99.6%), compared with conventional culture methods.⁸ We defined a patient as a carrier of MRSA or VRE if either culture or PCR assay results were positive.

A review of medical records was performed to collect data, including age, sex, underlying illness, diagnosis at hospital admission, admitting service, and admission source. For the year prior to admission, data regarding antibiotic use, hospital admission (Veterans Affairs and non–Veterans Affairs hospitals), use of medical device, and/or surgical procedure was recorded.

Univariate analyses were performed using Stata software, version 9.2 (StataCorp), to examine associations with MRSA carriage, VRE carriage, and MRSA or VRE carriage. We then calculated the sensitivity, specificity, and 95% CIs to assess the ability of single variables to identify patients colonized with MRSA or VRE.

Results

Of 422 patients screened, 421 were screened for MRSA, 381 were screened for VRE (41 [9.7%] refused to provide a perirectal swab sample), and 380 were screened for both MRSA and VRE. Of the 421 patients who were screened for MRSA, 43 (10.2%) had positive results, and of the 381 patients screened for VRE, 17 (4.5%) had positive results. Overall, 54 (13%) of the 422 screened subjects were carriers of either MRSA or VRE, and 6 (1.4%) tested positive for both MRSA and VRE. Demographic and clinical variables for MRSA and/or VRE carriers and noncarriers are shown in Table 1.

Table 1.  Characteristics of Patients Carrying Methicillin‐Resistant Staphylococcus aureus (MRSA), Vancomycin‐Resistant Enterococcus (VRE), and MRSA or VRE at Admission to a Veterans Affairs Medical Center, Compared With Noncarriers.

Open New Window

In seeking a simple predictor for the variable “MRSA or VRE carriage,” we confined our analysis to the 380 subjects who were screened for both MRSA and VRE. Among all variables, prior hospital admission within 1 year had the highest sensitivity (64%; 95% CI, 50%–78%), with a specificity of 52% (95% CI, 46%–57%), a positive predictive value of 17%, and a negative predictive value of 90% for MRSA or VRE carriage at hospital admission. If this screening criterion had been applied, testing at hospital admission of 192 (51%) of 380 patients would have been required and still would have missed 18 (35%) of 51 patients who were carriers of either MRSA or VRE. Combinations of variables resulted in either increased sensitivity at the expense of near universal screening or increased specificity at the expense of unacceptably low sensitivity (data not shown). Use of this prediction rule for either MRSA or VRE carriage alone demonstrated similar performance characteristics (Table 2).

Table 2.  Indices of Validity and Utility for a History of Hospital Admission Within the Previous Year as a Predictor of Colonization With Methicillin‐Resistant Staphylococcus aureus (MRSA), Vancomycin‐Resistant Enterococcus (VRE), and MRSA or VRE at Admission to a Veterans Affairs Medical Center.

Open New Window

Discussion

Any prediction rule to guide active surveillance culturing for MRSA or VRE should be sensitive, in order not to miss carriers who could then serve as transmission sources. In addition, such a rule should be simple, because an overly complex rule that includes multiple variables or requires patient interviews may be difficult to apply to all hospital admissions. Earlier work by Furuno et al.⁵ suggested that a history of hospital admission within the previous year might fit the criteria of a simple yet sensitive (ie, 90%) predictor of MRSA or VRE carriage. Unfortunately, we found that this rule may not be easily generalizable to our VAMC population. The lower sensitivity that we found would result in missing an unacceptably high proportion of those patients who are colonized with MRSA or VRE on admission.

A limitation of our study was its reliance on medical records alone to acquire data on predictor variables. We did not perform patient interviews at hospital admission, as did Furuno et al.⁵ However, we felt that a rule that could be applied solely using electronic medical records would be preferable, and the VAMC system has long had in place an excellent computerized record system that should be extremely reliable in determining prior exposure to health care.

Additional limitations of our study include a relatively small sample size and the absence of separate derivation and validation cohorts. Limitations of our microbiological methods include reliance upon nares swab samples alone for MRSA detection, which misses those patients who carry MRSA at other sites (eg, the throat and the perirectal region) without nares carriage^9,10 and accepting either positive PCR or culture results as true‐positive results, which could misclassify false‐positive PCR results.

Our study underscores the importance of carefully assessing the performance of any prediction rule in various healthcare settings. An additional concern relates to the changing epidemiology of drug‐resistant pathogens, particularly MRSA. During our study, <3% of all MRSA isolates belonged to the USA300 strain, with the vast majority being related to the USA100 health care–associated strain (data not shown)⁸. If community‐associated MRSA carriage becomes more common among the veteran population, prediction rules that rely on healthcare exposures may become even less useful.

In summary, a history of hospital admission within 1 year was the most sensitive single predictor of VRE or MRSA carriage at admission in our veteran population. However, this predictor still missed more than one‐third of carriers and required screening of more than one‐half of admitted patients. We conclude that universal screening may be required in our VAMC patient population if the aim is to detect most or all MRSA and VRE carriers. Since we began this study, the VA healthcare system has implemented universal screening for MRSA for all newly admitted patients.