JSTOR

Study Overview

NorthShore University HealthSystem (formerly Evanston Northwestern Healthcare; Evanston, IL) is a 3‐hospital, 850‐bed acute care organization with approximately 40,000 annual admissions. Since August 1, 2005, universal surveillance for MRSA colonization at the time of admission has been practiced. All nonneonatal patients are evaluated for colonization at admission by real‐time polymerase chain reaction (PCR; Becton Dickinson) analysis of swab samples of both nares.^12,13 Although PCR is very sensitive, it does have a relatively high false‐positive rate. Accordingly, all MRSA‐positive test results are confirmed by culture on Columbia colistin‐nalidixic acid agar with 5% sheep blood (Remel). Results of surveillance testing were classified as positive for MRSA if PCR and culture detected MRSA in a nares swab sample obtained at admission or if PCR detected MRSA in a nares swab sample obtained at admission and culture detected MRSA in a sample obtained from any body site in the past 12 months.¹⁴ Cultures of specimens from other body sites were performed at the discretion of the physician (ie, if there was clinical indication to do so). Patients with either positive results of surveillance testing or a MRSA‐positive clinical culture within 2 days after admission were considered to be MRSA carriers. MRSA isolates recovered from cultures of nares specimens (approximately 90% of all isolates in this study) were tested for high‐level mupirocin resistance, using a validated in‐house PCR for detection of mupA.¹⁵

It is a hospital guideline to attempt decolonization of all MRSA carriers by means of the following regimen: application (by finger or by cotton swab) of mupirocin calcium 2% twice daily to the nares and washing or showering with chlorhexidine gluconate 4% every 2 days (for a total of 3 doses) during the 5‐day period of mupirocin treatment. It is recommended that patients who are discharged before completing therapy be sent home with prescriptions to complete the regimen. Recommendations do not differ for patients with pressure ulcers or ongoing MRSA infection. During the study period, decolonization therapy was prescribed at the discretion of the attending physician, and we observed considerable variability in its use.

This investigation comprises 2 retrospective cohort studies. Study 1 evaluated the impact of decolonization therapy on subsequent colonization in a cohort of patients who were carrying MRSA at the time of admission and were later readmitted and retested. Study 2 sought to determine the impact of decolonization therapy on subsequent development of MRSA infection in a cohort of patients who were carrying MRSA but did not have clinical infection at the time of admission.

Subjects

Study 1: decolonization. The study interval was November 1, 2006, through December 31, 2007. All nonneonatal patients admitted overnight during this interval were eligible for study inclusion if the following criteria were met: (1) surveillance testing was performed at the time of initial admission, (2) either surveillance testing or a clinical culture performed during the first 2 days of hospitalization was positive for MRSA, and (3) the patient was subsequently readmitted during the study interval, and a surveillance test was performed during that second admission. Patients were excluded if they were discharged after the first admission with a prescription for mupirocin or chlorhexidine, because we could not determine how many doses of these medications were actually administered to these individuals. Patients were only eligible for inclusion once, regardless of their number of admissions.

Study 2: infection prevention. The study interval was the same as in study 1. All nonneonatal patients admitted overnight during this interval were eligible for inclusion if the following criteria were met: (1) surveillance testing was performed, and (2) they did not have a clinical culture indicative of MRSA infection within 30 days before and 3 days after surveillance testing. As in study 1, patients were excluded if they were discharged with a prescription for mupirocin or chlorhexidine. Patients were only eligible for inclusion once, regardless of their number of admissions. Only the last admission during the study period for each eligible patient was chosen, so that the follow‐up period would not include multiple episodes of decolonization therapy. All eligible patients were followed through March 17, 2008 (the latest date for which we had complete data at the time of analysis), for the onset of MRSA infection. MRSA infection was detected through a review of the microbiology laboratory information system for inpatient or outpatient clinical cultures yielding MRSA any time between initial testing that revealed MRSA and March 17, 2008. All MRSA‐positive cultures were reviewed according to standardized criteria to determine whether they represented true infection.³ Importantly, extensive follow‐up data were available for these patients. Our healthcare organization comprises 3 geographically proximate hospitals and more than 60 outpatient sites; all hospitals and most outpatient sites send all relevant specimens to our central laboratory for microbiological analysis. Of the patients included in this study, 85% had multiple visits to these inpatient or outpatient sites during the study period. Thus, the potential for detecting a subsequent MRSA infection by culture was high.

Data Collection

For all study patients, the electronic medical record, laboratory information system, and hospital administrative databases were used to determine past MRSA‐positive cultures, admission characteristics, age, sex, race, home address, recent admission history, and International Classification of Diseases, Ninth Revision (ICD‐9) diagnostic codes. MRSA‐positive clinical cultures were established or rejected as representing infection on the basis of systematic chart review, as described above. The method of Elixhauser et al.¹⁶ was used to detect certain comorbidities (cancer, chronic lung disease, congestive heart failure, diabetes mellitus, and renal disease) on the basis of ICD‐9 codes. The presence of a pressure ulcer was established through medical record abstraction, as was temperature on admission. Long‐term care facility (LTCF) residence was determined through review of nursing admission notes and home addresses of all patients. Only centers that primarily provide nursing care were considered LTCFs. Pharmacy records were reviewed to identify the receipt (while an inpatient) of any doses of MRSA‐active antimicrobial agents (clindamycin, daptomycin, fluoroquinolones, linezolid, rifampin, tetracyclines, tigecycline, trimethoprim‐sulfamethoxazole, or vancomycin) or any MRSA‐inactive antimicrobial agents (aminoglycosides, β‐lactams, macrolides, or metronidazole). Institutional review board approval was obtained for all data collection.

Statistical Analysis

For each study, patients were grouped into 1 of the following 4 groups according to the number of doses of mupirocin received: 0 doses, 1–3 doses, 4–7 doses, or 8 or more doses. These 4 groups of patients were compared with respect to demographic characteristics, comorbidities, and admission and clinical characteristics, using the Kruskal‐Wallis test (for length of stay) and χ² test (for all other characteristics).

Because treatment assignment was nonrandom, multivariable analysis was used to control for differences between treatment groups. For study 1, multiple logistic regression was used to determine the effect of treatment on continued colonization. For study 2, Cox proportional hazards modeling was used to evaluate the effect of treatment on the time to MRSA infection. For both studies, all variables with a univariate P value of less than .05 were entered into a model, and variables were then removed 1 at a time until only significant predictors remained. To ensure appropriate adjustment for confounding, if parameter estimates for the effect of mupirocin (or chlorhexidine) changed by more than 10% upon removal of a nonsignificant variable, that variable was retained in the model even if its final P value was .05 or greater. For each study, 2 versions of the multivariable model were generated, the first “forcing” the inclusion of mupirocin dose as an independent variable (ie, a group was included regardless of whether the univariate P value was less than .05) and the second “forcing” the inclusion of the chlorhexidine dose. In a separate analysis, the times to infection among patients who did and those who did not receive mupirocin treatment were compared using the Wilcoxon 2‐sample test.

For study 1, assuming a colonization persistence rate of 50% among untreated patients, approximately 66 subjects were required in each of the 4 mupirocin groups to ensure a power of 80% and an α level of 0.05 for detecting a 50% difference in treatment effect between groups. Assuming an infection rate of 20%,¹⁷ approximately 130 patients were required in each treatment group to ensure a power of 80% and an α level of 0.05 for detecting a 50% effect of mupirocin on infection.

Study 1: Decolonization

During the study period, there were 48,203 overnight nonneonatal admissions. Testing for MRSA was done during 43,504 admissions (90.3%). There were 1,357 unique patients found to be MRSA carriers. Of these, 483 were readmitted during the study interval and retested for MRSA colonization. A total of 76 readmitted patients were excluded from study participation because they were discharged receiving mupirocin and/or chlorhexidine therapy during their initial hospitalization. The remaining 407 patients constituted the cohort in this study (Figure 1).

(62 KB)

Figure 1.  Flow chart for studies to evaluate the impact of methicillin‐resistant Staphylococcus aureus (MRSA) decolonization therapy on subsequent MRSA colonization (study 1) or MRSA infection (study 2). CHG, chlorhexidine.

Open New Window

Most patients (69%) were older than 70 years, most (88%) were white, and most (91%) were admitted to the internal medicine service. A majority (64%) had been admitted previously in the past year, and many (41%) had diabetes mellitus (Table 1). Compared with patients who received at least 8 doses of mupirocin, patients who received none were more likely to be young and nonwhite; they were also less likely be febrile on admission, to be directly admitted to the intensive care unit, and to have chronic lung or heart disease. They also had a shorter mean length of stay (Table 1).

Table 1.  Characteristics of study patients

Open New Window

Patients were readmitted a mean (± standard deviation [SD]) of days and a median of 48 days (range, 2–371 days) after first testing positive for MRSA. At the time of readmission, 144 (47.8%) of 301 patients who received any quantity of mupirocin were still colonized, in contrast to 67 (63.2%) of 106 patients who did not receive mupirocin ( ). On multivariable analysis, residence in a LTCF (odds ratio [OR], 1.8 [95% confidence interval {CI}, 1.1–3.2]) and the presence of a pressure ulcer (OR, 2.3 [95% CI, 1.2–4.4]) both predicted continued colonization. In this model, the receipt of 1–3, 4–7, and 8 or more doses of mupirocin were all associated with a significant reduction in the odds of continued colonization, compared with no receipt of mupirocin (Table 2), although the overall effect of including mupirocin receipt in the final multivariable model was associated with a P value of .076. High‐level mupirocin resistance in an MRSA isolate predicted continued colonization (OR, 4.1 [95% CI, 1.6–10.7]). The impact of mupirocin receipt on the odds of continued colonization appeared to be greatest 30–60 days after the initial admission (Figure 2). In a separate multivariable model that included chlorhexidine dose instead of mupirocin dose, there was no significant relationship between chlorhexidine use and subsequent colonization ( ; data not shown).

Table 2.  Findings of Univariate Analysis and Multivariable Modeling to Predict Persistent Colonization Among 407 Known Carriers of Methicillin‐Resistant Staphylococcus aureus (MRSA) at Their First Readmission Subsequent to MRSA Detection (Study 1)

Open New Window

(25 KB)

Figure 2.  Duration of methicillin‐resistant Staphylococcus aureus (MRSA) colonization in 407 carriers (study 1). For the mupirocin‐treated patients, the percentage who were still colonized with MRSA at 30–60 days differed significantly from the percentage of untreated patients who were still colonized at <30 days ( ) and >120 days ( ).

Open New Window

Study 2: Infection Prevention

Of 1,357 unique patients admitted with MRSA during the study period, 243 were excluded because they were already infected at the time of admission. Of the remaining 1,114 patients, 181 were excluded because they were discharged receiving mupirocin and/or chlorhexidine therapy. The remaining 933 patients constituted the cohort in this study (Figure 1). The characteristics of these patients were similar to those of patients in study 1. Compared with patients who received 8 or more doses of mupirocin, patients who received none were more likely to be young and male; they were also less likely be febrile on admission, to be directly admitted to the intensive care unit, or to have various comorbidities. They also had a shorter mean length of stay (Table 1).

Patients were followed up for a mean (±SD) of days and a median of 269 days (range, 4–502 days). Sixty‐nine (7.4%) of these initially uninfected patients subsequently developed at least 1 MRSA infection during the follow‐up period. Twelve (17.4%) were bacteremic (Table 3). In the multivariable analysis, independent risk factors for infection were length of stay (hazard ratio [HR], 1.2 per 5 additional days [95% CI, 1.0–1.4]), chronic lung disease (HR, 1.7 [95% CI, 1.0–2.8]), and receipt of antimicrobial agents not active against MRSA (HR, 1.8 [95% CI, 1.1–3.1]). Receipt of mupirocin did not affect the risk of subsequent development of infection. However, there was a trend toward an impact of mupirocin treatment on infection timing. Median times to infection were 15.5 days (range, 4–193 days) among patients who received no mupirocin and 50 days (range, 4–289 days) among patients who received any mupirocin ( ), even though the median durations of follow‐up were the same (272 and 285 days, respectively [range, 4–502 days for both]; ).

Table 3.  Site of First Infection for Patients Carrying Methicillin‐Resistant Staphylococcus aureus (MRSA) Who Subsequently Developed MRSA Infection

Open New Window

In a separate multivariable model that included chlorhexidine dose instead of mupirocin dose, there was no significant relationship between chlorhexidine use and risk of infection ( ; data not shown). To adjust for the variation in the follow‐up time for different patients, we conducted a sensitivity analysis in which event times were truncated at 1 year. Model results were nearly identical to those from the analyses that used all available data (data not shown).

Table 4.  Findings of Univariate Analysis and Multivariable Modeling to Predict Methicillin‐Resistant Staphylococcus aureus (MRSA) Infection in 933 Carriers (Study 2)

Open New Window