Evaluation of an Intervention Designed to Decrease the Rate of Nosocomial Methicillin‐Resistant Staphylococcus aureus Infection by Encouraging Decreased Fluoroquinolone Use
Objective. Society for Health Care Epidemiology guidelines recommend decreasing the use of fluoroquinolone antibiotics in institutions where methicillin‐resistant Staphylococcus aureus (MRSA) is endemic. We evaluated whether an intervention to limit fluoroquinolone use was associated with a lower rate of nosocomial MRSA infection and summarized changes in antibiotic use, changes in other variables potentially correlated with a lower rate of MRSA infection, and rates of nosocomial infections due to other pathogens.
Design. Single‐center quasi‐experimental design. A time series of nosocomial MRSA infections was measured at monthly intervals from July 2001 through June of 2004; there were 80 MRSA infections recorded. Segmented regression analysis (ie, quasi‐Poisson generalized linear models) was used to evaluate variables possibly associated with the nosocomial MRSA infection rate.
Setting. An 87‐bed Veterans Affairs teaching hospital with an extended‐care facility.
Intervention. A physician‐directed computer‐generated intervention designed to limit the use of fluoroquinolone antibiotics was initiated, and institutional changes in antibiotic use and nosocomial MRSA infection rates were tracked.
Results. After the intervention, fluoroquinolone use decreased by approximately 34%, and levofloxacin use decreased by approximately 50%. Decreased fluoroquinolone use was offset by increased cephalosporin, piperacillin‐tazobactam, and trimethoprim‐sulfamethoxazole use. The nosocomial MRSA infection rate decreased from 1.37 to 0.63 episodes per 1,000 patient‐days after the study intervention (
). Coagulase‐negative Staphylococcus and Enterococcus infection rates also decreased. However, the rate of infection with gram‐negative organisms increased. The rate of MRSA infection was positively correlated with levofloxacin use (
) and azithromycin use (
), whereas it was negatively correlated with summer season (
). In a subsequent model, the rate of MRSA infection was negatively correlated with the study intervention (
).
Conclusion. Reduction in the institutional use of fluoroquinolones may be associated with a lower nosocomial MRSA infection rate.
Received March 15, 2005; accepted August 8, 2005; electronically published February 8, 2006.
Several recent studies have described an association between fluoroquinolone use and methicillin‐resistant Staphylococcus aureus (MRSA) colonization or infection.1‐9 This association is noteworthy, because the fluoroquinolone levofloxacin ranks second only to cefazolin nationally as the antibiotic most frequently prescribed to hospital inpatients.10
In May 2003, the Society for Healthcare Epidemiology of America (SHEA) published guidelines for preventing nosocomial transmission of multidrug‐resistant strains of Staphylococcus aureus and Enterococcus organisms.11 One of the SHEA recommendations suggested that institutions where MRSA is endemic should consider limiting the use of broad‐spectrum antibiotics, especially fluoroquinolones. In July 2003, a computer‐generated intervention designed to limit the use of fluoroquinolone antibiotics for treatment of inpatients was implemented. The intervention was a prompt that was inserted next to fluoroquinolone selections on the electronic order‐entry screen accessed by physicians for prescription of antibiotics to inpatients. The prompt described the SHEA recommendation regarding fluoroquinolone use in institutions where MRSA is endemic. Subsequent changes in antibiotic use, potential nonantibiotic risk factors for MRSA infection, and nosocomial infections were tracked and analyzed with segmented regression analysis. The purpose was to evaluate whether the intervention to limit fluoroquinolone use was associated with a lower rate of nosocomial MRSA infection and to summarize changes in antibiotic use, changes in other variables potentially correlated with a lower rate of MRSA infection, and rates of nosocomial infections due to other pathogens.
Methods
The Veterans Affairs Medical Center in Boise, Idaho, is a university‐affiliated, 87‐bed teaching hospital that consists of a 10‐bed medical‐surgical intensive care unit, a 23‐bed medical‐surgical step‐down unit, a 14‐bed general medical surgical unit, a 9‐bed psychiatric unit, and a 31‐bed attached extended care unit. There appeared to be an association between addition of levofloxacin to the hospital formulary in the late fall of 1998 and a subsequent increase in isolation of MRSA. Local guidelines for antibiotic use and in‐service training sessions for residents were structured to educate physicians about this apparent association. However, limited published data were available at that time to support the clinical observation, and the interventions were only temporarily successful (ie, for 1‐2 months) at reducing fluoroquinolone use.
In July 2003, a computer‐generated intervention to limit the use of fluoroquinolone antibiotics was created on the basis of the SHEA guidelines.11 The intervention was a prompt that was inserted next to fluoroquinolone selections on the electronic order‐entry screen accessed by physicians to prescribe medications. In addition, if a physician chose to order a fluoroquinolone, the subsequent screen asked them to confirm their need for a fluoroquinolone. The prompt asked physicians to prescribe alternative antibiotic agents when possible in place of fluoroquinolones in accordance with local antibiotic use guidelines of the infectious diseases service at the Boise Veterans Affairs Medical Center (Appendix, Table A).
The study used a single‐center retrospective/prospective quasi‐experimental design.19 The number of nosocomial MRSA infections was measured at monthly intervals from July 2001 through June of 2004 for time‐series analysis; 80 infections were detected. The study design was inspected by the human subjects committee at the University of Washington (Seattle) and was determined to be exempt from review.
Data on nosocomial infections were obtained from monthly infection control reports. Nosocomial infection surveillance at this facility is hospital wide and population based. One infection control practitioner performed active retrospective and prospective surveillance coupled with postdischarge surveillance of patients who had undergone surgery. Screening for nosocomial infections was performed through daily review of hospital admissions and discharges, intravenous antibiotic use by patients admitted to the emergency department, and laboratory reports with case confirmation by review of medical records. Medical records of patients who had undergone surgery were reviewed monthly for reports of nosocomial infection. Standard Centers for Disease Control and Prevention case definitions for classification of nosocomial infections were used.20 An infection was assumed to be caused by MRSA if cultures of blood, intravenous line, sputum, urine, tissue, or stool obtained at the time of symptom development yielded MRSA.
Infection control measures during the study period consisted of standard universal precautions, with patients who had known MRSA colonization or infection placed under contact isolation precautions. Active surveillance cultures were not systematically performed during the study period, and there were no institutional policies governing decolonization of MRSA carriers. Both nosocomial infection surveillance and infection control measures were relatively constant throughout the study, unless otherwise specified.
Explanatory variables that could influence the rate of nosocomial MRSA infection were identified from published research and professional experience.21,22 Variables were formulated to measure time trends.
Data on antibiotic use were collected from dispensing records and converted to be expressed in defined daily doses (DDDs).23 Formulary data for the following intravenous and oral antibiotics were obtained: amoxicillin, amoxicillin‐clavulanate, ampicillin, ampicillin‐sulbactam, azithromycin, cefazolin, cefoxitin, cefuroxime, ceftriaxone, cefotaxime, cefpodoxime, ceftazidime, cephalexin, ciprofloxacin, clindamycin, dicloxacillin, doxycyline, imipenem, gatifloxacin, gentamicin, levofloxacin, linezolid, metronidazole, nafcillin, piperacillin‐tazobactam, tetracycline, trimethoprim‐sulfamethoxazole, and vancomycin.
Nonantibiotic data included the number of central venous catheter (CVC)–days, ventilator‐days, and patient‐days of hospitalization; the nurse staffing ratio, defined as the ratio of actual to target hours worked by nursing staff per day, which was averaged to get monthly levels; and hand hygiene measures used, determined by reviewing purchase orders for alcohol foam dispensers and liters of 0.5% triclosan medicated soaps. Seasonal effect variables were included based on clinical observation.
Antibiogram data, in addition to organism‐specific nosocomial infection data, were presented as a secondary outcome measure. Antibiogram data were calculated in accordance with NCCLS guidelines.24 Only the first isolate of a given species per patient per analysis period (irrespective of the anatomical sampling site or the antimicrobial susceptibility profile of the isolate) was included in the antibiogram. Surveillance data and results of cultures of environmental samples were not included. The method for the tabulation of isolates in the antibiogram has been used since before 1994. Because antibiogram data were only available for cumulative 6‐month reporting intervals, these data were not analyzed with multivariate techniques.
In addition to the study intervention, there were several nonstudy interventions that could have impacted the rate of nosocomial MRSA infection. First, use of chlorhexidine‐based skin preparations for CVC insertion and use of chlorhexidine‐infused adhesive skin dressings (biopatches) administered after CVC placement were initiated in January 2003 and June 2003, respectively. Data on quantities of these products that were used were obtained from purchase orders. Second, in April 2003, there was an outbreak of norovirus infection in the hospital. The infection control service intervened by educating staff to use standard hand hygiene techniques involving soap and water, instead of alcohol foam, and by using “deep‐cleaning” methods to disinfect environmental surfaces.
Changes in antimicrobial use and other nonantibiotic hospital care explanatory variables were determined by segmented regression analysis.25 The slope of the regression line (ie, the time trend) before the study intervention, the change in level after the intervention, and the change in the slope after the study intervention were measured for each variable; the denominator for all variables except the nursing staff ratio was 1,000 patient‐days. The full model, a subset model, or no model was determined for each variable by means of a backward selection procedure. If the change in the slope after the intervention was statistically significant (
), then both lower‐order terms (ie, the slope during the preintervention period and the change in level during the postintervention period) were kept in the final model, regardless of their P values. If the change in the slope was not statistically significant, then the final segmented regression model included the slope for the entire study period and/or the change in level (if the change was statistically significant).
Because of the low number of infections, changes in pathogen‐specific nosocomial infection rates were analyzed by segmented quasi‐Poisson generalized linear models.26 Quasi‐Poisson models control for potential overdispersion or underdispersion of data. To test whether the study intervention had a statistically significant negative correlation with the MRSA infection rate, the following approach was taken. First, a best model without the study intervention variable (hereafter, “the testing model”) was developed, and then a test was performed by adding the study intervention variable to the testing model. Fluoroquinolone variables (ie, levofloxacin use and ciprofloxacin use) were also tested in the same way. The testing model was developed by backward selection, which initially included all variables determined be significant at an α level of .20 by univariate regression analysis. Both time‐lagged and unlagged variables were considered in model selection. Antibiotics prescribed to treat MRSA infection (ie, vancomycin and linezolid) were excluded from model selection.4 Backward selection removed variables with a P value of greater than .10 one at a time. For consistency with the cutoff α level of .10 for inclusion of a variable in the model, 90% confidence intervals were reported for regression coefficients. For consistency with the study objective to test whether the study intervention was associated with a lower rate of nosocomial MRSA infection, study intervention and fluoroquinolone analyses were 1‐sided tests.
When a high pair‐wise correlation between variables was observed, the model was evaluated for confounding. Confounding was defined as either a change of more than 20% in the correlation coefficient of the test variable or as a resulting P value that was not statistically significant (
). To test for confounding, variables were analyzed in a model with and a model without potential confounders.
A final check was made for confounding between the study intervention variable and select nonstudy intervention variables. Select nonstudy intervention variables included the environmental cleaning intervention and those not included in the testing model for which there was a relatively large difference in level between the preintervention period and the postintervention period. Each of these nonstudy intervention variables was added to a final intervention model. The process was repeated for a final levofloxacin model.
All statistical analyses were performed using the open‐source statistics language R, version 1.91.27
Results
There were 67 nosocomial MRSA infections before the intervention and 13 after the intervention. Of the 67 infections during the preintervention period, 34 (51%) were lower respiratory tract infections, 13 (19%) were bloodstream infections (BSIs), 10 (15%) were urinary tract infections, 9 (13%) were skin and deep structure (ie, tendon, joint, or bone) infections, and 1 (<2%) was enterocolitis. Of the 13 infections during the postintervention period, 5 (38%) were lower respiratory tract infections, 0 were BSIs, 5 (38%) were urinary tract infections, 1 (8%) was a skin and structure infection, 1 (8%) was a peripheral intravenous catheter infection, and 1 (8%) was endophthalmitis. The annual rate of MRSA infection decreased for all infection types except urinary tract infection, and the most pronounced decrease was observed for BSIs.
Changes in antimicrobial use and other potential explanatory variables during the 24‐month period before and the 12‐month period after the intervention are summarized in Table 1. Over the entire study period, both linezolid and piperacillin‐tazobactam use increased, whereas vancomycin use decreased. The study intervention was associated with statistically significant decreases in the use of several antibiotics. Notably, use of aminopenicillin β‐lactamase inhibitors, all fluoroquinolones, and levofloxacin decreased, whereas use of first‐generation cephalosporins and trimethoprim‐sulfamethoxazole increased. After the intervention, negative changes in the slope of the regression lines for levofloxacin and aminopenicillin use were observed, and positive changes in the slopes were observed for third‐generation cephalosporin use and ciprofloxacin use. Total fluoroquinolone use decreased by 34% (from 129 to 85 DDDs per 1,000 patient‐days), and levofloxacin use decreased by 50% (from 116 to 58 DDDs per 1,000 patient‐days).
In January 2004, the National Veterans Affairs formulary contract stipulated the replacement of levofloxacin with gatifloxacin. However, this policy was implemented slowly at the local level beginning in February 2004, and levofloxacin was prescribed until the end of April 2004. During the preintervention period, fluoroquinolone use involved levofloxacin 90% of the time and ciprofloxacin 10% of the time. After the intervention, fluoroquinolone use involved levofloxacin 69% of the time, ciprofloxacin 22% of the time, and gatifloxacin 9% of the time.
There were significant differences between the preintervention period and the postintervention period for several nonantibiotic variables, as well. Over the entire study period, purchase of chlorhexidine skin preparation increased, and the number of ventilator‐days, purchase of alcohol foam, and the nursing staff ratio decreased. Positive changes in the slope of the regression lines were observed for all of the nonantibiotic variables except triclosan soap purchases (for which the change was negative) and ventilator‐days (for which the slope was unchanged). Evaluation of the influence of CVC‐directed nonstudy interventions and other potential confounders on the reduction in the rate of nosocomial MRSA infection are addressed in subsequent sections.
The Figure illustrates changes in the rate of MRSA, as documented in the hospital antibiogram, at 6‐month intervals. The percentage of S. aureus isolates that were resistant to methicillin increased steadily during 1984‐2000 and fluctuated seasonally between 34% and 42% since 2000. After the study intervention, the MRSA percentage reported via antibiogram findings decreased from 41% to 29%, with a slight increase to 30% at the end of the study period. After completion of the study, antibiogram analysis on December 30, 2004, revealed an MRSA percentage of 25%.
Figure. Antibiogram showing the methicillin‐resistant Staphylococcus aureus (MRSA) percentage (defined as the no. of MRSA isolates/no. of S. aureus isolates submitted for testing) every 6 months from January 1984 through December 2004 at the Boise Veterans Affairs Medical Center (Boise, ID) (
data points). CL = confidence limit (average); LCL = lower confidence limit; SHEA = Society for Healthcare Epidemiology;11 UCL = upper confidence limit.
Table 2 summarizes antibiogram‐based susceptibility rates for the 6 pathogens most commonly recovered before the study intervention, as well antibiogram‐based antibiotic susceptibility rates for these pathogens 6 and 12 months after the study intervention. These 6 pathogens accounted for approximately 80% of the isolates reported in the hospital antibiogram. In general, β‐lactam susceptibility rates increased among Staphylococcus species. Fluoroquinolone susceptibility rates increased among Staphylococcus species and gram‐negative organisms. Gatifloxacin replaced levofloxacin before creation of the third antibiogram, which revealed a higher rate of susceptibility to gatifloxacin among S. aureus, compared with levofloxacin. Piperacillin‐tazobactam activity against gram‐negative organisms remained unchanged, but the activity of trimethoprim‐sulfamethoxazole against E. coli decreased.
Changes in nosocomial infection rates associated with specific pathogens before and after the study intervention are summarized in Table 3. The overall rate of nosocomial infection increased throughout the study, even after correcting for the norovirus outbreak (β 
; 
). The rate of nosocomial MRSA infection decreased from 1.37 cases per 1,000 patient‐days before the intervention to 0.63 cases per 1,000 patient‐days after the intervention (
). Multivariate models that describe the relationship between the rate of MRSA infection and potential explanatory variables are discussed in later sections.
There were also large decreases in the rates of infection due to coagulase‐negative Staphylococcus species (40% decrease from the preintervention rate) and Enterococcus species (43.93% decrease from the preintervention rate). The rate of nosocomial infection with gram‐negative organisms increased by 22.66% after the intervention, which was significantly associated with the increase in the rate (
) and the decrease in the slope (
). Infections due to gram‐negative Enterobacteriaceae increased by 19.26%, and those due to nonfermentative gram‐negative bacilli increased by 49.2%, although individually, these groups of pathogens were not associated with significant changes in slope or infection rate.
The model selection process involved creation of a testing model (model 1) that included aminopenicillin use lagged 1 month (lag 1), azithromycin use (lag 1), and an indicator for summer season (defined as the period from July through September) (Table 4). Aminopenicillin and azithromycin use were positively correlated with the rate of MRSA infection, whereas summer exhibited a protective effect. Next, levofloxacin use (lag 1) was added to the testing model to test for a positive correlation with the MRSA infection rate (model 2; 
). Results of the test were inconclusive, because a high pair‐wise correlation between levofloxacin and aminopenicillin use (
) suggested potential confounding. Confounding was confirmed by removing aminopenicillin use from the model (model 3; 
for levofloxacin use), which resulted in a large increase of 28.66% in the levofloxacin coefficient. Ciprofloxacin use was not associated with the MRSA infection rate (
). Gatifloxacin was not used before the intervention. Model 4 tested for a negative correlation between the intervention and the rate of MRSA infection (
). The study intervention was also confounded with aminopenicillin use, because removing aminopenicillin use from the model (model 5; 
for the study intervention) resulted in large decreases in the β coefficient and P value (0.8489 and .14, respectively). The observed confounding between aminopenicillin use and the study intervention and levofloxacin use is potentially explained by the large decrease of 43.93% in the rate of enterococcal infection after the study intervention. After the intervention, piperacillin‐tazobactam (the use of which increased by 46.05%) appears to have been used as one substitute for levofloxacin (the use of which decreased by 50.33%). An increase in piperacillin‐tazobactam use would be expected to result in a lower rate of enterococcal infection, which would, in turn, decrease the demand for aminopenicillins. In support of this hypothesis, piperacillin‐tazobactam use (lag 1) was associated with a lower rate of enterococcal infection (β = −0.0243; 
). The estimated dispersion was 1.12 for model 1; 1.09 for model 2; 1.18 for model 3; 1.12 for model 4; and 1.28 for model 5. Given the slight overdispersion observed, the quasi‐Poisson model is more conservative than the classical Poisson model with respect to conferring statistical significance on variables. In addition, there was no autocorrelation to control for in models 1‐5. The residuals from each of these models had insignificant negative autocorrelation (
for all, by the Ljung‐Box test).
Several select variables (ie, nonstudy interventions) exhibited relatively large differences of more than 20% between the preintervention and postintervention periods but were removed during selection of the testing model. The environmental surface disinfection intervention was also removed. Table 5 summarizes the extent to which the select variables are confounded with levofloxacin use (model 3) and the study intervention (model 5). The select variables were added one at a time to models 3 and 5. Inclusion of all select variables except chlorhexidine‐based biopatch use resulted in small changes in the β coefficients and P values for levofloxacin use and the study intervention. The addition of chlorhexidine biopatch use to models 3 and 5 resulted in large changes in the β coefficients and P values for levofloxacin use (a decrease of 0.3266 and an increase of .08, respectively) and the study intervention (increases of 0.6853 and .29, respectively). Given the changes in chlorhexidine‐based CVC interventions, hand hygiene education, and environmental surface disinfection in the 6 months before the study intervention, it is possible that inclusion of these data in analyses of the preintervention period could bias the study intervention effect toward the null hypothesis of no reduction in the rate of MRSA infection. In other words, if these nonstudy interventions were successful in reducing the rate of MRSA infection, then changes in the β coefficients and P values for levofloxacin use and the study intervention should be statistically significant after omission of nonstudy intervention data from the 6‐month period before the study intervention. Refitting models 3 and 5 without nonstudy intervention data from January through June 2003 led to similar results. The P value for levofloxacin use in model 3 increased from .01 to .03. The P value for the study intervention in model 5 increased from .04 to .06. This is evidence against the overall effectiveness of the nonstudy interventions.
The plausibility that use of chlorhexidine‐based skin preparations was responsible for the observed decrease in the rate of MRSA infection was unlikely, given that none of the 80 case of nosocomial MRSA infection during the entire study fit the criteria for a CVC infection or a CVC‐associated BSI. Overall, the rate of CVC infection did not change significantly between the preintervention and postintervention periods (1.2 vs 1.1 infections per 1,000 patient‐days). However, the number of CVC‐related BSI decreased slightly from 4 infections (0.07 infections per 1,000 patient‐days) before the intervention to 2 infections (0.04 infections per 1,000 patient‐days) after the intervention.
Discussion
In this study, a computer‐generated intervention directed at physicians during the electronic order‐entry process for prescriptions was associated with a significant decrease in fluoroquinolone use, which, in turn, was associated with a reduction in the rate of nosocomial MRSA infection. This intervention, in conjunction with activities of the antibiotic management team, such as dissemination of local prescribing guidelines, educational promotion, and positive feedback relative to the decreased MRSA infection rate, resulted in a sustained decrease in fluoroquinolone use. In contrast, activities of the antibiotic management team alone were unsuccessful.
After the intervention, marked reductions in the number of isolates reported in the antibiogram and the rate of nosocomial MRSA infection were observed. The rates of nosocomial MRSA infection and coagulase‐negative Staphylococcus infection decreased by 53% and 40%, respectively. Infection with Enterococcus species also decreased by 44%, which appeared to be related to the increased use of piperacillin‐tazobactam. Conversely, the rate of nosocomial infection with gram‐negative organisms significantly increased by 23%, with infection due to nonfermentative gram‐negative bacilli organisms constituting the largest relative increase.
The best model for describing nosocomial MRSA infection that included a fluoroquinolone consisted of 3 terms: levofloxacin use (lag 1), azithromycin use (lag 1), and summer season. Summer had a protective effect, whereas use of these antibiotics was positively correlated with MRSA infection. A model containing the study intervention in place of levofloxacin use produced a similar 3‐term significant model. With the exception of chlorhexidine‐based biopatch use, nonantibiotic variables were not confounded with levofloxacin use or the study intervention.
Major strengths of this quasi‐experimental study include use of time‐series and segmented regression analyses. Interrupted time‐series analysis with segmented regression is a statistical technique that can control for the effects of multiple variables, including lagged time and seasonal effects, and it is the method of choice for analyzing antibiotic resistance interventions.25,28 Unfortunately, many published analyses of antibiotic interventions had a poor study design and lacked statistical validity.29 Strengths of this time‐series analysis include the relatively long period of observation (36 months).30 Analysis of at least 12 monthly data points after the intervention allowed the effects of seasonal variation to be evaluated. This analysis also compared changes of variables before and after the intervention in terms of the slope of their regression line and their level, which are necessary components of control for quasi‐experimental design. The approach to testing the fluoroquinolone variables and study intervention controlled for other significant effects, and potential confounding was fully assessed. Additional strengths of this analysis are the inclusion of data on antibiotic use and nonantibiotic healthcare measures classically associated with nosocomial MRSA infections.
General limitations of this analysis are consistent with those of other quasi‐experimental studies and include the potential for regression to the mean, maturation, and confounding for important effects.19 Regression to the mean refers to erroneously concluding that a change in the outcome is the result of an intervention when it may be due to chance. Given the consistent MRSA percentages reported in antibiograms between 2000 and the start of the study intervention, regression to the mean is an unlikely explanation for the decrease in MRSA infection. Maturation effects refer to natural changes in disease patterns or outcomes as a result of seasonality. This analysis included sufficient observation periods to identify the protective effect of summer on nosocomial MRSA infection. The major limitation of this study is its limited ability to control for potential confounding effects: because the use of a number of nonstudy interventions changed from the preintervention to the postintervention periods, such interventions could have acted as confounders with the study intervention.
Changing one aspect of antibiotic use policy frequently results in changes in use of multiple antibiotics, and in this analysis aminopenicillin use was confounded with the study intervention and fluoroquinolone use. It is possible that decreased use of aminopenicillins is associated with a decreased rate of MRSA infection. However, ampicillin and amoxicillin accounted for only 7% of antibiotic use before the study intervention. Aminopenicillin β‐lactamase inhibitors and other penicillins, which accounted for a larger proportion of total antibiotic use, were not significant explanatory variables in the testing model. A more plausible explanation is that aminopenicillins are the treatment of choice for enterococcal infection, which decreased by 44% during the postintervention period. The decrease in enterococcal infection was associated with the increased use of piperacillin‐tazobactam. Taken together, these data suggest that the decrease in aminopenicillin use was unlikely to have played a significant role in the reduction of the MRSA infection rate.
Confounding between the study intervention and the use of chlorhexidine biopatches for CVC placement is more difficult to assess. MRSA‐associated BSI decreased substantially after the intervention, and although CVC‐related infection rates did not change, CVC‐associated BSI rates decreased slightly after initiation of biopatch use. However, none of the 80 nosocomial MRSA cases during the entire study fit the criteria for a CVC infection or CVC‐associated BSI. Although it is unlikely that a large component of the reduction in the MRSA infection rate can be attributed to this nonstudy intervention, it is not possible to dismiss a potential association between use of chlorhexidine biopatches and the reduction in nosocomial MRSA infection during the postintervention period.31
There were several other nonantibiotic variables that exhibited differences between the preintervention and postintervention periods. Alcohol foam purchases increased over time, suggesting that hand hygiene practices were improving. However, alcohol foam use was not associated with nosocomial MRSA infection. The number of ventilator‐days increased throughout the study, which was consistent with increased use of noninvasive ventilatory methods beginning in 2002.32 With rare exception, our Veterans Affairs facility limits mechanical ventilation to patients in the ICU, whereas fluoroquinolone is used institution wide. The number of ventilator‐days was not associated with the overall rate of nosocomial MRSA infection or confounded with the fluoroquinolone‐related variables in multivariate models.
Other potential limitations include the inability to control for illness comorbidities, the use of purchase order data as surrogate markers for hand hygiene compliance, the lack of surveillance data to distinguish onset of colonization before onset of infection, and the lack of isolates for molecular analysis.
Nosocomial MRSA spread is difficult to control, as demonstrated in a recent large survey of approximately 670 US hospitals in which only 4% of respondents reported decreasing rates of MRSA infection.33 The substantial decrease in the nosocomial MRSA infection rate in our institution is noteworthy, because system‐wide surveillance to identify colonized patients for placement in isolation was not undertaken. Our facility did undergo widespread environmental cleaning several months before the intervention. However, environmental cleaning was not associated with the MRSA infection rate. Although environmental cleaning has been attempted to contain MRSA outbreaks in localized environments, a recent report suggested that even relocation of all patients to a new facility had minimal impact on MRSA colonization rates.34
The role of antibiotic stewardship in the containment of MRSA is also poorly defined. There are several reports of antibiotic‐related interventions that, although limited in scope, resulted in decreased rates of MRSA infection.35‐37 In several cases, the antibiotic interventions were not distinct from other interventions, and the study designs had methodological flaws.36,37 A number of case‐control and ecological analyses have identified both fluoroquinolone and macrolide antibiotics as risk factors for colonization and infection with MRSA.1‐9 In several of these studies, the magnitude of risk association between fluoroquinolone use and the rate of MRSA infection appeared to be quite large.1‐3 To our knowledge, there has been only 1 interventional study that specifically investigated the limitation of the fluoroquinolone class of agents in decreasing MRSA infection. Using an interrupted time‐series design, Charbonneau et al.38 found that a 97% decrease in fluoroquinolone use in a 2000‐bed French teaching hospital was associated with a reduction in nosocomial MRSA infections during a 12‐month period. These findings led to the recommendation in the SHEA guidelines11 to limit the use of broad‐spectrum antibiotics, especially fluoroquinolones.
The increase in nosocomial gram‐negative infection as a result of decreased fluoroquinolone use and increased cephalosporin and piperacillin‐tazobactam use is not surprising. A case‐control study designed to identify risk factors for third‐generation cephalosporin resistance in gram‐negative nosocomial pathogens identified β‐lactamase–inhibitor combinations and third‐generation cephalosporins as agents associated with increased risk, whereas fluoroquinolones were found to be protective.39
Several areas of further study are warranted. First, although several biological explanations for the association between fluoroquinolone use and MRSA have been proposed, there is still a large gap in the understanding of how exposure to these agents facilitates MRSA colonization and infection at the unit level and the patient level.40,41 A careful study of frequent use of surveillance cultures, patterns of antibiotic use, and other risk factors, combined with molecular analysis of recovered S. aureus strains for determination of antimicrobial resistance and treatment adherence factors, may provide further insight into why this relationship exists. Second, to date, almost all of the epidemiological data regarding the association between fluoroquinolones and MRSA involve older drugs. Gatifloxacin and moxifloxacin possess more potent activity against gram‐positive organisms, and further study is needed to determine whether use of these agents is associated with an increased risk of MRSA colonization and infection, compared with use of levofloxacin and ciprofloxacin. Third, because nearly all antibiotics provide an ecological advantage to certain bacteria, further research should focus on restructuring policy on broad‐spectrum antibiotic use in institutions to minimize the overall rate of nosocomial infection. Finally, spread of community‐acquired MRSA is increasing dramatically. Further investigation of risk factors associated with antibiotic use is warranted, particularly because the widespread increase in outpatient use of respiratory tract fluoroquinolones and macrolides generally paralleled the increase in the spread of community‐acquired MRSA.
In conclusion, an electronic intervention designed to decrease fluoroquinolone use was successful. The overall rate of nosocomial infection increased over the entire study period. Against this trend, large reductions in organism‐specific infection rates were observed for MRSA, well as for coagulase‐negative Staphylococcus and Enterococcus species. Nosocomial infection due to gram‐negative bacilli increased during the postintervention period. The rare if MRSA infection was positively correlated with levofloxacin use and negatively correlated with summer season. A subsequent interrupted time‐series model demonstrated evidence of an association between the study intervention and a reduction in the nosocomial MRSA infection rate. Our findings support the SHEA recommendation that institutions where nosocomial MRSA infection is endemic should consider restricting the use of fluoroquinolone antibiotics.11
Appendix
Use of intravenous therapy as initial, empirical treatment for infection is frequently necessary. However, an accurate diagnosis of infection is of paramount importance to prevent missing important pathogens, minimize antibiotic adverse events, and decrease nosocomial infections. Therapy should be modified as necessary on the basis of culture results and clinical response. Preferred regimens are composed of single antibiotics or combinations that have demonstrated clinical efficacy, provide appropriate broad‐ or narrow‐spectrum coverage, and are considered to be relatively safe or cost‐effective. In many cases, no single regimen is preferred. Physicians are encouraged to alternate the antibiotic regimens they prescribe, to limit selective pressure on any one particular regimen Patient‐specific characteristics and disease processes need to be considered when choosing appropriate antimicrobial therapy. The Boise Veterans Affairs Medical Center guidelines for use of antibiotic therapy are given in Table A.
References
- 1. Weber SG, Gold HS, Hooper DC, Karchmer AW, Carmeli Y. Fluoroquinolones and the risk of methicillin resistant Staph aureus in hospitalized patients. Emerg Infect Dis 2003; 9:1415‐1422.
- 2. Graffunder EM, Venezia RA. Risk factors associated with nosocomial methicillin‐resistant Staphylococcus aureus (MRSA) infection including previous use of antimicrobials. J Antimicrob Chemother 2002; 49:999‐1005.
- 3. Hori S, Sunley R, Tami A, Grundmann H. The Nottingham Staphylococcus aureus population study: prevalence of MRSA among the elderly in a university hospital. J Hosp Infect 2002; 50:25‐29.
- 4. Monnet DL, MacKenzie FM, Lopez‐Lozano JM, et al. Antimicrobial drug use and methicillin‐resistant Staphylococcus aureus, Aberdeen, 1996‐2000. Emerg Infect Dis 2004; 10:1432‐1441.
- 5. Crowcroft NS, Ronveaux O, Monnet DL, Mertens R. Methicillin‐resistant Staphylococcus aureus and antimicrobial use in Belgian hospitals. Infect Control Hosp Epidemiol 1999; 20:31‐36.
- 6. Dziekan G, Hahn A, Thune K. Methicillin‐resistant Staphylococccus aureus in a teaching hospital: investigation of nosocomial transmission using a matched case‐control study. J Hosp Infect 2000; 46:263‐270.
- 7. Harbarth S, Liassine N, Dharan S, Herrault P, Auckentbaler, Pittet D. Risk factors for persistent carriage of methicillin‐resistant Staphylococcus aureus. Clin Infect Dis 2000; 31:1380‐1385.
- 8. Manhold C, von Rolbieki U, Brase R, et al. Outbreaks of Staphylococcus aureus infections during treatment of late onset pneumonia with ciprofloxacin in a prospective randomized study. Intensive Care Med 1998; 24:1327‐1330.
- 9. McDonald LC, Lauderdale TL, Lo RJ, Tsai JJ, Hung CC. Colonization of HIV infected outpatients in Tiawan with methicillin‐resistant and methicillin‐susceptible Staphylococcus aureus. Int J STD AIDS 2003; 14:473‐477.
- 10. Johnson CK, Polk RE, Edmond M, Wenzel R. Trends in antimicrobial use from 1999 to 2001 in 36 US hospitals: a SCOPE‐MMIT report. In: Program and abstracts of the 42nd Annual Meeting of the Interscience Conference on Antimicrobial Agents and Chemotherapy; 2002; San Diego, CA. Abstract K‐1005.
- 11. Muto CA, Jernigan JA, Ostrowsky BE, et al. SHEA guideline for preventing nosocomial transmission of multidrug‐resistant strains of Staphylococcus aureus and enterococcus. Infect Control Hosp Epidemiol 2003; 24:362‐378.
- 12. Niederman MS, Mandell LA, Anzueto A, et al. The American Thoracic Society. Guidelines for the management of adults with community‐acquired pneumonia: diagnosis, assessment of severity, antimicrobial therapy, and prevention. Am J Respir Crit Care Med 2001; 163:1730‐1754.
- 13. Bartlett JG, Gorbach SL. Treatment of aspiration pneumonia and primary lung abscess: penicillin G vs clindamycin. JAMA 1975; 234:935‐937.
- 14. Marik PE. Aspiration pneumonitis and aspiration pneumonia. New Engl J Med 2001; 344:665‐667.
- 15. Fagon JY, Chastre J, Wolff M, et al. Invasive and noninvasive strategies for management of suspected ventilator‐associated pneumonia: a randomized trial. Ann Intern Med 2000; 132:621‐630.
- 16. Chan R, Hemeryck L, O'Regan M, Clancy L, Feely J. Oral versus intravenous antibiotics for community acquired lower respiratory tract infection in a general hospital: open, randomised controlled trial.BMJ 1995; 310:1360‐1362.
- 17. Dewan NA, Rafique S, Kanwar B, et al. Acute exacerbation of COPD: factors associated with poor treatment outcome. Chest 2000; 117:662‐671.
- 18. Baron TH, Morgan DE. Acute necrotizing pancreatitis. New Engl J Med 1999; 340:1412‐1417.
- 19. Harris A, Bradham D, Baumgartner M, et al. The use and interpretation of quasi‐experimental design in infectious diseases. Clin Infect Dis 2004; 38:1586‐1591.
- 20. Anon. National Nosocomial Infections Surveillance (NNIS) System report, data summary from January 1990–May 1999, issued June 1999. Am J Infect Control 1999; 27:520‐532.
- 21. Hartstein A, Mullingan MW. Methicillin. In: Mayall CG, ed. Resistant Staphylococcus aureus in hospital epidemiology and infection control. 2nd ed. Philadelphia: Lippincott, Williams & Wilkins; 1999.
- 22. Cooper BS, Stone SP, Kibbler CC. Systematic review of isolation policies in the hospital management of methicillin‐resistant Staphylococcus aureus: a review of the literature with epidemiological and economic modeling. Health Technology Assessment 2003; 7:1‐139.
- 23. World Health Organization Collaborating Center for Drug Statistics Methodology Web site. Available at: http://www.whocc.no/atcddd/. Accessed August 4, 2004.
- 24. NCCLS. Performance standards for antimicrobial susceptibility testing. Wayne, PA: NCCLS; 2004. NCCLS document M100‐S14.
- 25. Neter J, Kutner MH, Wasserman W, Nachtsheim CJ, Neter J. Applied linear statistical models. 4th ed. New York, NY: McGraw‐Hill/Irwin; 1996.
- 26. McCullagh P, Nelder JA. Generalized linear models. 2nd ed. Boca Raton, FL: Chapman & Hall/CRC; 1989.
- 27. R Development Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: http://www.r‐project.org. Accessed March 10, 2004.
- 28. Ansari, F, Kirsteen G, Nathawani D, et al. Outcome of an intervention to improve hospital antibiotic prescribing: interrupted time series with segmented regression analysis. J Antimicrob Chemother 2003; 52:842‐848.
- 29. Ramsay C, Brown E, Hartman G, Davey P. Room for improvement: a systematic review of the quality of evaluations of interventions to improve hospital antibiotic prescribing. J Antimicrob Chemother 2003; 52:764‐771.
- 30. The Cochrane Effective Practice and Organization of Care Group (EPOC). The data collection checklist (2002). Cochrane EPOC methods papers. Available at: http://www.epoc.uottawa.ca/. Accessed November 15, 2003.
- 31. Farr BM. Nosocomial infections related to use of intravascular devices inserted for short‐term vascular access in hospital epidemiology and infection control. In: Mayall CG, ed. Resistant Staphylococcus aureus in hospital epidemiology and infection control. 2nd ed. Philadelphia: Lippincott, Williams & Wilkins; 2003.
- 32. Mehta S, Hill NS. Noninvasive ventilation. Am J Respir Crit Care Med 2001;163:540‐577.
- 33. Davis KA, Stewart JJ, Crouch HK, Florez CE, Hospenthal DR. Methicillin‐resistant Staphylococcus aureus (MRSA) nares colonization at hospital admission and its effect on subsequent MRSA infection. Clin Infect Dis 2004; 39:776‐782.
- 34. Vierti NJ, Dooley DP, Davis CE, Longfield JN, Meier PA, Whelen AC. The effect of moving to a new hospital facility on the prevalence if methicillin resistant Staphylococcus aureus. Am J Infect Control 2004; 32:262‐267.
- 35. Frank MO, Battieger BE, Sorensen SJ, et al. Decrease in expenditures and select nosocomial infections following implementation of an antimicrobial‐prescribing improvement program. Clin Perform Qual Health Care 1997; 5:180‐188.
- 36. Landman D, Chockalingham M, Quale JM. Reduction in the incidence of methicillin‐resistant Staphylococcus aureus and ceftazidime‐resistant Klebsiella pneumoniae flowing changes in a hospital antibiotic formulary. Clin Infect Dis 1999; 28:1062‐1066.
- 37. May AK, Melton SM, McGwin G, Cross JM, Mose SA, Rue LW. Reduction of vancomycin‐resistant enterococcal infections by limitation of broad‐spectrum cephalosporin use in a trauma and burn intensive care unit. Shock 2000; 14:259‐264.
- 38. Charbonneau P, Thibon JJ, Parienti MJ, et al. Impact of a 12‐month fluoroquinolone restriction on MRSA incidence in a French university hospital. In: Program and abstracts of the 43rd Annual Meeting of the Interscience Conference of Antimicrobial Chemotherapy; 2003; Chicago, IL. Abstract K‐1743.
- 39. Schwaber MJ, Cosgrove SE, Gold H, Kaye KS, Carmeli Y. Fluoroquinolones protective against cephalosporin resistance in gram‐negative nosocomial pathogens. Emerg Infect Dis 2004; 10:94‐99.
- 40. Bisognano C, Vaudaux PE, Lew DP, Ng EY, Hooper DC. Increased expression of fibronectin‐binding proteins by fluoroquinolone‐resistant Staphylococcus aureus exposed to subinhibitory levels of ciprofloxacin. Antimicrob Agents Chemother 1997; 41:906‐913.
- 41. Evans ME, Titlow WB. Levofloxacin selects fluoroquinolone‐resistant methicillin‐resistant Staphylococcus aureus less frequently than ciprofloxacin. J Antimicrob Chemother 1998; 41:285‐288.
-
This article is the result of work supported with resources and the use of facilities at the Boise Veterans Affairs Medical Center, and it is partially funded by an unrestricted educational grant from Wyeth Pharmaceuticals.