Antimicrobial Resistance Trends and Outbreak Frequency in United States Hospitals

We assessed resistance rates and trends for important antimicrobial-resistant pathogens (oxacillin-resistant Staphylococcus aureus [ORSA], vancomycin-resistant Enterococcus species [VRE], ceftazidime-resistant Klebsiella species [K-ESBL], and ciproﬂoxacin-resistant Escherichia coli [QREC]), the frequency of outbreaks of infection with these resistant pathogens, and the measures taken to control resistance in a stratiﬁed national sample of 670 hospitals. Four hundred ninety-four (74%) of 670 surveys were returned. Resistance rates were highest for ORSA (36%), followed by VRE (10%), QREC (6%), and K-ESBL (5%). Two-thirds of hospitals reported increasing ORSA rates, whereas only 4% reported decreasing rates, and 24% reported ORSA outbreaks within the previous year. Most hospitals (87%) reported having implemented measures to rapidly detect resistance, but only ∼ 50% reported having provided appropriate resources for antimicrobial resistance prevention (53%) or having implemented antimicrobial use guidelines (60%). The most common resistant pathogen in US hospitals is ORSA, which accounts for many recognized outbreaks and is increasing in frequency in most facilities. Current practices to prevent

lance programs provide an excellent overview of rates of antimicrobial resistance in US hospitals [2][3][4][5][8][9][10][11][12][13], these programs are extremely resource-and time-intensive and have important limitations. For example, most surveillance programs that track nosocomial pathogens are overrepresented by larger teaching hospitals [2,14]. Fewer data exist from representative samples that include smaller and/or nonteaching hospitals. In addition, ongoing surveillance systems track overall rates of antimicrobial resistance; no existing program assesses the frequency of recognized outbreaks due to antimicrobial-resistant organisms in US hospitals. Finally, few data exist that describe the extent to which hospitals adhere to guidelines for the control of antimicrobial-resistant pathogens [15]. Several guidelines have been published [16][17][18], yet antimicrobial resistance rates continue to increase. Most recently, the Centers for Disease Control and Prevention (CDC; Atlanta, GA) unveiled a campaign with recommendations to prevent and control antimicrobial resistance [19].
We performed a large representative survey of US Surrounding area with nonmetropolitan population base 97 (20.2) Veterans Affairs hospital 98 (19.8) hospitals, stratified by the number of beds, geographic region, and teaching status. Specific goals of this study were (1) to estimate antimicrobial resistance rates and outbreak frequency associated with 4 epidemiologically important antimicrobialresistant pathogens, and (2) to investigate the extent to which hospitals adopt measures for prevention and control of antimicrobial resistance.

MATERIALS AND METHODS
Sample population. The American Hospital Association annual survey data set was used to identify the sample. Stratification variables were the number of beds (50-99 beds, 100-199 beds, and у200 beds), teaching status (member or nonmember of the Council of Teaching Hospitals), and geographic region (table 1). All acute care Veterans Affairs Medical Centers (VAMCs) were included, with 4-5 non-VAMCs selected for each VAMC in these strata. Facilities were excluded if they had !50 beds, were not "general medical and surgical," or were not accredited by the Joint Commission on Accreditation of Healthcare Organizations. The sample included 670 hospitals. Survey. The survey was mailed to the clinical microbiology laboratory director at each selected hospital. The mailing included the survey, a cover letter, and letters of support for the research (see Acknowledgments). The study procedures, cover letter, and questionnaire were approved by the University of Iowa institutional review board. A reminder postcard was mailed to nonresponders 3-4 weeks after the first survey mailing. A follow-up mailing, which included another survey and cover letter, was sent 4-6 weeks after the postcard mailing. Six weeks after the second survey mailing, nonresponders were contacted by phone and asked to participate. Fourteen weeks after the phone calls were made, a letter, accompanied by a 30min phone card, was sent to respondents to thank them for their participation. Concurrently, a phone card and letter were sent to nonresponders, which served as a final incentive to participate.
Survey development was informed by 2 pilot studies, the second of which included 108 hospitals nationally, to refine the questions and establish methods for reliably estimating resistance rates on an interval scale. The survey instrument addressed (1) prevalence (among all unique clinical isolates from both nosocomial and community onset infections, with duplicate patient isolates excluded) of oxacillin resistance among S. aureus (ORSA), vancomycin resistance among enterococci (VRE), ceftazidime resistance among Klebsiella species (K-ESBL), and quinolone (ciprofloxacin) resistance among Escherichia coli (QREC); (2) three-year trends in these resistances; (3) frequency of recognized outbreaks of these resistant pathogens; (4) availability, frequency, and distribution of reports on the occurrence of antimicrobial resistance (i.e., antibiograms); (5) representation of the microbiology laboratory on the hospital infection-control committee; (6) promptness of notification of pertinent personnel when important resistances are detected; (7) time from bacterial isolation to susceptibility test reporting; and (8) extent of the hospitals' implementation of measures to provide feedback on the occurrence of antimicrobial resistance, implement guidelines for antimicrobial use, recognize and promptly report trends in resistance, rapidly detect resistant pathogens in patient specimens, and provide appropriate resources to prevent resistance. Extent of hospital implementation and support for measures to control resistance was reported using a 5-point Likert scale ("not at all," "very little," "some," "great," and "very great"). All participants were asked to include their most recent antibiogram to allow for validation of reported resistance rates.
Analyses. Point estimates for the overall resistance rates were calculated by the ridit method, using the range category reported by each facility to calculate an overall mean resistance rate for each organism [20]. Resistance rates were compared across geographic region, teaching status, and VA status using the Cochran-Mantel-Haenszel mean score statistic. Resistance rates were compared across bed number using the Jonckheere-Terpstra test [21], which measured the null hypothesis of no association between bed number and resistance rate. The significance level was set at .05, and all P values were 2-tailed. All statistical analyses were performed with SAS (SAS Institute) and StatXact (Cytel Software) software.

RESULTS
A total of 494 hospital laboratories (74%) responded. The characteristics of participating hospitals (table 1) were representative of the sampling frame. Almost two-thirds (64%) of hospitals included a copy of their antibiogram.  Figure 1 shows the proportion of hospitals reporting various levels of epidemiologically important antimicrobial resistances among clinical isolates. Levels were highest for ORSA and lowest for K-ESBL. Overall point estimates of resistance rates were as follows: 36% of S. aureus were resistant to oxacillin, 10% of enterococci were resistant to vancomycin, 6% of E. coli were resistant to ciprofloxacin, and 5% of Klebsiella species were resistant to ceftazidime. Table 2 compares point estimates of resistance rates that we observed for the 4 marker antimicrobial resistant pathogens with antimicrobial resistance rates from several microbiological surveillance programs. The 2 programs with results that most closely approximated the data obtained in this survey, SENTRY and The Surveillance Network (which reported resistance rates for all [i.e., community and nosocomial] recently recovered clinical isolates), demonstrated resistance rates associated with the 4 antimicrobials that were remarkably close to our point estimates. Figure 2 summarizes resistance rates according to the stratification variables bed number, teaching status, and geographic region. Each resistance rate increased significantly as the number of beds increased. In addition, teaching hospitals reported significantly higher resistance rates than did nonteaching hospitals. The major regional differences in resistance included higher ORSA and QREC rates in the South, a higher rate of K-ESBL in the Northeast, and lower resistance rates (particularly for ORSA) in the Mountain/Pacific region. Table 3 summarizes the 3-year trends for each of the 4 antimicrobial resistances reported by participating hospitals and the percentage of participating centers that had reported an outbreak of one of the surveyed antimicrobial-resistant pathogens during the previous 5 years. ORSA was the pathogen that was most commonly associated with an increase in incidence, whereas the incidence of K-ESBL increased in the fewest hospitals. Less than 10% of hospitals reported a decreasing rate of resistance to any of the surveyed antimicrobials. ORSA was the most common antimicrobial-resistant pathogen to have caused recognized outbreaks, which had occurred in almost one-half of the participating centers during the previous 5 years. Outbreaks of VRE infection were also frequent, occurring in nearly one-third of the hospitals. Less than 10% of the centers had reported an outbreak of K-ESBL or QREC infection during the previous 5 years.
Hospital clinical microbiology laboratories reported varying levels of support for antimicrobial resistance prevention and control efforts (table 4). More than 80% of laboratory directors reported that they regularly participate in infection-control committee meetings; 190% reported that they had developed and disseminated an antibiogram. More than 80% of the centers compiling antibiograms reported that they updated them at least yearly.
Most hospital laboratories (78%) reported susceptibility test results within 24 h of bacterial isolation. More than 90% of laboratory directors reported notification of infection-control personnel immediately (by phone or page) when an epidemiologically important resistant pathogen (i.e., ORSA or VRE) was detected. Two-thirds reported prompt notification of both the attending physician and the nursing unit. The level of implementation of recommended measures to control resistance was highest for rapid detection of resistant organisms and was lower for implementing guidelines for antimicrobial use, providing appropriate resources to prevent antimicrobial resistance, providing feedback on the occurrence of antimicrobial resistance, and reporting significant antimicrobial resistance trends (table 4). Antibiograms were submitted by 317 hospitals. The 4 monitored resistance rates obtained from the antibiograms were within 0-3 percentage points of reported rates for all combinations of organisms and antimicrobials (rates for antibiograms vs. survey reports, respectively, were as follows: ORSA, 39% vs. 36%; VRE, 11% vs. 10%; K-ESBL, 6% vs. 5%; and QREC, 6% vs. 6%).

DISCUSSION
The emerging threat of antimicrobial resistance has been well documented [2][3][4][5][6][7][8][9][10][11][12][13]. Using a large, stratified and representative sample, we established that very few US hospitals have been successful in halting or reversing the emergence of antimicrobial resistance-especially among S. aureus, one of the most common and devastating human pathogens [22]. Moreover, resistant pathogens are frequently implicated in hospital outbreaks. The financial and human [22][23][24][25] costs associated with the unchecked spread of antimicrobial-resistant pathogens demand a more aggressive approach to prevention and control.
Multiple recommendations have been developed to help prevent and control antimicrobial resistance [15][16][17][18][19]26]. Any prevention and control recommendations must be based on reliable and valid surveillance that documents the extent and distribution of the problem of antimicrobial resistance. Surveillance will also prove important for monitoring the success of prevention and control efforts.
Most large-scale antimicrobial resistance surveillance pro-grams are labor intensive, and larger tertiary care hospitals are overrepresented [2,14]. Although these programs provide highquality data, and some even provide central reference laboratory confirmation of antimicrobial susceptibility testing results [2], there is interest in using alternative surveillance methods that use electronic data transfer from laboratories [8] or hospitalderived antibiograms [27,28] to estimate antimicrobial resistance rates. Using a survey of a large, stratified, random sample of US hospitals, we found that rates of important resistances closely approximated those reported from more-labor-intensive surveillance programs. The comparability of our data with data generated from existing surveillance programs argues for the establishment of a nationwide reporting system for antimicrobial resistance and outbreak frequency. In addition, our data indicate a very high level of antibiogram preparation and relatively frequent updating of antibiograms. If confidentiality is assured and participation is relatively simple, such a system will be invaluable for tracking antimicrobial resistance trends and outbreak frequency in real time on national, regional, state, and local levels. This could be part of the proposed National Healthcare Safety Network, an internet-based reporting and information system in development by the CDC in collaboration with the Agency for Healthcare Research and Quality, the Centers for Medicare and Medicaid Services, and the US Food and Drug Administration (http://www.cdc.gov/ncidod/hip).
Our survey also provides important data regarding temporal trends in antimicrobial resistance. As demonstrated by CDC investigators, using aggregate multicenter surveillance data to assess trends in antimicrobial resistance does not take into account outlier data and tends to overestimate increases in the prevalence of antimicrobial resistance [29]. Using conservative statistical methods that took into account changes within in- , for difference in each P ! .0001 resistance rate by hospital bed number); teaching status (B; , for difference in each resistance rate by teaching status); and geographic region P ! .01 (C; , dividual hospitals, Fridkin et al. [29] found that significant increases in antimicrobial resistance in a sample of 23 hospitals were only noted for oxacillin resistance among S. aureus and for quinolone resistance among Pseudomonas aeruginosa and E. coli. In our survey, we asked each hospital for their antimicrobial resistance trends for epidemiologically important organisms. Our data suggest that previously described increasing trends in the frequency of ORSA infection [2,3] are not because of data primarily from outlier hospitals. Rather, more than twothirds of this large stratified sample of hospitals report that the rate of oxacillin resistance among S. aureus is increasing, whereas !5% of hospitals report a decreasing trend. Although fewer hospitals (26%-48%) reported increasing trends in the other 3 resistances, it is discouraging to note that !10% of hospitals reported a decreasing trend for any of the resistances we examined. It seems clear that current practices to prevent and control antimicrobial resistance in hospitals are not working.
Our survey represents a rare attempt to determine the frequency with which outbreaks of antimicrobial resistant pathogens occur in US hospitals [30,31]. Although relatively few nosocomial infections are thought to be outbreak related [30,31], when outbreaks do occur, they can result in substantial morbidity, mortality, and resource utilization. In addition, outbreaks are, by definition, "special cause" events and should be preventable. An outbreak, therefore, almost always reflects poorly on infection control and prevention practice in a hospital and can lead to a great deal of anxiety, turmoil, and unfavorable media attention [32]. Almost one-half of the hospitals surveyed had recognized an outbreak of ORSA infection during the 5 years before our survey, and almost 25% had recognized such an outbreak during the previous year. Although the frequency of outbreaks was lower for the other resistant pathogens, they nonetheless occurred commonly-1 in 4 hospitals reported an outbreak of VRE infection during the previous 5 years. Because outbreaks may not be recognized and often resolve without specific intervention [22], these data almost certainly underestimate the problem.
Given the data we present with regard to rates of antimicrobial resistance and outbreak frequency, it is important to determine the extent to which hospitals have implemented guidelines recommended for antimicrobial resistance preven-  tion and control. Previous work suggests that adoption of recommended antimicrobial use guidelines is inadequate [15]. However, our study represents the largest sample of hospitals surveyed to assess measures implemented to prevent and control antimicrobial resistance. These data suggest that most hospitals have implemented measures to detect and report antimicrobial resistance (on-site testing, rapid detection, and antibiogram with regular updating, etc.). However, fewer facilities report vigorous implementation of guidelines for antimicrobial use or provision of adequate resources for prevention and control efforts. More research is needed to assess in greater detail how current guidelines for prevention and control of antimicrobial resistance are implemented, monitored, and enforced.
There are limitations to relying on self-report for surveillance of antimicrobial resistance. First, although the data may be accurate for common, clinically important resistances that interest laboratory directors and infection-control practitioners, the data may not be as accurate for early detection of emergingresistance phenotypes. Second, self-reported data not accom-panied by central reference laboratory testing of the organisms does not allow for independent confirmation of the resistance phenotype or additional testing (e.g., molecular typing to examine genetic relatedness of strains and evidence of patientto-patient transmission). Third, self-reported data generally include all organisms tested in a laboratory (antibiogram data), rather than separating the data by nosocomial versus community acquisition. However, data from the CDC's Intensive Care Antimicrobial Resistance Epidemiology program suggest that, with the exception of ORSA, hospital antibiogram data closely approximate resistance rates among nosocomial pathogens [27].
The data we report have important implications. A more aggressive approach to the prevention and control of antimicrobial resistance is needed. Improved antibiotic utilization to limit selective pressure [33] and improved adherence to infection-control practices (especially hand hygiene [34,35]) should form the bedrock of this approach. However, large multicenter trials to examine innovative approaches to control antimicrobial resistance (e.g., earlier detection, isolation, and/or decolonization of infected or colonized persons) are indicated.