Antimicrobial Nonsusceptibility of Gram-Negative Bloodstream Isolates, Veterans Health Administration System, United States, 2003–2013

Bacteremia caused by gram-negative bacteria is associated with serious illness and death, and emergence of antimicrobial drug resistance in these bacteria is a major concern. Using national microbiology and patient data for 2003–2013 from the US Veterans Health Administration, we characterized nonsusceptibility trends of community-acquired, community-onset; healthcare-associated, community-onset; and hospital-onset bacteremia for selected gram-negative bacteria (Escherichia coli, Klebsiella spp., Pseudomonas aeruginosa, and Acinetobacter spp.). For 47,746 episodes of bacteremia, the incidence rate was 6.37 episodes/10,000 person-years for community-onset bacteremia and 4.53 episodes/10,000 patient-days for hospital-onset bacteremia. For Klebsiella spp., P. aeruginosa, and Acinetobacter spp., we observed a decreasing proportion of nonsusceptibility across nearly all antimicrobial drug classes for patients with healthcare exposure; trends for community-acquired, community-onset isolates were stable or increasing. The role of infection control and antimicrobial stewardship efforts in inpatient settings in the decrease in drug resistance rates for hospital-onset isolates needs to be determined.

D espite advances in public health and medical care, bacteremia is still a major cause of illness and death (1)(2)(3). Bacteremia caused by gram-negative bacteria is a frequent cause of severe sepsis and septic shock (4,5) and poses serious therapeutic challenges. Treatment options are limited because of increased infections with multidrug-resistant, gram-negative bacteria in community and hospital settings. Understanding the epidemiology of infections with gram-negative bacteria is needed to improve guidelines providing empiric therapy recommendations and adequately allocate resources toward infection control and antimicrobial stewardship programs.
Descriptions of the epidemiology of gram-negative bacteremia have been limited by care settings, referral bias, small geographic regions, short study duration, or case identification by use of administrative code data (6)(7)(8)(9)(10). As modern healthcare systems become increasingly diverse and complex, large-scale studies are needed that accurately estimate the burden of antimicrobial resistance and chronologic trends of gram-negative bacteremia (11)(12)(13).
The Veterans Health Administration (VHA) is the largest healthcare system in the United States and has ≈8 million veterans enrolled (14). The VHA uses an integrated electronic health record and a nationwide data repository. Microbiological results from all VHA facilities have been added to the data repository, which enables identification of all cases of gram-negative bacteremia across the entire VHA system. These results enable examination of national trends in gram-negative bacteremia in a geographically dispersed and diverse population.
We previously reported incidence rates of communityonset and hospital-onset bacteremia caused by 3 species of gram-negative bacteria in a cohort of patients admitted to the VHA system during 2003-2013 (15). In this article, we expand on that study by reporting trends in antimicrobial nonsusceptibility (intermediate or resistant) among bacteremia isolates of 4 selected gram-negative bacteria (Escherichia coli, Klebsiella spp., Pseudomonas aeruginosa, and Acinetobacter spp.) in the national VHA healthcare system.

Study Population
We analyzed a retrospective cohort of all veterans who were admitted to acute-care units at VHA hospitals during January 2003-December 2013 and who had positive blood cultures for E. coli, Klebsiella spp., P. aeruginosa, or Acinetobacter spp. between 48 hours before admission and time of discharge. We used these 4 bacterial species a priori because studies consistently showed that these organisms represent most gram-negative bacteria causing bacteremia across various healthcare settings (3,(16)(17)(18)(19)(20)(21)(22). If 1 patient had multiple positive blood cultures for the same species during 1 hospital admission, we included only the first isolate. To estimate the population at risk for community-onset bacteremia, we identified the number of veterans who had visited VHA clinics and emergency departments inside the catchment areas of acute inpatient care facilities by calendar year (outpatient denominator for community-onset bacteremia). To estimate the population at risk for hospital-onset bacteremia, we calculated patientdays in acute inpatient care units (inpatient denominator for hospital-onset bacteremia).
Data for 130 VHA acute-care hospitals in the 48 contiguous states, the District of Columbia, and Puerto Rico were used in this study (23). Using the US Department of Agriculture and Department of Health and Human Services Rural-Urban Commuting Areas System, we found that 107 facilities were located in urban areas, 22 in rural areas, and 1 in a highly rural area (24). These VA hospitals also serve as referral centers for >1,400 outreach clinics. Acute inpatient capacities were 10-260 beds. Total acute inpatient capacity was ≈10,000 acute-care beds, including 1,900 authorized intensive care unit beds (25). We did not include stays at mental health, rehabilitation, and nursing home care units. All but 3 included hospitals have on-site microbiology laboratories, and they are required to conduct quality control/quality assessment routinely per requirements of VHA-designated accreditation organizations and to use methods and equipment approved by the Food and Drug Administration (FDA; Silver Spring, MD, USA).
The institutional review board at the University of Iowa and the research and development committee of the Iowa City Veterans Affairs Health Care System approved this study. A waiver of informed consent was issued for this retrospective analysis.

Data Source
We obtained data through Veterans Affairs Informatics and Computing Infrastructure, which includes data extracted from the VHA integrated electronic medical record system. Susceptibility results in microbiology reports (susceptible/ intermediate/resistant) were recorded in a standardized manner, and isolates were classified as nonsusceptible if they were reported as intermediate resistance or resistant. MICs or sizes of inhibition zones were not typically available.

Definitions
Bacteremia episodes were classified according to Centers for Diseases Control and Prevention (Atlanta, GA, USA) criteria as community-onset and hospital-onset. This classification was based on time of the first positive blood culture as a standard definition (26).
An episode of bacteremia was considered to be community-onset when the first positive blood culture was collected between 48 hours before and <48 hours after admission. For episodes of hospital-onset bacteremia, the first positive blood culture was collected >48 hours after admission.
We further categorized community-onset bacteremia episodes as community-acquired, community-onset and healthcare-associated, community-onset on the basis of healthcare exposure in the VHA before hospital admission. Episodes were classified as healthcare-associated, community-onset when the patient had been admitted to an acutecare facility <90 days before onset of bacteremia; was a resident of a nursing home or rehabilitation facility; was receiving renal replacement therapy; or received wound care or specialized nursing care in an outpatient setting or at home in the 30 days before onset of bacteremia (19). If the patient with community-onset bacteremia did not meet these criteria, the episode was classified as community-acquired, community-onset. Information for healthcare exposure outside the VHA before admission was available only if the VHA was the payer of care.
We categorized antimicrobial drugs listed on susceptibility reports into antimicrobial classes by using interim standard definitions for acquired resistance produced by an international panel of experts from the Centers for Disease Control and Prevention and the European Centre for Disease Prevention and Control (Solna, Sweden) (27). We provide included antimicrobial agents and antimicrobial classes for each organism (online Technical Appendix Tables 1-3, https://wwwnc.cdc.gov/EID/ article/23/11/16-1214-Techapp1.pdf). Isolates were considered nonsusceptible if they were not susceptible to >1 agent in an antimicrobial class (27).

Measurements
We focused primarily on annual proportions and trends of antimicrobial nonsusceptibility rates for the 4 most clinically relevant antimicrobial classes (carbapenems, extended-spectrum cephalosporins, aminoglycosides, and fluoroquinolones) used in gram-negative bacteremia management. We also evaluated incidence rates per 10,000 outpatients for community-onset bacteremia and incidence rates per 10,000 patient-days and per 1,000 admissions for hospital-onset bacteremia for each organism. To enable comparisons with previously reported studies, we calculated age-standardized incidence rates by using a direct method, used US population data for 2000 as a standard population for community-onset bacteremia (28), and reported incidence rates with 2 denominators for hospital-onset bacteremia.

Statistical Analysis
We summarized characteristics of the study population and incidence rates for all 3 categories by using descriptive statistics. We calculated 95% CIs by using the Clopper-Pearson exact method (29) for crude incidence rates. Proportions among acquisition categories were compared by using Fisher exact tests. To assess annual trends in proportions of nonsusceptible isolates, we used the Cochrane-Armitage χ 2 test for trend. All p values were 2-sided, and values <0.05 were considered statistically significant. All statistical analyses were performed by using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Study Population Demographics and Overall Incidence Rates
During 2003-2013, the VHA provided 53,576,096 personyears of outpatient care and 30,015,733 patient-days of acute inpatient care. We obtained demographic characteristics of the study population (Table 1). A total of 47,746 episodes of gram-negative bacteremia occurred during the study. Most (96.4%) patients were male, and communityacquired, community-onset cases accounted for 42.3% of cases of gram-negative bacteremia, followed by healthcareassociated, community-onset (29.2%) and hospital-onset (28.5%) ( Table 2; online Technical Appendix Figure 1).

Trends for Antimicrobial Drug Susceptibilities
Susceptibility data were available for >96% of all organism-antimicrobial drug combinations. The only exception was carbapenem susceptibility of Acinetobacter spp., which was available for 90% of isolates tested.
Incidence rates for fluoroquinolone-nonsusceptible isolates rapidly increased during the first half of the study Carbapenem nonsusceptibility was reported for 3.0% of Klebsiella spp. isolates ( Figure 2). We observed a trend toward increasing nonsusceptibility rates (2.1% in 2003-2007 and 3.7% in 2008-2013, p<0.01 by test for trend).

Discussion
We analyzed 47,746 cases of gram-negative bacteremia in the VHA system over an 11-year period and report 3 major findings. First, for E. coli, rates of nonsusceptibility to aminoglycosides, fluoroquinolones, and extended-spectrum cephalosporins increased across all care settings. Second, for Klebsiella spp., rates of carbapenem nonsusceptibility 1820 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 23, No. 11, November 2017 increased across all care settings, and nonsusceptibility to other antimicrobial classes increased for community-onset isolates and was stable or decreased for isolates from patients with healthcare exposure. Third, for P. aeruginosa and Acinetobacter spp., rates of nonsusceptibility to aminoglycosides, fluoroquinolones, and extended-spectrum cephalosporins mostly decreased, but only for isolates from patients with healthcare exposure. Overall incidence rates for gram-negative bacteremia were comparable with those from previous populationbased studies from the United States and Europe (3,13,(20)(21)(22)30). We recently reported major changes in hospital-onset bacteremia, which was possibly caused by expansion of horizontal infection control programs in the VHA system during this time (15). An additional finding in our more recent study was the decreased proportion of nonsusceptible bloodstream isolates across nearly all antimicrobial classes for Klebsiella spp., P. aeruginosa, and Acinetobacter spp. in patients with healthcare exposure. However, trends for community-onset bloodstream isolates were stable or showed increased nonsusceptibility.
Our findings suggest that efforts to improve infection control practices within the VHA might have had effects on antimicrobial susceptibilities for isolates from patients with healthcare exposure, although the extent of these effects needs to be determined. With increased nonsusceptibility rates among community-acquired, community-onset isolates, we generally expect upward trends of nonsusceptibility rates for healthcare-associated and hospital-onset isolates caused by increased colonization pressure (31,32). However, our results showed improvements in nonsusceptibility rates for isolates of Klebsiella spp., P. aeruginosa, and Acinetobacter spp. in hospital-onset and healthcare-associated, community-onset infections, which might suggest successful interruption of transmission in inpatient settings. Infection control programs within the VHA have been expanded over the past 10 years, especially since introduction of the Methicillin-Resistant Staphylococcus aureus Prevention Initiative in 2007 (33). Our recent study showed a substantial decrease in nosocomial, gram-negative bacteremia after implementation of this initiative, which suggests a collateral benefit beyond the originally intended scope of this initiative (15). In addition, the VHA released a mandatory policy for hand hygiene practices in 2011 (34).
Antimicrobial stewardship efforts within the VHA, which have been evolving over the past 10 years, might have contributed to increased antimicrobial susceptibilities. The National VA Antimicrobial Stewardship Task Force, established in 2011, has been leading systemwide educational efforts, and a mandate for inpatient stewardship was enacted in 2014. However, even before these broad interventions, facilities were actively performing stewardship. According to a survey of VHA facilities conducted by the VA Healthcare Analysis and Information Group, 86 (66.2%) of 130 VHA facilities had either a formal or informal antimicrobial stewardship policy in place in 2012 (35). At the time of the survey, 14 (16.3%) of these facilities had a stewardship policy in place for 2 years, 7 (8.1%) for 3 years, and 33 (38.4%) for >4 years.
We speculate that increased rates of antimicrobial susceptibility were not observed for E. coli healthcareassociated, community-onset and hospital-onset isolates because patients were probably colonized with E. coli before admission and these commensal strains might have developed or acquired resistance in the community setting. Antimicrobial drug resistance can develop in E. coli by its exposure to retail meat (36) and antimicrobial drug exposure outside VHA hospitals. Antimicrobial stewardship efforts within the VHA, like most healthcare systems, have typically not focused on outpatients. In addition, many non-VHA hospitals, which might have cared for patients in this cohort but were not captured by our analysis, did not have stewardship programs during this study (37,38).
A notable exception to these trends is increased rates of carbapenem nonsusceptibility for Klebsiella spp. in community and hospital settings. The increase in carbapenem-resistant Klebsiella spp. was also reported outside the VHA (39)(40)(41). This observation deserves attention because it might indicate a lapse of infection control efforts at the VHA. The VHA implemented internal guidelines for carbapenem-resistant Enterobacteriaceae to address this issue in 2015, and additional research is needed to determine whether this intervention has had any positive effect on this major threat (42).
The strength of our study was inclusion of all bacteremia episodes from the entire VHA system, which enabled us to perform population-based analysis, including diverse care settings, for wide geographic regions. Our study demonstrated the advantage of an integrated health informatics infrastructure with a clinical data warehouse within the VHA system. To accurately monitor the burden and trends  of antimicrobial resistance, more investments in data integration informatics infrastructure need to be made. This study had several limitations. First, most patients seen within the VHA system were male. Therefore, results might not be generalizable to the rest of the population. Second, because information was collected through patient medical records, detailed microbiological information, such as break points of antimicrobial susceptibility testing, molecular typing, or mechanisms of resistance, was often not available. Third, the Clinical and Laboratory Standards Institute (CLSI) revised its break point recommendations for extended-spectrum cephalosporins and carbapenems in 2010 (43), and antimicrobial susceptibility trends in the later part of the study might have been affected by this change. However, the FDA did not revise its recommendations to match those of the CLSI until 2013 (44). Because most microbiology laboratories in the VHA system use commercial systems that are regulated by the FDA for susceptibility testing, it is less likely that the change of CLSI break points substantially affected our observations. Fourth, this study focused on nationwide overall trends, and results might reflect local trends in each specific region. Previous studies showed geographic variabilities in incidence rates and antimicrobial susceptibilities for gram-negative bacteria (45)(46)(47). We also recently reported regional variations in fluoroquinolone nonsusceptibilities for E. coli isolates and plan to analyze geographic variabilities for other organisms and antimicrobial agents in future studies (48). Fifth, resistance can develop in gram-negative bacteria during patient therapy, and assessing susceptibilities for the primary isolate from each episode might underestimate the prevalence of nonsusceptibilities. Because there were diversities in reporting practices of susceptibilities for subsequent isolates across the system, it was not feasible to collect information for those isolates in a standardized manner.
In conclusion, we observed an increase in antimicrobial drug nonsusceptibility for E. coli in all healthcare exposure categories and stable or decreased rates of nonsusceptibility for Klebsiella spp., P. aeruginosa, and Acinetobacter spp. in patients with healthcare exposure. The extent to which these findings might reflect VHA infection control and antimicrobial stewardship activities remains to be determined. Despite encouraging decreases in nonsusceptibility for pathogenic gram-negative bacteria in acute-care hospital settings, increased attention to infection prevention and antimicrobial stewardship is needed for outpatient and nonacute inpatient settings.

The Public Health Image Library (PHIL)
The Public Health Image Library (PHIL), Centers for Disease Control and Prevention, contains thousands of public health-related images, including highresolution (print quality) photographs, illustrations, and videos.
PHIL collections illustrate current events and articles, supply visual content for health promotion brochures, document the effects of disease, and enhance instructional media. PHIL images, accessible to PC and Macintosh users, are in the public domain and available without charge.