Abstract
Background
Extremely drug-resistant gram-negative bacilli (XDR-GNB) increasingly cause healthcare-associated infections (HAIs) in intensive care units (ICUs).
Methods
A matched case-control (1:2) study was conducted from February 2007 to January 2010 in 16 ICUs. Case and control subjects had HAIs caused by GNB susceptible to ≤1 antibiotic versus ≥2 antibiotics, respectively. Logistic and Cox proportional hazards regression assessed risk factors for HAIs and predictors of mortality, respectively.
Results
Overall, 103 case and 195 control subjects were enrolled. An immunocompromised state (OR=1.55, p=0.047) and exposure to amikacin (OR=13.81, p<0.001), levofloxacin (OR=2.05, p=0.005), or trimethoprim-sulfamethoxazole (OR=3.42, p=0.009) were factors associated with XDR-GNB HAIs. Multiple factors in both case and control subjects significantly predicted increased mortality at different time intervals after HAI diagnosis. At 7 days, liver disease (Hazard Ratio [HZ]=5.52), immunocompromised state (HR=3.41), and bloodstream infection (HR=2.55) predicted mortality; at 15 days, age (HR=1.02 per year increase), liver disease (HR=3.34), and immunocompromised state (HR 2.03) predicted mortality; and at 30 days, age (HR=1.02 per one year increase), liver disease (HR=3.34), immunocompromised state (HR=2.03), and hospitalization in a medical ICU (HR=1.85) predicted mortality.
Conclusions
HAIs caused by XDR-GNB were associated with potentially modifiable factors. Age, liver disease, and immunocompromised state, but not XDR-GNB HAIs, were associated with mortality.
Introduction
Antibiotic-resistant gram-negative bacilli (GNB) are increasingly common causes of healthcare-associated infections (HAIs) in intensive care units (ICUs) [1] and are associated with higher mortality rates, longer hospitalizations, and increased healthcare expenditures [2, 3]. Effective treatment for extremely drug-resistant (XDR) GNB infections is challenging due to limited therapeutic options [4].
In this study, we examined the epidemiology and outcomes of HAIs caused by XDR-GNB in the 16 ICUs affiliated with our medical center. We performed a case-control study to identify risk factors associated with XDR-GNB infections compared with non-XDR-GNB infections. We hypothesized that exposure to carbapenem agents would be associated with HAIs caused by XDR-GNB. In addition, we performed a survival analysis to explore if predictors for death changed 7, 15, and 30 days after diagnosis of an HAI. We hypothesized that HAIs caused by XDR-GNB would be associated with an increased hazard for mortality and that the effect would be most pronounced at 7 days, rather than at 15 or 30 days.
Materials and Methods
Study Design and Study Setting
This study was a prospective cohort study with a nested, matched case-control study. It was conducted from February 2007 to January 2010 in the 16 ICUs affiliated with NewYork-Presbyterian (NYP) Hospital located in New York City. NYP is a 2,278-bed (383 ICU-bed) tertiary-care facility affiliated with two medical schools, Columbia University College of Physicians and Surgeons and Weill Cornell Medical College. Study ICUs included medical (n=5), surgical (n=6), burn (n=1), and pediatric/neonatal (n=4) ICUs and had approximately 14,800 annual patient admissions. Institutional Review Board approval was obtained from Columbia University Medical Center and Weill Cornell Medical College with a waiver of informed consent.
Study Subjects and Case Definitions
The cohort was defined as all patients admitted to the study ICUs during the study period. Case subjects were defined as patients hospitalized in the ICU with healthcare-associated bloodstream infections (BSIs), pneumonia (PNA), or urinary tract infections (UTIs) caused by XDR-Acinetobacter spp., Klebsiella pneumoniae, or Pseudomonas aeruginosa (defined below). Control subjects were defined as patients hospitalized in the ICU with HAIs caused by non-XDR Acinetobacter spp., K. pneumoniae, or P. aeruginosa. HAIs were diagnosed using the Centers for Disease Control & Prevention's National Hospital Safety Network (NHSN) definitions [5], but modified to include antimicrobial treatment. When feasible, case and control subjects were matched (1:2) by the following matching hierarchy: campus (Columbia or Cornell), type of ICU (medical or surgical), type of infection (BSI, PNA, or UTI), date of culture, and pathogen (Acinetobacter spp., K. pneumoniae, or P. aeruginosa). Patients were excluded if their infections developed < 48 hours after hospital admission, were a non-study type of infection, e.g., skin and soft tissue infection, or were caused by a non-study pathogen.
XDR-GNB were the species described above, susceptible to ≤1 antimicrobial agent or only susceptible to imipenem and meropenem as determined by commercial broth microdilution susceptibility panels (described below). Non-XDR-GNB were susceptible to ≥2 antimicrobial agents. Susceptibility to tigecycline and polymyxin B were not included in the definitions of XDR- and non-XDR-GNB, as these agents were not consistently tested at the study sites. MICs were interpreted according to the Clinical and Laboratory Standards Institute breakpoints in effect during the study period [6-8].
Potential subjects were identified prospectively using EpiPortal, a web-based surveillance system developed by the NYP Department of Infection Prevention & Control and Department of Information Technology and Columbia University Department of Biomedical Informatics [9]. EpiPortal integrates data from different electronic systems (e.g., microbiology laboratories, pharmacy, medical records) to identify patients with epidemiologically significant organisms including multidrug-resistant pathogens. The electronic medical record of each potential subject was reviewed by a study physician to confirm case or control status and to determine the presence of comorbid conditions, antibiotic exposures, and medical device use (central venous catheter, mechanical ventilation, and/or urinary catheter). Demographic and microbiological data were also obtained from the electronic medical record.
At the Columbia campus, blood culture samples from adults were inoculated into BD Bactec Plus Aerobic/F and Bactec Lytic/10 Anaerobic/F bottles, while pediatric samples were inoculated into Bactec Peds Plus/F bottles (Becton Dickinson, Franklin Lakes, NJ). At the Cornell campus blood culture samples obtained from adults and children were inoculated into BactT Alert bottles (bioMérieux, Durham, NC). Respiratory and urine samples were plated onto MacConkey agar at both study sites. During the study period, the clinical microbiology laboratories used the Vitek 2 system (bioMérieux, Durham, NC) as the primary method of antibiotic susceptibility testing (AST). The laboratory on the Columbia campus used the Vitek 2 AST GN09 prior to May 2009 and afterwards used GN35. The laboratory on the Cornell campus used Vitek 2 AST GN13 prior to January 2009 and afterwards used GN28 for Klebsiella and Acinetobacter spp. and GN31 for Pseudomonas aeruginosa. Both laboratories performed Etests (bioMérieux, Durham, NC) to determine susceptibility to polymyxin B and tigecycline for XDR strains if requested, and at Cornell, Etests for tigecycline were regularly performed after January 2009.
Risk Factors for HAIs and Predictors of Mortality
Risk factors evaluated for HAIs caused by XDR-GNB vs. non-XDR-GNB included age, sex, race and ethnicity; days of ICU and hospital stay prior to infection; comorbid conditions (defined below); exposure to antibiotics administered during hospitalization in the 30 days prior to infection; and use of medical devices in the 7 days prior to infection. Comorbid conditions were defined using APACHE II/III classification [10]. Briefly, liver disease was defined as biopsy-proven cirrhosis or portal hypertension; respiratory disease was defined as a chronic process resulting in severe exercise restriction; cardiovascular disease was defined as symptoms of cardiac insufficiency at rest; renal impairment was defined as the use of chronic dialysis; and immunocompromised state was defined as conditions that increased susceptibility to infection (e.g., leukemia/lymphoma, metastatic cancer) or receipt of immunosuppressant medications (e.g., chemotherapy, high dose steroids).
Potential predictors of mortality were infection with an XDR-GNB, age, sex, comorbid conditions, type of ICU, duration of ICU stay prior to infection, pathogen, type of infection, and time to effective therapy (defined below).
Outcomes
The onset of HAIs was defined as the first day of positive culture(s). Several outcomes related to antibiotic treatment were compared among case vs. control subjects. These included: (1) duration of therapy (calendar days) with ≥1 antibiotic(s) with GNB activity administered following HAI diagnosis; (2) the number of antibiotics with GNB activity; (3) time to effective therapy with ≥1 antibiotic(s) to which the infecting organism was susceptible in vitro, including tigecycline and polymyxin B; and (4) duration of effective therapy. Effective therapy was considered “not received” if the time to effective therapy was >7 days. In addition, the proportion of case vs. control subjects with persistently positive blood cultures (i.e., positive cultures for >1 calendar day) within 7 days of the first blood culture was determined. During the hospital admission in which the HAI was diagnosed, mortality was determined 7, 15, and 30 days after the HAI was diagnosed.
Statistical Analysis
To assess risk factors for HAIs, conditional logistic regression was used for bivariate analyses. Using a backward elimination approach, multivariable conditional logistic regression was used to examine potential risk factors associated with HAIs caused by XDR-GNB. The final model included age, sex, and length of stay prior to infection, and all risk factors significant at p<0.05.
To assess predictors of mortality, Cox proportional hazards regression was used for bivariate analyses. Using a backward elimination approach, multivariable Cox proportional hazards regression was used to examine potential predictors of risk of mortality at a specific time point (i.e., hazard). Three survival analyses were conducted to explore if predictors changed when the observation time for subjects was censored at 7, 15, and 30 days following HAI diagnosis. The final models included age and case status, and all predictors significant at p<0.05. Statistical analyses were completed in SAS 9.2 for Windows (SAS Institute Inc., Cary, NC).
Results
Subjects
During the study period, 103 case subjects and 195 control subjects were identified; 92 case subjects were matched to 2 control subjects and 11 were matched to 1 control subject. The demographic and clinical characteristics of subjects are shown in Table 1. Six case and 8 control subjects were <18 years old, including one case from the neonatal ICU. Consistent with the matching strategy, comparable proportions of subjects were hospitalized at each campus and type of ICU. Pneumonia was the most common HAI, followed by BSI. While the proportion of case and control subjects with HAIs caused by K. pneumoniae was similar, the proportions of infections caused by Acinetobacter spp. and P. aeruginosa were significantly different among case and control subjects (p<0.001); few HAIs were caused by XDR-P. aeruginosa or by non-XDR-Acinetobacter spp.
Table 1. Characteristics of Case vs. Control Subjects with Healthcare-associated Infections Caused by Gram-Negative Bacilli.
| Characteristic | Case Subjects (n=103) | Control Subjects (n=195) | p-valuea |
|---|---|---|---|
| n (%) | |||
| Age group | 0.701 | ||
| < 50 yrs | 22 (21.4) | 44 (22.6) | |
| 50–75 yrs | 57 (55.3) | 94 (48.2) | |
| > 75 yrs | 24 (23.3) | 57 (29.2) | |
| Sex | 1.00 | ||
| Male | 59 (57.3) | 111 (56.9) | |
| Female | 44 (42.7) | 84 (43.1) | |
| Race | 0.43 | ||
| White | 34 (33.0) | 47 (24.1) | |
| Black | 8 (7.8) | 21 (10.8) | |
| Asian/Pacific Islander | 2 (1.9) | 4 (2.1) | |
| Other | 15 (14.6) | 30 (15.4) | |
| Unknown | 44 (42.7) | 93 (47.7) | |
| Ethnicity | 0.63 | ||
| Hispanic | 12 (11.7) | 18 (9.2) | |
| Non-Hispanic | 30 (29.1) | 51 (26.2) | |
| Unknown | 61 (59.2) | 126 (64.6) | |
| Campusb | 1.00 | ||
| Columbia University Medical Center | 49 (47.6) | 96 (49.2) | |
| Weill Cornell Medical Center | 54 (52.4) | 99 (50.8) | |
| ICU typeb | 0.82 | ||
| Medical ICU | 53 (51.5) | 103 (52.8) | |
| Surgical and Burn ICU | 50 (48.5) | 92 (47.2) | |
| Type of infectionb | 1.00 | ||
| Bloodstream | 34 (33.0) | 68 (34.9) | |
| Pneumonia | 50 (47.6) | 92 (47.2) | |
| Urinary tract | 19 (19.4) | 35 (17.9) | |
| Pathogenb | <0.001 | ||
| Klebsiella pneumoniae | 48 (46.6) | 100 (51.3) | |
| Acinetobacter spp. | 49 (47.6) | 21 (10.8) | |
| Pseudomonas aeruginosa | 6 (5.8) | 74 (37.9) | |
Bivariate conditional logistic regression
Variable included in matching hierarchy
ICU= Intensive care unit
Antibiotic Susceptibilities of GNB Isolates
The antimicrobial susceptibilities of the GNB isolates from case and control subjects are shown in Table 2. Consistent with the case definitions, a greater proportion of non-XDR-GNB isolates were susceptible to aminoglycoside, fluoroquinolone, and β-lactam agents than XDR-GNB. Susceptibility to these antimicrobial classes varied from 0% to 16% among XDR isolates and from 86% to 99% among non-XDR isolates. Most XDR isolates had tigecycline MICs ≤2 μg/mL (68%, 58/85 tested) and polymyxin B MICs ≤2 μg/mL (90%, 75/83 tested).
Table 2. Comparison of Selected Susceptibility Profiles of Gram-Negative Bacilli Causing Healthcare-associated Infections among Case and Control Subjects.
| Antimicrobial Agent | Case Subjects (n=103) | Control Subjects (n=195) |
|---|---|---|
| Susceptible n/Na (%) | Susceptible n/Na (%) | |
| Amikacin | 15/96 (15) | 193/195 (99) |
| Cefepime | 7/98 (7) | 166/193 (86) |
| Ciprofloxacin | 0/50 (0) | 86/96 (90) |
| Gentamicin | 16/102 (16) | 178/196 (91) |
| Imipenem | 1/67 (1) | 129/142 (91) |
| Levofloxacin | 0/102 (0) | 167/195 (86) |
| Meropenem | 6/82 (7) | 127/140 (91) |
| Piperacillin-tazobactam | 1/51 (2) | 76/87 (87) |
| Polymyxin Bb | 75/83 (90) | 14/14 (100) |
| Tigecyclinec | 58/85 (68) | 5/9 (56) |
| Trimethoprim-sulfamethoxazole | 1/101 (1) | 100/121 (83) |
| Tobramycin | 3/82 (4) | 162/170 (95) |
n= number susceptible isolates and N= number of isolates with available results
Interpretive criteria do not exist for the Enterobacteriacae when testing polymyxin B; therefore, the susceptible breakpoint of ≤2 μg/mL approved for P. aeruginosa and Acinetobacter spp. was employed for these organisms.
Interpretive criteria do not exist for the Acinetobacter spp. when testing tigecycline; therefore, the susceptible breakpoint of ≤2 μg/mL approved for Enterobacteriacae was employed for these organisms.
Risk factors for XDR-GNB HAIs
The proportion of case and control subjects with comorbid conditions and device use is shown in Table 3. Compared to control subjects, case subjects were more likely to have chronic respiratory conditions and to require mechanical ventilation, but did not have a longer hospital or ICU length of stay prior to infection.
Table 3. Risk Factors for Healthcare-associated Infections (HAIs) with Gram-Negative Bacilli: Comorbid Conditions and Medical Device Usea.
| Risk factors | Case Subjects N=103 | Control Subjects N=195 | p-value |
|---|---|---|---|
| Comorbid Conditions | n (%) | ||
| None | 47 (45.6) | 113 (57.9) | 0.12 |
| Liver disease | 5 (4.9) | 9 (4.6) | 0.90 |
| Immunocompromised state | 33 (32.0) | 41 (21.0) | 0.11 |
| Cardiovascular disease | 9 (8.7) | 14 (7.2) | 0.70 |
| Chronic respiratory disease | 12 (11.7) | 7 (3.6) | 0.028 |
| Chronic renal impairment | 15 (14.6) | 15 (7.7) | 0.22 |
| Length of stay prior to HAI | |||
| Hospital, mean ± SD (days) | 28.8 ± 31.2 | 20.3 ± 23.9 | 0.13 |
| ICU, mean ± SD (days) | 11.5 ± 9.5 | 11.2 ± 17.3 | 1.00 |
| Device use within 7 days of HAI | |||
| Mechanical ventilation | 85 (82.5) | 124 (63.6) | 0.011 |
| Urinary catheter | 93 (90.3) | 156 (80.0) | 0.17 |
| Central venous catheter | 81 (78.6) | 143 (73.3) | 0.48 |
Bivariate Analyses
ICU= Intensive care unit
Inpatient antibiotic use during the 30 days prior to infection differed among case and control subjects as shown in Table 4. In the bivariate analyses, case subjects were more likely to have been exposed to several antimicrobial agents including amikacin, a carbapenem agent, linezolid, piperacillin-tazobactam, polymyxin B, tigecycline, trimethoprim-sulfamethoxazole, and vancomycin.
Table 4. Antimicrobial Use Within 30 days Prior to Healthcare-associated Infections.
| Antimicrobial Agent | Case Subjects N=103 | Control Subjects N=195 | OR (95%CI) | p-valuea |
|---|---|---|---|---|
| n (%) | ||||
| Amikacin | 13 (12.6) | 1 (0.5) | 13.07 (2.96, 57.66) | 0.001 |
| Ampicillin | 9 (8.7) | 6 (3.1) | 1.91 (0.84, 4.32) | 0.12 |
| Ampicillin-sulbactam | 9 (8.7) | 12 (6.1) | 1.22 (0.64, 2.34) | 0.54 |
| Aztreonam | 7 (6.8) | 6 (3.1) | 1.59 (0.70, 3.60) | 0.27 |
| Carbapenems | 33 (32.0) | 20 (10.3) | 2.32 (1.44, 3.74) | 0.001 |
| Cefazolin | 9 (8.7) | 24 (12.3) | 0.84 (0.47, 1.47) | 0.54 |
| Cefepime | 20 (19.4) | 19 (9.7) | 1.53 (0.94, 2.48) | 0.09 |
| 3rd generation cephalosporinb | 13 (12.6) | 15 (7.7) | 1.36 (0.76, 2.41) | 0.30 |
| Clindamycin | 3 (2.9) | 6 (3.1) | 1.00 (0.42, 2.40) | 1.00 |
| Daptomycin | 6 (5.8) | 5 (2.6) | 1.63 (0.66, 4.00) | 0.29 |
| Gentamicin | 14 (13.6) | 15 (7.7) | 1.28 (0.75, 2.18) | 0.36 |
| Levofloxacin | 21 (20.4) | 21 (10.8) | 1.51 (0.94, 2.42) | 0.09 |
| Linezolid | 22 (21.4) | 12 (6.2) | 2.37 (1.33, 4.20) | 0.003 |
| Macrolidec | 19 (18.5) | 19 (9.7) | 1.53 (0.87, 2.68) | 0.14 |
| Metronidazole | 25 (24.3) | 38 (19.5) | 1.13 (0.77, 1.64) | 0.54 |
| Oxacillin | 6 (5.8) | 12 (6.2) | 0.85 (0.42, 1.70) | 0.64 |
| Piperacillin-tazobactam | 73 (70.9) | 101 (51.8) | 1.46 (1.04, 2.06) | 0.029 |
| Polymyxin B | 11 (10.7) | 0 (0) | 1.13 (1.28, 3.54) | <0.001 |
| Tigecycline | 10 (9.7) | 2 (1.0) | 5.16 (1.65, 16.13) | 0.005 |
| Tobramycin | 31 (30.1) | 34 (17.4) | 1.57 (0.99, 2.47) | 0.05 |
| Trimethoprim-sulfamethoxazole | 14 (13.6) | 4 (2.1) | 3.62 (1.56, 8.39) | 0.003 |
| Vancomycin | 78 (75.3) | 104 (53.3) | 1.57 (1.12, 2.22) | 0.001 |
Bivariate Analyses
Cefotaxime, ceftriaxone, ceftazidime
Erythromycin or azithromycin
In the final multivariable analyses, four variables were identified as significant risk factors for HAIs caused by XDR-GNB as shown in Table 5. These included immunocompromised state and exposure to the antimicrobial agents amikacin, levofloxacin, or trimethoprim-sulfamethoxazole. Comorbid conditions and device use were not identified as risk factors.
Table 5. Risk Factors Associated with Healthcare-associated Infections Caused by Extremely Drug-resistant Gram-Negative Bacilli, Multivariable analysis.
| Risk Factora | OR (95% CI) | p-value |
|---|---|---|
| Immunocompromised state | 1.55 (1.01, 2.39) | 0.047 |
| Amikacin | 13.81 (2.96, 64.47) | 0.001 |
| Levofloxacin | 2.05 (1.24, 3.40) | 0.005 |
| Trimethoprim-sulfamethoxazole | 3.42 (1.36, 8.60) | 0.009 |
Age, sex, and hospital stay were forced into the final model and were not significant
Antibiotic Treatment and Persistently Positive Blood Cultures
The mean duration of antibiotic therapy was similar among case (15.7 days) and control (13.4 days) subjects (p=0.41). However, more antimicrobial agents with GNB activity were administered to case (mean 3.8 antibiotics) than to control (mean 3.1 antibiotics) subjects (p=0.001). Although the mean duration of effective therapy did not differ between case (11.1 days) and control (9.8 days) subjects (p=0.21), the mean time to effective therapy was longer for case (3.0 days) than control (1.3 days) subjects (p<.001). Furthermore, fewer case (83%) than control (96%) subjects received effective therapy within 7 days of their first positive blood culture (p<0.001). Among those who survived at least one week following their first positive blood culture, 12% (3/25) of case and 16% (7/44) of control subjects had persistently positive blood cultures (p=0.66).
Mortality
More case (59%) than control (31%) subjects died during their hospital stay (p<0.001). Among those who died, the mean survival after HAI was similar among case (22.6 days) and control (27.1 days) subjects (p=0.44). Among cases, 11 deaths occurred within 7 days of infection and 21 deaths occurred >30 days after infection. For those with BSIs, mortality was higher for case (77%, 26/34) than control (31%, 21/68) subjects (p<0.001). Similarly, for those with PNA, mortality was higher for case (58%, 29/50) than control (36%, 33/92) subjects (p=0.010). However, mortality was similar among case (32%, 6/19) and control (20%, 7/35) subjects with UTIs (p=0.53).
The multivariable Cox proportional hazards regression for 7-, 15-, and 30-day mortality is presented in Table 6. Case status was not an independent predictor of mortality at any of these time intervals, but an immunocompromised state or liver disease was an independent predictor. BSI was a significant predictor for 7-day mortality only, while older age was a significant predictor for 15- and 30-day mortality. Type of pathogen and time to effective therapy were not independent predictors of mortality.
Table 6. Predictors of Mortality after Healthcare-associated Infection (HAI) Diagnosis with Censoring at 7-, 15-, and 30-days.
| Predictors of Mortality | Multivariable Cox Proportional Hazards Analysis | |
|---|---|---|
| Hazard Ratio (95% CI) | p-value | |
| Death within 7 days | ||
| Case statusa | 0.94 (0.42, 1.95) | 0.87 |
| Age (per one year increase)a | 1.01 (0.99, 1.04) | 0.22 |
| Liver disease | 5.52 (1.79, 16.99) | 0.003 |
| Immunocompromised state | 3.41 (1.67, 6.96) | 0.001 |
| Bloodstream infectionb | 2.55 (1.32, 4.86) | 0.005 |
| Death within 15 days | ||
| Case statusa | 1.39 (0.84, 2.31) | 0.20 |
| Age (per one-year increase)a | 1.02 (1.00, 1.04) | 0.014 |
| Liver disease | 3.24 (1.30, 8.07) | 0.012 |
| Immunocompromised state | 2.66 (1.57, 4.50) | <.001 |
| Death within 30 days | ||
| Case statusa | 1.39 (0.90, 2.15) | 0.14 |
| Age (per one-year increase)a | 1.02 (1.01, 1.03) | 0.002 |
| Liver disease | 3.34 (1.56, 7.12) | 0.001 |
| Immunocompromised state | 2.03 (1.33, 3.11) | 0.001 |
| Medical intensive care unitc | 1.85 (1.22, 2.86) | 0.005 |
Case status and age were included a priori in the model
Compared to urinary tract infection
Compared to surgical intensive care unit
Discussion
This is one of the largest recent studies to describe the epidemiology of HAIs caused by XDR-GNB among patients hospitalized in ICUs and to assess relevant outcomes including predictors of mortality. To further delineate the impact of HAIs caused by XDR-GNB, we performed a matched case-control study adjusting for previously identified predictors of HAIs caused by resistant pathogens including several comorbid conditions, use of medical devices, and length of stay [11]. We demonstrated that an immunocompromised state or previous treatment with amikacin, levofloxacin, or trimethoprim-sulfamethoxazole within 30 days of infection were risk factors for HAIs caused by XDR-GNB. While in-hospital mortality was higher among case subjects, XDR-GNB HAIs did not predict mortality at 7, 15, or 30 days after HAI diagnosis. However, BSIs caused by either XDR- or non-XDR-GNBs did predict mortality at 7 days.
Contrary to our hypothesis, we did not find that treatment with carbapenem agents was a risk factor for XDR infection. Several previous studies have also assessed antimicrobial exposures as risk factors for infection and/or colonization with XDR GNB, but have not had consistent findings. Henceforth in this discussion, we will use the term multi-drug resistant (MDR) GNB, as it is the term most commonly used by the authors cited, though definitions of XDR and MDR GNB may vary. Use of fluoroquinolone agents has been associated with HAIs caused by K. pneumoniae carbapenemase (KPC)-producing strains or carbapenem-resistant A. baumannii [12, 13]. Exposure to imipenem [14, 15], piperacillin-tazobactam [14], vancomycin [14, 15], or aminoglycoside agents [14, 15] has also been associated with detection of imipenem-resistant P. aeruginosa from clinical cultures. Furthermore, exposure to trimethoprim-sulfamethoxazole has been associated with MDR- Stenotrophomonas maltophilia [16], colonization with trimethoprim-sulfamethoxazole-resistant Enterobacteriaceae [17] and UTIs caused by trimethoprim-sulfamethoxazole-resistant Escherichia coli [18].
While an immunocompromised state has been described as a risk factor for HAIs [19], the association with MDR-GNB infection is less clear. Steroid use during the previous 30 days has been associated with infections caused by extended spectrum β-lactamase (ESBL)-producing E. coli and K. pneumoniae [20]. Similarly receipt of antineoplastic, immunosuppressive, and immunomodulating agents has been associated with acquisition of MDR-GNB [21]. Conversely, others did not find that immune suppression (defined as solid or hematological malignancy, leukopenia, or chronic use of immunosuppressive agents) was associated with MDR-GNB infections [22].
However, studies of risk factors for MDR-GNB infections and/or colonization are difficult to compare. Study designs, study populations, local epidemiology, and definitions of resistance vary widely [23, 24]. Furthermore, risk factors may be affected by the route of acquisition of MDR-organisms (MDROs); MDROs may arise from de novo selection via antibiotic exposure or be transmitted from other patients, healthcare personnel, or the healthcare environment [25]. Additional factors may increase the risk of progression from colonization to infection such as medical devices, breakdown in mucosal barriers, or impaired immune function [11].
While others have reported that HAIs caused by MDR-GNBs were associated with increased mortality [26, 27], XDR case status was not a predictor of mortality in the current study, despite more deaths in case subjects. Notably, many deaths among case subjects occurred >30 days after infection and few deaths occurred within 7 days of infection, as would be expected for HAI-attributable mortality. Our survival analysis implicated two comorbid conditions, liver disease and an immunocompromised state, as independent predictors of mortality at 7, 15 and 30 days after HAIs. We speculate that the severity of illness associated with these comorbid conditions accounted for the increased risk of mortality. Others have also found that higher APACHE II scores and Charlson comorbidity indices have also been associated with an increased mortality risk [28].
We explored several predictors of mortality related to antibiotic treatment. We did not find that a delay in effective therapy impacted mortality—again likely confounded by the effect of comorbidities. Similarly, two previous studies have shown that timely administration of antibiotics was not associated with survival among patients with BSIs caused by carbapenem-resistant K. pneumoniae [2, 29]. In contrast, delays in appropriate therapy have been associated with mortality in patients with MDR E. coli and P. aeruginosa bacteremia [30, 31]. In the current study, while the mean duration of therapy was similar among case and control subjects, case subjects received more unique antibiotics. Thus, treatment of XDR-GNB most likely results in more antibiotic exposures and further antibiotic resistance.
This study had limitations. It was performed at a large, tertiary care hospital system in New York City and findings may not be generalizable to other settings; NYC is known to be an epicenter for XDR-GNB infections in ICUs [32]. Our definition of XDR-GNB was crafted prior to the recent international consensus definition which could further limit the generalizability of our findings [33]. We did not determine clonality and therefore could not distinguish if the infections were endemic or epidemic. The diagnosis of pneumonia, even using NHSN diagnostic criteria, lacks both sensitivity and specificity [34]. We did not assess the potential impact of removal of central venous catheters which may have impacted outcomes. Our matching hierarchy may have led to overmatching and selection bias [35]. The use of control subjects infected with susceptible GNBs may have inflated the odds ratios for antibiotic exposures since patients previously treated with antibiotic agents may be less likely to be infected with a susceptible organism [36]. Lastly, while comorbid conditions were associated with mortality, attributable mortality was no assessed.
Conclusion
XDR-GNB infections have emerged as a clinical threat to hospitalized patients, particularly to those in the ICU. We have demonstrated that XDR-GNB infections were associated with exposures to several antibiotics, some of which may be amenable to antibiotic stewardship [37]. Predictors for mortality after HAIs with XDR-GNB were not modifiable, as mortality was more likely to be associated with age and underlying diseases.
Acknowledgments
Financial Support. This work was supported by the Centers for Disease Control and Prevention [R01 CI000537], the National Institute of Nursing Research [T90 NR010824] to S.A.C., and the Clinical and Translation Science Center at Weill Cornell Medical College [KL2RR024997] to S.A.W.
Footnotes
Conflicts of Interest. All authors report no conflicts of interest relevant to this article.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Lautenbach E, Polk RE. Resistant gram-negative bacilli: A neglected healthcare crisis? Am J Health Syst Pharm. 2007;64:S3–21. doi: 10.2146/ajhp070477. [DOI] [PubMed] [Google Scholar]
- 2.Patel G, Huprikar S, Factor SH, Jenkins SG, Calfee DP. Outcomes of carbapenem-resistant Klebsiella pneumoniae infection and the impact of antimicrobial and adjunctive therapies. Infect Control Hosp Epidemiol. 2008;29:1099–1106. doi: 10.1086/592412. [DOI] [PubMed] [Google Scholar]
- 3.Giske CG, Monnet DL, Cars O, Carmeli Y. Clinical and economic impact of common multidrug-resistant gram-negative bacilli. Antimicrob Agents Chemother. 2008;52:813–821. doi: 10.1128/AAC.01169-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Falagas ME, Bliziotis IA. Pandrug-resistant Gram-negative bacteria: the dawn of the post-antibiotic era? Int J Antimicrob Agents. 2007;29:630–636. doi: 10.1016/j.ijantimicag.2006.12.012. [DOI] [PubMed] [Google Scholar]
- 5.Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control. 2008;36:309–332. doi: 10.1016/j.ajic.2008.03.002. [DOI] [PubMed] [Google Scholar]
- 6.Clinical and Laboratory Standards Institute. Performance standards for antimicrobial susceptibility testing: 17th informational supplement M100-S17. Clinical and Laboratory Standards Institute; Wayne, Pa: 2007. [Google Scholar]
- 7.Clinical and Laboratory Standards Institute. Performance standards for antimicrobial susceptibility testing: 18th informational supplement M100-S18. Clinical and Laboratory Standards Institute; Wayne, Pa: 2008. [Google Scholar]
- 8.Clinical and Laboratory Standards Institute. Performance standards for antimicrobial susceptibility testing: 19th informational supplement M100-S19. Clinical and Laboratory Standards Institute; Wayne, Pa: 2009. [Google Scholar]
- 9.Larson EL, Cohen B, Ross B, Behta M. Isolation precautions for methicillin-resistant Staphylococcus aureus: Electronic surveillance to monitor adherence. Am J Crit Care. 2010;19:16–26. doi: 10.4037/ajcc2009467. [DOI] [PubMed] [Google Scholar]
- 10.Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest. 1991;100:1619–1636. doi: 10.1378/chest.100.6.1619. [DOI] [PubMed] [Google Scholar]
- 11.Safdar N, Maki DG. The commonality of risk factors for nosocomial colonization and infection with antimicrobial-resistant Staphylococcus aureus, enterococcus, gram-negative bacilli, Clostridium difficile, and Candida. Ann Intern Med. 2002;136:834–844. doi: 10.7326/0003-4819-136-11-200206040-00013. [DOI] [PubMed] [Google Scholar]
- 12.Gasink LB, Edelstein PH, Lautenbach E, Synnestvedt M, Fishman NO. Risk factors and clinical impact of Klebsiella pneumoniae carbapenemase-producing K. pneumoniae. Infect Control Hosp Epidemiol. 2009;30:1180–1185. doi: 10.1086/648451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Le Hello S, Falcot V, Lacassin F, Mikulski M, Baumann F. Risk factors for carbapenem-resistant Acinetobacter baumannii infections at a tertiary care hospital in New Caledonia, South Pacific. Scand J Infect Dis. 2010;42:821–826. doi: 10.3109/00365548.2010.496087. [DOI] [PubMed] [Google Scholar]
- 14.Harris AD, Smith D, Johnson JA, Bradham DD, Roghmann MC. Risk factors for imipenem-resistant Pseudomonas aeruginosa among hospitalized patients. Clin Infect Dis. 2002;34:340–345. doi: 10.1086/338237. [DOI] [PubMed] [Google Scholar]
- 15.Fortaleza CM, Freire MP, Filho Dde C, de Carvalho Ramos M. Risk factors for recovery of imipenem- or ceftazidime-resistant Pseudomonas aeruginosa among patients admitted to a teaching hospital in Brazil. Infect Control Hosp Epidemiol. 2006;27:901–906. doi: 10.1086/507288. [DOI] [PubMed] [Google Scholar]
- 16.Ansari SR, Hanna H, Hachem R, Jiang Y, Rolston K, Raad I. Risk factors for infections with multidrug-resistant Stenotrophomonas maltophilia in patients with cancer. Cancer. 2007;109:2615–2622. doi: 10.1002/cncr.22705. [DOI] [PubMed] [Google Scholar]
- 17.van der Veen EL, Schilder AG, Timmers TK, et al. Effect of long-term trimethoprim/sulfamethoxazole treatment on resistance and integron prevalence in the intestinal flora: a randomized, double-blind, placebo-controlled trial in children. J Antimicrob Chemother. 2009;63:1011–1016. doi: 10.1093/jac/dkp050. [DOI] [PubMed] [Google Scholar]
- 18.Brown PD, Freeman A, Foxman B. Prevalence and predictors of trimethoprim-sulfamethoxazole resistance among uropathogenic Escherichia coli isolates in Michigan. Clin Infect Dis. 2002;34:1061–1066. doi: 10.1086/339491. [DOI] [PubMed] [Google Scholar]
- 19.Halwani M, Solaymani-Dodaran M, Grundmann H, Coupland C, Slack R. Cross-transmission of nosocomial pathogens in an adult intensive care unit: incidence and risk factors. J Hosp Infect. 2006;63:39–46. doi: 10.1016/j.jhin.2005.10.012. [DOI] [PubMed] [Google Scholar]
- 20.Zaoutis TE, Goyal M, Chu JH, et al. Risk factors for and outcomes of bloodstream infection caused by extended-spectrum beta-lactamase-producing Escherichia coli and Klebsiella species in children. Pediatrics. 2005;115:942–949. doi: 10.1542/peds.2004-1289. [DOI] [PubMed] [Google Scholar]
- 21.Alexiou VG, Michalopoulos A, Makris GC, Peppas G, Samonis G, Falagas ME. Multi-drug-resistant gram-negative bacterial infection in surgical patients hospitalized in the ICU: a cohort study. Eur J Clin Microbiol Infect Dis. 2012;31:557–566. doi: 10.1007/s10096-011-1347-2. [DOI] [PubMed] [Google Scholar]
- 22.Nseir S, Di Pompeo C, Diarra M, et al. Relationship between immunosuppression and intensive care unit-acquired multidrug-resistant bacteria: a case-control study. Crit Care Med. 2007;35:1318–1323. doi: 10.1097/01.CCM.0000261885.50604.20. [DOI] [PubMed] [Google Scholar]
- 23.Falagas ME, Koletsi PK, Bliziotis IA. The diversity of definitions of multidrug-resistant (MDR) and pandrug-resistant (PDR) Acinetobacter baumannii and Pseudomonas aeruginosa. J Med Microbiol. 2006;55:1619–1629. doi: 10.1099/jmm.0.46747-0. [DOI] [PubMed] [Google Scholar]
- 24.Falagas ME, Karageorgopoulos DE. Pandrug resistance (PDR), extensive drug resistance (XDR), and multidrug resistance (MDR) among Gram-negative bacilli: need for international harmonization in terminology. Clin Infect Dis. 2008;46:1121–1132. doi: 10.1086/528867. [DOI] [PubMed] [Google Scholar]
- 25.Ray AJ, Hoyen CK, Taub TF, Eckstein EC, Donskey CJ. Nosocomial transmission of vancomycin-resistant enterococci from surfaces. JAMA. 2002;287:1400–1411. doi: 10.1001/jama.287.11.1400. [DOI] [PubMed] [Google Scholar]
- 26.Aloush V, Navon-Venezia S, Seigman-Igra Y, Cabili S, Carmeli Y. Multidrug-resistant Pseudomonas aeruginosa: risk factors and clinical impact. Antimicrob Agents Chemother. 2006;50:43–48. doi: 10.1128/AAC.50.1.43-48.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gudiol C, Tubau F, Calatayud L, et al. Bacteraemia due to multidrug-resistant Gram-negative bacilli in cancer patients: risk factors, antibiotic therapy and outcomes. J Antimicrob Chemother. 2011;66:657–563. doi: 10.1093/jac/dkq494. [DOI] [PubMed] [Google Scholar]
- 28.Lye DC, Earnest A, Ling ML, et al. The impact of multidrug resistance in healthcare-associated and nosocomial Gram-negative bacteraemia on mortality and length of stay: cohort study. Clin Microbiol Infect. 2012;18:502–508. doi: 10.1111/j.1469-0691.2011.03606.x. [DOI] [PubMed] [Google Scholar]
- 29.Nguyen M, Eschenauer GA, Bryan M, et al. Carbapenem-resistant Klebsiella pneumoniae bacteremia: factors correlated with clinical and microbiologic outcomes. Diagn Microbiol Infect Dis. 2010;67:180–184. doi: 10.1016/j.diagmicrobio.2010.02.001. [DOI] [PubMed] [Google Scholar]
- 30.Peralta G, Sanchez MB, Garrido JC, et al. Impact of antibiotic resistance and of adequate empirical antibiotic treatment in the prognosis of patients with Escherichia coli bacteraemia. J Antimicrob Chemother. 2007;60:855–863. doi: 10.1093/jac/dkm279. [DOI] [PubMed] [Google Scholar]
- 31.Lodise TP, Patel N, Kwa A, et al. Predictors of 30-day mortality among patients with Pseudomonas aeruginosa bloodstream infections: impact of delayed appropriate antibiotic selection. Antimicrob Agents Chemother. 2007;51:3510–3515. doi: 10.1128/AAC.00338-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Manikal Vivek M, Landman D, Saurina G, Oydna E, Lal H, Quale J. Endemic carbapenem-resistant Acinetobacter species in Brooklyn, New York: Citywide prevalence, interinstitutional spread, and relation to antibiotic usage. Clin Infect Dis. 2000;31:101–106. doi: 10.1086/313902. [DOI] [PubMed] [Google Scholar]
- 33.Magiorakos AP, Srinivasan A, Carey RB, et al. Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin Microbiol Infect. 2012;18:268–81. doi: 10.1111/j.1469-0691.2011.03570.x. [DOI] [PubMed] [Google Scholar]
- 34.Klompas M. Does this patient have ventilator-associated pneumonia? JAMA. 2007;297:1583–1593. doi: 10.1001/jama.297.14.1583. [DOI] [PubMed] [Google Scholar]
- 35.Marsh JL, Hutton JL, Binks K. Removal of radiation dose response effects: an example of over-matching. BMJ. 2002;325:327–330. doi: 10.1136/bmj.325.7359.327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Harris AD, Samore MH, Lipsitch M, Kaye KS, Perencevich E, Carmeli Y. Control-group selection importance in studies of antimicrobial resistance: examples applied to Pseudomonas aeruginosa, Enterococci, and Escherichia coli. Clin Infect Dis. 2002;34:1558–1563. doi: 10.1086/340533. [DOI] [PubMed] [Google Scholar]
- 37.Zhou JJ, Patel SJ, Jia H, et al. Clinicians' knowledge, attitudes, and practices regarding infections with multidrug-resistant gram-negative bacilli in intensive care unit. Infect Control Hosp Epidemiol. 2013;34:274–283. doi: 10.1086/669524. [DOI] [PMC free article] [PubMed] [Google Scholar]