We performed a post hoc analysis of individual patients included in a previous study [15]. In the current study, a case-control-control design was used to evaluate risk factors associated with MDR-ESBL-EC in community-onset infections. Study subjects were prospectively identified as previously described [15] through a computerized database maintained by the clinical microbiology laboratory. The study was conducted between September 2010 and May 2011 at Samsung Medical Center in Seoul (a 1,950-bed tertiary care university hospital) and at Samsung Changwon Hospital (a 700-bed community-based university-affiliated hospital in Changwon, Korea). Patients with community-onset infection due to E. coli were included in the study. The inclusion and exclusion criteria were the same as reported previously [15]. A case patient was defined as a person whose clinical sample yielded MDR-ESBL-EC. A control group I (CG I) patient was defined as a person whose clinical sample yielded ESBL-EC that did not meet the criteria for MDR. A control group II (CG II) patient was defined as a patient with a non-ESBL-EC infection. For each patient who presented with an infection caused by ESBL-EC, one CG II was randomly chosen from a group of outpatients from whom non-ESBL-EC had been isolated from a clinical sample that had been sent to the same laboratory for culturing during the week following presentation of the case patient. The CG II patients were matched according to the culture specimen and acquisition unit. The patient data collected included age, sex, underlying disease, severity of underlying diseases as classified by McCabe and Jackson criteria [16], presence of severe sepsis or septic shock at presentation, severity of illness at presentation according to the Pitt bacteremia score [2], and receipt of any antimicrobial therapy prior to onset of the infection. The presence of any of the following comorbid conditions was also documented: neutropenia, recent surgical procedure within the prior 3 months, corticosteroid use within the prior 1 month, immunosuppressive therapy within the prior 1 month, or the presence of an indwelling urinary catheter or percutaneous tube. Because the study was observational, the attending physician decided on indications for cultures, other tests, and treatments based on each patient’s individual clinical situation. The study was approved by the Institutional Review Board of Samsung Medical Center, Seoul, Korea. Written informed consent was not required because of the observational nature of the study.

The community-onset infections were further classified as healthcare-associated (HCA) or community-associated (CA). Episodes were considered to be HCA if any of the following criteria were present: a history of a 48 hours or longer hospital admission in the previous 90 days, hemodialysis, intravenous medication, home wound care in the previous 30 days, or residence in a nursing home or long-term care facility [9,17]. Community-onset infections that did not meet these criteria were considered to be CA. MDR-ESBL-EC was defined as an ESBL-EC that demonstrated in vitro resistance to all of the following three antibiotics or antibiotic classes: TMP-SMX, FQs (i.e., ciprofloxacin and levofloxacin), and gentamicin. Non-MDR-ESBL-EC was defined as ESBL-EC that did not meet the criteria for MDR. Community-onset infection was defined as an infection diagnosed within the first 48 hours of hospitalization. The sites of infection were determined by the physicians on the basis of clinical evaluation and the isolation of EC from the presumed portal of entry [9,18]. Primary bacteremia was defined according to Centers for Disease Control and Prevention definitions [18]. Patients with immunosuppression included those who were on immunosuppressive therapy (chemotherapeutic agents, immunosuppressive agents, or radiation therapy). Neutropenia was defined as an absolute neutrophil count < 500 neutrophils/mm³. Sepsis was defined as the presence of a systemic inflammatory response syndrome (any two of the following: tachypnea > 20 breaths per minute, white blood cell count < 4,000 or > 12,000 cells/µL, heart rate > 90 beats per minute, and fever > 38.0°C or hypothermia < 36.0°C) in addition to a documented or presumed infection [19]. Severe sepsis was defined as sepsis with one or more clinical signs of organ dysfunction and septic shock defined as sepsis with hypotension despite adequate fluid resuscitation along with the presence of perfusion abnormalities [19]. Prior antibiotic therapy was defined as the receipt of any systemic antibiotics for more than 48 hours in the preceding 30 days.

Identification of microorganisms, antimicrobial susceptibility testing, and ESBL confirmatory testing were performed as previously described [15]. Quality control was performed using the strains E. coli 25922 and Pseudomonas aeruginosa 27853.

For the characterization of ESBL types, polymerase chain reactions (PCRs) and sequencing of PCR products were performed using available stored isolates, as described previously [15]. ESBL-related genes, such as TEM, SHV, CTX-M, and OXA, were amplified by PCR from clinical isolates as described in previous studies [20,21]. The types of ESBL genes were identified by comparing the sequences to those in the database of G. Jacoby and K. Bush (http//lahey.org/Studies/).

Available MDR-ESBL-EC isolates were evaluated for genetic relatedness by pulsed-field gel electrophoresis (PFGE). For PFGE, agarose-embedded bacterial genomic DNA was digested with 20 U Xba l. The restriction fragments were separated by electrophoresis in 0.5 × Tris/borate/ethylenediaminetetraacetic acid (TBE) buffer. The subsequent PFGE analyses were performed on a CHEF-MAPPER XA apparatus (Bio-Rad Laboratories, Hercules, CA, USA) as described previously [22]. The isolates were considered genetically related if the Dice coefficient correlation was 80% or greater, which corresponds to the possibly related criterion of Tenover et al. [23].

The Student t test or the Mann-Whitney test was used to compare continuous variables, and the chi-square or Fisher exact test was used to compare categorical variables. We used a backward stepwise logistic regression analysis to control for the effects of confounding variables. All risk factors with a p < 0.1 at the bivariate level were included in the multivariate logistic model predicting MDR-ESBL-EC infection. All variables for which the p value was less than 0.05 in the multivariate analysis were retained in the final model. Interactions between variables were not introduced into the models. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated. All p values were two-tailed, and p < 0.05 was considered statistically significant. IBM SPSS version 19 (IBM Co., Armonk, NY, USA) was used for these analyses.

During the study period, a total of 108 patients with community-onset ESBL-EC infections were identified and all of them were included in this study. Of the 108 isolates, 30 (27.8%) were due to MDR-ESBL-EC. The mean age (± standard deviation) of the study population was 62.4 ± 16.0 years, and 70 (64.8%) were women. The most common underlying disease was diabetes mellitus (26.9% [28/108]) and the most common site of infection was the urinary tract (67.6% [73/108]). To analyze risk factors associated with community-onset MDR-ESBL-EC, data from the 30 patients with MDR-ESBL-EC were compared to those of the 78 patients with non-MDR-ESBL-EC (CG I) and 100 patients with non-ESBL-EC infections (CG II). The clinical characteristics of the case and control groups and risk factors for community-onset MDR-ESBL-EC infections are summarized in Table 1. Variables such as age, sex, underlying disease, and severity of illness at presentation were similar between the case and control groups. The source of isolates did not differ significantly between case and control groups. Compared with CG I, significant factors associated with community-onset MDR-ESBL-EC infections were immunosuppressive therapy and prior receipt of FQs (all p < 0.05). Compared with CG II, HCA infection, severity of illness at presentation, immunosuppressive therapy, and prior receipt of antibiotics, especially FQs, were significantly associated with MDR-ESBL-EC infections.

To identify independent risk factors associated with community-onset MDR-ESBL-EC infections, a logistic regression analysis was performed including variables with an association of p < 0.1 at the univariate level. The independent risk factors associated with community-onset MDR-ESBL-EC infections are shown in Table 2. Immunosuppressive therapy (OR, 10.47; 95% CI, 1.07 to 12.05; p = 0.044 in model 1), prior receipt of FQs (OR, 3.16; 95% CI, 1.11 to 8.98; p = 0.31 in model 1; OR, 15.53; 95% CI, 2.86 to 84.27; p = 0.01 in model 2), and HCA infection (OR, 5.98; 95% CI, 2.26 to 15.86; p < 0.01 in model 2) were found to be independent factors associated with community-onset MDR-ESBL-EC infections.

The 108 community-onset ESBL-EC isolates demonstrated variable resistance to other antibiotics. The prevalence of resistance to piperacillin/tazobactam (36.7% vs. 29.5%, p = 0.472) and amoxicillin/clavulanic acid (73.3% vs. 59.0%, p = 0.189) were similar for the MDR-ESBL-EC group and the non-MDR-ESBL-EC group. However, strains resistant to tobramycin were more frequent in the MDR-ESBL-EC group than the non-MDR-ESBL-EC group (93.3% vs. 44.9%, p < 0.001) (Table 3). None of the strains were resistant to imipenem or meropenem. Both MDR- and non-MDR-ESBL-EC isolates were highly susceptible to amikacin (100% vs. 98.7%, p = 1.000). Ninety-nine isolates of ESBL-EC were available for further microbiological study. All of these isolates produced ESBLs from the CTX-M family. We compared MDR-ESBL-EC with non-MDR-ESBL-EC and the distribution of ESBL genes is summarized in Table 4. CTX-M-15 was the most common type of MDR-ESBL-EC (59.1% [13/22]), while CTX-M-14 was the most common non-MDR-ESBL-EC (41.6% [32/77]). CTX-M-15 was significantly associated with MDR-ESBL-EC (59.1% vs. 32.5%, p = 0.028).

PFGE of 22 MDR-ESBL-EC strains was performed to characterize the clonality of these isolates. Because two isolates (test number 1 and 19) among the 22 MDR-ESBL-EC did not provide DNA bands by the PFGE method used in this study, we analyzed only 20 isolates. Applying a similarity index of 80% to the PFGE results, four clonal clusters were identified, consisting of two or three isolates for each cluster. The 10 remaining isolates showed greater clonal diversity (Fig. 1). The microbiological characteristics of MDR-ESBL-EC that were available for further microbiological evaluation are shown in Table 5.