Settings

This study was conducted at Siriraj hospital, a 2200-bed, tertiary-care university hospital in Bangkok, Thailand. The study protocol including waiver of informed consent for using patients' clinical isolates were approved by Siriraj Institutional Review board. From July 1, 2007 - December 31, 2008, all patients in whom at least one blood culture positive for NFGNB were prospectively identified through the microbiology laboratory database. Only patients who met the criteria of systemic inflammatory response syndrome (SIRS) within 24 hours before or after obtaining blood culture were enrolled in the study. If NFGNB were isolated on multiple occasions from the same patient, only the first episode of bacteremia was included.

Microbiologic procedures and isolate identification

Microbiological and susceptibility results of all patients were retrieved from the microbiology laboratory database. All blood cultures during the study period were processed by the BactT/ALERT system (bioMeriéux) according to manufacturer’s protocol. Identification to the species level was achieved by using the Vitek 2 and/or API 20NE systems (bioMeriéux). All clinical isolates that could not be satisfactorily identified by the Vitek 2 or API 20NE were further characterized by 16S rDNA sequencing. Approximately 800 bp at the 5’ terminal of 16S rDNA gene was PCR amplified and sequenced using primers and protocols as described elsewhere [12, 13]. The sequencing results were compared with databases in the GenBank (the nucleotide-nucleotide Basic Local Alignment Search Tool or BLAST).

Susceptibility testing was performed by Disk Diffusion or Broth Microdilution method as appropriate. Susceptibilities to each antimicrobial agent were determined according to criteria established by the Clinical and Laboratory Standards Institute [14, 15]. Since the Clinical and Laboratory Standards Institute (CLSI) criteria for polymyxin E and tigecycline are not available, susceptibilities to tigecycline and polymyxin E were interpreted by using the European Committee on Antimicrobial Susceptibility Testing (EUCAST) clinical breakpoints [16]. We could not report the susceptibility results of some NFGNB species because the standard interpretation method is not available. Names of antimicrobial agent were abbreviated as follows: ceftazidime (CAZ), cefipime (FEP), ciprofloxacin (CIP), colistin or polymyxin E (CST), gentamycin (GEN), cotrimoxazole (SXT), imipenem-cilastatin (IMP), piperacillin-tazobactam (TZP) and tigecycline (TIG).

Data collection

Medical records were retrospectively reviewed to obtain data including age, sex, hospital service, previous hospitalization, comorbidities, presence of a central venous catheter at the onset of infection, use of antimicrobial therapy or immunosuppressive agents in the preceding 30 days, suspected source(s) of bacteremia and infection-related mortality within 28 days after the onset of bacteremia (ID-mortality). The presence of the following comorbidities at the time of bacteremia was documented: cardiovascular diseases, chronic renal diseases, chronic liver diseases, chronic lung diseases, diabetes mellitus, neurological diseases, hematologic malignancy, solid cancer, HIV infection and receiving immunosuppressive therapy. We also recorded a number of blood cultures obtained, a number of positive blood cultures, other pathogens identified in the same set of blood cultures and hospital days both before and after the onset of bacteremia. We considered the date of obtaining the first positive blood culture as the onset of bacteremia.

Definitions

Pseudomonas aeruginosa and Acinetobacter baumannii were considered the common NFGNB while all other NFGNB pathogens were considered the uncommon NFGNB.

Multidrug resistance was defined as resistance to carbapenems, second and third generation-cephalosporins, anti-pseudomonas penicillins, fluoroquinolones and aminoglycosides.

A bacteremic episode was considered to be hospital-acquired if one of the following criteria was true; 1) it occurred at least 48 hours after the admission and did not present or incubate at the time of admission; 2) it presented on the admission but the patient had been transferred from another medical center or long-term care facility, had spent at least 48 hours in the given facility; or 3) it presented on the admission but the patient has been hospitalized within the past 2 weeks [17].

A site of infection was determined by using the CDC definitions of nosocomial infections [17]. The site of infection was considered a suspected source of bacteremia if a similar NFGNB had also been isolated from that site within 24 hours before or after the onset of bacteremia. Therefore, each patient may have more than one suspected sources of bacteremia.

Receiving immunosuppressive agents was documented if the patient had a history of corticosteroid use (receipt of prednisone at a dosage of 20 mg per day (or equivalent) for at least 2 weeks) and/or history of receipt of chemotherapeutic agents in the preceding 30 days.

ID-mortality was defined as death in the setting of clinical evidence of active infection (elevated WBC and or elevated body temperature) and death within 5 days of last positive culture result.

Statistical analysis

Descriptive statistics were used to express overall results. Univariate analysis was performed to determine an unadjusted association between the uncommon NFGNB bacteremia and other variables. Categorical variables were compared by using chi-square or Fisher’s exact test while continuous variables were compared by using t-test or Wilcoxon rank sum test as appropriate. To identify factors that independently associated with the uncommon NFGNB bacteremia, we subsequently built a multivariate logistic model by the stepwise method which is a combination of backward elimination and forward selection approaches. Variables were included in the multivariable model if they presented a p-value ≤0.20 in univariate analysis and then removed from the final multivariable model if they did not exhibit an adjusted p-value <0.05. A 2-tailed p-value of <0.05 was considered significant. All statistical calculations were performed by using STATA, version 12 (Stata Corp, College Station, TX).

Distribution of causative pathogens

During the study period, there were a total of 221 patients with NFGNB bacteremia. Two patients (0.9%) had blood cultures positive for two species of NFGNB in a same set. The first patient had bacteremia caused by Chryseobacterium meningosepticum and Elizabethkingia meningoseptica. The second patient had bacteremia caused by Ralstonia spp. and Burkholderia mallei. Therefore, we detected 223 clinical isolates of NFGNB in 221 unique patients. In addition to a NFGNB pathogen, 25 patients (11.3%) also had a blood culture positive for a non-NFGNB pathogen, including Escherichia coli (n = 7), Klebsiella pneumoniae (n = 5), Enterobacter spp. (n = 3), Enterococcus faecalis (n = 2), Staphylococcus aureus (n = 2), alpha-hemolytic streptococci (n = 1) and fungi (n = 5). The most common combination was P. aeruginosa and E. coli (n = 7), followed by P. aeruginosa and K. pneumoniae (n = 5).

Susceptibility patterns

Clinical characteristics

Antimicrobial therapy

Due to a very high prevalence of multi-drug resistant pathogens, only half of patients (50.7%, 112/221) received at least one antimicrobial agent that active against causative pathogen(s) on the onset date. Percent of patients who received adequate antimicrobial therapy on the onset date was 66.1% in P. aeruginosa, 31.5% in A. baumannii, 16.7% in S. maltophilia, 100% in B. pseudomallei and 58.8% among the rest.

Clinical outcomes

The ID-mortality was 38.9% (86/221) in all study subjects, 52.5% (71/135) in the common NFGNB group and 17.4% (15/86) in the uncommon NFGNB group. The ID-mortality was highest among patients with A. baumannii bacteremia (63.0%, 46/73), followed P. aeruginosa bacteremia (40.3%, 25/52). There was no significant difference in the ID-mortality between those with monomicrobial bacteremia vs. polymicrobial bacteremia (39.7% vs. 33.3%; p = 0.53). Details of the ID-mortality across the causative pathogens are shown in Figure 1.

Factors associated with uncommon NFGNB bacteremia

To identify factors associated with bacteremia caused by uncommon NFGNB, we subsequently compared characteristics of patients with uncommon NFGNB bacteremia (S. maltophilia, B. pseudomallei and all other uncommon NFGNB species) to patients with common NFGNB bacteremia (P. aeruginosa and A. baumannii). When baseline characteristics of patients in both groups were compared, a median length of hospital stay prior to the onset of bacteremia was slightly lower in the uncommon NFGNB group (1.5 day vs. 10 days). However, this difference did not reach statistical significance. Patients in the uncommon NFGNB group were less likely to have underlying hematologic malignancy as well as exposure to beta-lactams, aminoglycosides and metronidazole. However, the uncommon NFGNB group had a higher prevalence of underlying cardiovascular diseases.