Data collection and patients
This study was conducted at the following five medical centers in Taiwan: Taipei Veterans General Hospital (TVGH, 2900 beds), Tri-Service General Hospital (TSGH, 1712 beds), Mackay Memorial Hospital (MMH, 2055 beds), and National Taiwan University Hospital (NTUH, 2200 beds) in Northern Taiwan and Changhua Christian Hospital (CCH, 1676 beds) in Central Taiwan. Data was collected between July 2012 and June 2016. The inclusion criteria for A. baumannii bacteremic pneumonia were: (1) ≥1 positive blood culture for A. baumannii which could not be attributed to an infection source other than the lower respiratory tract; (2) a clinical course compatible with the diagnosis of pneumonia, including a new pulmonary infiltrate plus one additional criterion (fever ≥38°C, blood leukocytosis ≥10000 cells/mm3 or leucopenia ≤3000 cells/mm3), together with one or more of the following conditions: new cough, change in sputum color, chest pain, and dyspnea; (3) ≥1 positive respiratory sample (sputum, endotracheal aspirate, or broncho-alveolar lavage [BAL]) for A. baumannii collected within 48 hours before or after the first positive blood culture for A. baumannii. Patients below 20 years of age or without a complete medical record were excluded. The study protocol was approved by the Research Ethics Committee of all participating hospitals. Informed consent was waived because of the retrospective nature of the study and the analysis used anonymous clinical data.
Study variables and definitions
The following data were collected from patient’s medical records: demographic information, comorbid conditions, duration of hospital and intensive care unit (ICU) stays, time, dose and route of antimicrobial therapy, the use of ventilator, and procedures (central venous catheters, arterial catheters, foley catheter, nasogastric tube, hemodialysis, and tracheostomy) at the time of bacteremia.
The onset of bacteremia was defined as the day the positive blood culture of A.baumannii was collected. The bacteremic pneumonia was considered acquired in the ICU if the positive respiratory sample for A.baumannii and positive blood culture for A.baumannii were both obtained at least 48 hours after ICU admission. Previous ICU admission was defined as being admitted to ICU within 4 weeks prior to the onset of bacteremia.
Immunosuppressive therapy was defined as receiving cytotoxic agents within 6 weeks, corticosteroids at a dosage equivalent to or higher than 15 mg of prednisolone daily for 1 week within 4 weeks, or other immunosuppressive agents within 2 weeks of the onset of bacteremia. Chronic kidney disease was defined as an estimate glomerular filtration rate <60 mL/min/1.73 m2. Neutropenia referred to an absolute neutrophil count <0.5 × 109 neutrophils/L. Recent surgery was defined as undergoing an operation within 4 weeks of the onset of bacteremia. Previous ventilator use was defined as mechanical ventilation use for more than 3 days in the past 4 weeks. The severity of patient illness was evaluated using the Acute Physiology and Chronic Health Evaluation (APACHE) II score within 24 hours before bacteremia onset.
Appropriate antimicrobial therapy was defined as administration of the antimicrobial agent to the pathogen susceptible in vitro, within 48 hours after the onset of bacteremia, with an approved route and dosage appropriate for end organ(s) function. Antimicrobial therapy that did not meet this definition was considered inappropriate. Monotherapy with an aminoglycoside was also considered to be an inappropriate therapy. An antimicrobial agent (or antimicrobial agents)-based therapy was defined as treatment with the antimicrobial agent(s) alone or in combination with other antimicrobial agent(s). The colistin loading dose was 5 mg/kg colistin base activity, followed by 5 mg/kg/d colistin base activity divided over 8 or 12 hours in patients with normal renal function. For those with impaired renal function, the dosage was adjusted according to renal function as previously described [19, 20]. The loading dose of tigecycline was 100 mg, followed by a maintenance dose of 50 mg every 12 hours. The primary outcome was all-cause 14-day mortality after the onset of A. baumannii bacteremia.
Microbiological studies
The presumptive identification of the isolates to the level of the A. baumannii complex was determined using the Vitek 2 system (bioMérieux). All A. baumannii complex bloodstream isolates were regrown from storage, identified to species level, and tested for their susceptibility to various antimicrobials. A multiplex polymerase chain reaction method was used to identify A. baumannii to the genomic species level [21]. Polymicrobial bacteremia was defined as the concurrent isolation of one or more microorganisms other than A. baumannii from blood. Antimicrobial susceptibility to ampicillin-sulbactam, ceftazidime, cefepime, piperacillin-tazobactam, imipenem, meropenem, ciprofloxacin, levofloxacin, and amikacin was determined by the agar dilution method according to Clinical Laboratory Standards Institute criteria. Colistin minimal inhibitory concentrations (MICs) were determined by the broth macrodilution method to problems arising from the fact that the surface charge on the polystyrene microplate applied during manufacturing influences the level of colistin adsorption to the plate surface [22, 23] Tigecycline MICs were determined by the broth microdilution method using fresh medium [24]. Multidrug resistance (MDR) was defined as resistance to at least one agent in at least three of the following classes of antimicrobials: antipseudomonal cephalosporins, antipseudomonal carbapenems, ampicillin-sulbactam, fluoroquinolones, and aminoglycosides. Carbapenem resistance was defined as resistance to imipenem or meropenem. Extensive drug resistance (XDR) referred to non-susceptibility to imipenem or meropenem and all drug classes with the exception of colistin and tigecycline.
Statistical analysis
Chi-squared test or Fisher’s exact test was used to compare categorical variables. The Student’s t test or Man-Whitney rank sum test was used to analyze continuous variables. Logistic regression models were used to assess independent risk factors for 14-day mortality. Biologically plausible variables which were significantly associated with mortality (p ≤0.05) in the univariable analysis were included in the multivariable analysis. Stepwise logistic regression was used. Interactions between the APACHE II score and the covariates were assessed in the logistic regression model. APACHE II scores were categorized into four groups (APACHE II score ≤15, >15 and ≤25, >25 and ≤35, >35) in the logistic regression models based on their quartile distribution and previous studies [18, 25]. The time between bacteremia onset to mortality was analyzed using Kaplan-Meier survival analysis. A p-value <0.05 was considered to be statistically significant. All the analyses were processed with Stata software version 12.