Recent trends show an increase in the prevalence of nosocomial pneumonia (NP) caused by multidrug-resistant (MDR) Gram-negative bacteria, most commonly Pseudomonas aeruginosa with documented resistance to β-lactams, carbapenems, aminoglycosides, and fluoroquinolones [1-3]. Consequently, the therapeutic effectiveness of current therapies for bacterial NP is becoming increasingly limited, emphasizing the need for development of new and effective antimicrobials as well as novel strategies to prevent resistance emergence [4,5].

Nosocomial pneumonia due to P. aeruginosa (Pa-NP) is associated with considerable morbidity, prolonged hospitalization, increased costs, and mortality [6-8]. P. aeruginosa is one of the few pathogens independently associated with increased mortality among patients with sepsis or pneumonia in the ICU setting [6,9]. The mortality associated with Pa-NP is further increased when inappropriate initial antibiotic therapy (IIAT) is prescribed, usually due to the presence of MDR pathogens [10-13]. The overall impact of Pa-NP on clinical outcomes and healthcare costs underscores the importance of this nosocomial infection. Therefore, we performed a multinational study with the following objectives: first, to evaluate the prevalence of MDR Pa-NP and to identify clinical risk factors associated with MDR Pa-NP; second, to evaluate the influence of MDR status on patient outcomes.


Study design and ethical standards

We conducted a retrospective study in 12 hospitals in 5 countries (United States, n = 3; France, n = 2; Germany, n = 2; Italy, n = 2; and Spain, n = 3). Eligible patients were aged ≥18 years consecutively admitted for their index hospitalization within 36 months prior to study initiation in 2013. All eligible patients met a clinical diagnosis of NP defined as new or progressive infiltrates consistent with pneumonia on chest radiograph or computed tomography and either a temperature >38.3°C or leukocytosis >10,000 cells/mm3 or both. To be eligible, patients had to have P. aeruginosa cultured from at least one of the following respiratory specimens, including sputum, pleural fluid, flexible bronchoscopy with protected specimen brush, bronchoalveolar (BAL), transbronchial biopsy, nonbronchoscopic BAL, or tracheobronchial aspirate in intubated patients. Microbiologic cultures (qualitative or quantitative) had to be obtained within the 12-hour window before or the 12-hour window after the initiation of antibiotic(s) targeting P. aeruginosa. Each investigator obtained approval and a waiver of patient consent from an Independent Ethics Committee or Institutional Review Board at their institution before commencing the study. The list of all ethical bodies that approved the study can be found in the Acknowledgements section.

Endpoints and covariates

The primary endpoints examined were multidrug-resistance and hospital mortality. We collected important covariates including demographics, comorbidities (heart failure, diabetes mellitus, chronic obstructive pulmonary disease, chronic kidney disease, chronic liver disease, hematologic malignancy, solid tumor, HIV/AIDS, and dementia). In addition, important process-of-care variables, including ICU admission, mechanical ventilation, vasopressor administration, and the appropriateness of initial antibiotic therapy, were collected.


To be classified as MDR, the P. aeruginosa isolate had to be non-susceptible to one or more agents in three or more of the following antimicrobial categories, as determined by the European Center for Disease Prevention and Control (ECDC) and the Centers for Disease Control and Prevention (CDC): aminoglycosides, antipseudomonal carbapenems, antipseudomonal cephalosporins, antipseudomonal fluoroquinolones, antipseudomonal penicillins plus β-lactamase inhibitors, monobactams, phosphonic acids, and polymixins. To be classified as extensively drug-resistant (XDR), the P. aeruginosa isolate had to be non-susceptible to one or more agents in all but two or more of the aforementioned antimicrobial categories [14]. Antimicrobial treatment was deemed to be appropriate (AIAT) if at least one of the initially prescribed antibiotics was active against the identified P. aeruginosa isolate based on in vitro susceptibility testing and this antibiotic was administered within 24 hours after collection of the respiratory specimen [15].

Antimicrobial susceptibility testing

Microbiology laboratories performed antimicrobial susceptibility testing of isolates using disk diffusion or automated testing methods according to guidelines and breakpoints established by the Clinical Laboratory and Standards Institute (CLSI) [16] and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [17].

Statistical analyses

Continuous variables were reported as means with standard deviation or the median and interquartile range from non-normally distributed data. Differences between continuous variables were tested using Student’s t-test or the nonparametric Mann-Whitney U-test. Categorical data were summarized as proportions, and the Chi-square test or Fisher’s exact test for small samples was used to examine differences between groups. Univariate and multivariate logistic regression models were constructed to identify clinical risk factors associated with multidrug-resistance. All variables that showed a significant result in the univariate analysis (≤0.10) were included in the corresponding multivariate logistic regression analysis. All variables entered into the models were examined to assess for co-linearity, and interaction terms were tested. The model’s calibration was assessed with the Hosmer-Lemeshow goodness-of-fit test. A Cox proportional hazards model was constructed to determine variables independently associated with hospital mortality. This test was selected to exclude the influence of time-dependent covariates on hospital mortality and to adequately control for imbalances in baseline and clinical characteristics when constructing a survival curve. All tests were two-tailed, and a P-value <0.05 was deemed a priori to represent statistical significance. All analyses were performed with SPSS software, version 19.0 (IBM SPSS, Chicago, IL, USA).


Seven hundred and forty patients with Pa-NP met the inclusion criteria and were enroled in the study: 258 (34.9%) from the United States, 141 (19.1%) from France, 120 (16.2%) from Germany, 113 (15.3%) from Spain and 108 (14.6%) from Italy. The prevalence of multidrug resistance was 30.5%. The patients’ baseline and clinical characteristics are shown in Table 1. Patients with pneumonia caused by MDR strains of P. aeruginosa were significantly younger and were more likely to be admitted to the hospital from an inpatient rehabilitation facility compared to patients infected with non-MDR strains. Patients with MDR strains were significantly more likely to have received antibiotics in the 30 days prior to the diagnosis of pneumonia and were also more likely to have chronic obstructive pulmonary disease and diabetes mellitus. A significantly higher proportion of patients who were infected with an MDR strain received IIAT (37.9% versus 19.2%, P <0.001) and required ICU admission (79.6% versus 71.4%, P = 0.019) compared to those with a non-MDR strain.

Table 1 Clinical and epidemiological characteristics of multidrug (MDR) and non-multidrug resistant patients with Pseudomonas aeruginosa pneumonia

Susceptibility to all antibiotic classes tested was significantly lower in patients infected with MDR strains (Table 2). Antibiotic susceptibility by country is found in Table 3. Germany (44.2%) and Spain (43.4%) were found to have the highest prevalence of MDR, followed by France (33.3%), Italy (22.2%) and the United States (20.5%). Table 4 shows the results of a multivariable logistic regression model that identified the variables associated with pneumonia caused by MDR strains of P. aeruginosa. Decreasing age in increments of one year, diabetes mellitus, and ICU admission were independently associated with MDR P. aeruginosa pneumonia.

Table 2 Antibiotic susceptibility
Table 3 Antibiotic susceptibility by country
Table 4 Significant univariate and multivariate logistic regression analysis of predictors for multidrug-resistant (MDR) Pseudomonas aeruginosa pneumonia

The overall, hospital mortality was 35.7% (n = 264). Mortality was significantly different between the United States and European countries: United States, 22.5%; France, 37.6%; Germany, 41.7%; Spain, 46.9%; and Italy, 46.3%. Patients with MDR strains had a significantly higher in-hospital mortality rate compared to non-MDR infected patients (Table 1). A Cox proportional hazards model confirmed MDR status as an independent predictor of mortality (hazard ratio (HR) 1.39, 95% CI 1.05 to 1.83, P = 0.021) along with increasing age, heart failure, concomitant bacteremia, mechanical ventilation, and patients residing in Germany, Italy, and Spain (Table 5). Cox model-adjusted survival curve analysis controlling baseline and clinical imbalances confirmed the influence of MDR on in-hospital mortality (Figure 1).

Table 5 Cox proportional hazards model of significant predictors for in-hospital mortality
Figure 1
figure 1

Cox proportional hazards model curve comparing patients with multidrug-resistant (MDR)-Pseudomonas aeruginosa and those with non-MDR P. aeruginosa nosocomial pneumonia.


This international investigation representing the largest cohort study of Pa-NP demonstrated high prevalence of MDR at 30.5%. Infection caused by MDR P. aeruginosa was found to be an important determinant of hospital mortality, thus, it is critical for clinicians to identify patients at risk of MDR from the onset of infection. Our analysis suggests that the patient’s age, comorbid conditions specifically diabetes, and the severity of infection as indicated by the need for ICU admission predicts infection with a MDR strain of P. aeruginosa.

The prevalence of MDR Pa-NP is variable depending on the type of study performed and the participating institutions. A recent large epidemiologic study from the United States identified 205,526 P. aeruginosa isolates (187,343 pneumonia; 18,183 bloodstream infection (BSI)) and 95,566 Enterobacteriaceae specimens (58,810 pneumonia; 36,756 BSI) associated with infection [1]. Prevalence of MDR P. aeruginosa (MDR Pa) was approximately 15-fold greater than carbapenem-resistant-Enterobacteriaceae in both infection types. A net rise in MDR Pa as a proportion of all P. aeruginosa infections occurred from 2000 to 2009. Likewise, data from the National Healthcare Safety Network (NHSN) in the United States revealed an increased prevalence of MDR Pa VAP from the period 2007 to 2008 to the period 2009 to 2010, but, it should be noted the overall prevalence of MDR Pa was 17.7% in the latter time period, markedly less than our study [3]. The international composition of the participants is the most likely explanation for the higher prevalence of MDR strains in our study.

The literature also varies with respect to the outcomes of patients with MDR Pa-NP. Peña et al. examined a Spanish cohort of 91 episodes of ventilator-associated pneumonia (VAP) in 83 patients, 31 caused by susceptible P. aeruginosa and 60 by MDR Pa strains [18]. These investigators found that susceptible P. aeruginosa infections were more likely than MDR Pa episodes to receive AIAT and definitive antimicrobial therapy, and in a logistic regression model IIAT was identified as an independent risk factor for early mortality. A recent meta-analysis supports these findings by demonstrating that MDR status is an important determinant of mortality due to nosocomial infections attributed to Gram-negative bacteria, where P. aeruginosa and Acinetobacter species were the most common isolates [19]. Di Pasquale et al. recently found MDR status was not associated with a higher rate of ICU or hospital mortality in patients with ICU-acquired pneumonia. However, unlike our study, the etiology of infection was a mix of Gram-positive and Gram-negative pathogens and there was a small number of MDR Pa cases (n = 18) [20].

Increasing antimicrobial resistance in P. aeruginosa infections seems to be the most important predictor of outcome. In a recent Brazilian study of P. aeruginosa bacteremia isolates from 120 patients [21], 45.8% were resistant to carbapenems, and 23.3% expressed a metallo-β-lactamase gene, blaSPM-1 (57%) or blaVIM-type (43%). Cefepime-resistance, MDR status and XDR isolates were independently associated with IIAT, which was an important predictor of mortality. These studies support the importance of appropriate and timely antibiotic therapy as a potential determinant of outcome for serious infections attributed to P. aeruginosa. Given the association of antibiotic resistance with increasing administration of IIAT and greater hospital mortality, several strategies have been developed to improve upon the appropriateness of empiric therapy in patients at risk of infection with P. aeruginosa and other antibiotic-resistant pathogens.

A number of investigations have identified risk factors and scoring systems for infection with MDR pathogens, including MDR Pa [22-25]. Major limitations of such approaches are that the potential for IIAT remains, although potentially diminished, and the resultant overuse of broad-spectrum antibiotics in many patients because of the non-specificity of the scoring systems. Novel methods to improve early identification of pathogens and antibiotic susceptibilities are also entering the diagnostic arena. Such diagnostic technology advances offer the potential to maximize administration of appropriate antibiotic therapy while minimizing unnecessary antibiotic exposure. These approaches include the use of molecular methods (for example, polymerase chain reaction electrospray ionization mass spectrometry and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF), as well as advanced automated microscopy techniques that allow the identification of bacterial species, the presence of antibiotic resistance genes, and bacterial killing by specific antibiotics within 4 to 6 hours using direct specimen inoculation [26,27].

Our study has a number of limitations. As a retrospective cohort, it is prone to several forms of bias, most notably selection bias. We attempted to mitigate this by enroling consecutive patients fitting the predetermined enrolment criteria. Although we adjusted for known confounders, the possibility exists that some residual confounding remains, particularly confounding by indication. Another important limitation is the potential for patients to be enroled who did not have true pneumonia. Our use of clinical criteria along with microbiologic confirmation was an attempt to maximize the number of patients with Pa-NP in our cohort. Additionally, antimicrobial susceptibility testing was performed at the local hospital level. Therefore, the determination of MDR status may have varied more than if a single reference laboratory was used to determine the presence or absence of drug resistance. It is also important to note that although our results strongly suggest that the association of MDR status with increased risk of death is mechanistically related to the risk of receiving inappropriate empiric therapy, we cannot rule out that MDR Pa-NP may exert its lethal effect directly due to higher virulence, as has been suggested with other pathogens exhibiting higher minimum inhibitory concentration (MIC) to certain antimicrobials [28,29]. Because we examined hospital mortality rather than the more standard 28-day mortality as the primary outcome for our study, we may have overestimated the magnitude of this outcome. Last, individual antibiotics are commonly part of a regimen for the treatment of nosocomial pneumonia, therefore, independent analysis of the impact on AIAT may not represent the true prescribing practice at each site.


In summary, our study sheds light on variables associated with to MDR Pa-NP; namely decreasing age, diabetes mellitus and ICU admission. In addition, MDR status is an independent predictor of hospital mortality in patients with Pa-NP. Given the high rates of MDR Pa-NP, advances in rapid diagnosis and susceptibility analysis are needed to direct antibiotic treatment and potentially improve outcomes.

Key messages

  • Among patients with Pa-NP, presence of infection with MDR strains is an important independent predictor for hospital mortality.

  • Independent predictors of MDR strains of P. aeruginosa in this study included decreasing age, diabetes, and ICU admission.

  • Advances in rapid diagnostics and antibiotic susceptibility analysis are needed to direct antibiotic treatment and potentially improve outcomes of patients infected with MDR strains of P. aeruginosa.