Introduction

Hospital-acquired pneumonia is the second most frequent cause of hospital-acquired infection [1, 2, 3], accounting for 13–18% of nosocomial infections [4] and affecting 0.5–2.0% of hospitalized patients [5, 6]. The annual incidence is 5–10 cases per 1,000 admissions, but in ventilated patients the figure may be 20 times higher [1, 7, 8]. Hospital-acquired pneumonia is associated with the highest morbidity and mortality of all nosocomial infections; its crude mortality ranges from 30% to 70%, and its attributable mortality is as high as 33-50% [9, 10, 11, 12, 13]. The Study on the Efficacy of Nosocomial Infection Control (SENIC) and the National Nosocomial Infections Study (NNIS) investigations have found that hospital-acquired pneumonia occurs primarily in patients hospitalized for medical problems or recovering from abdominal or thoracic surgery [14, 15], and that about 50% of episodes occur in patients admitted to general wards [16]. In the ICU the incidence is higher than in wards, and most episodes are associated with the use of artificial airways (endotracheal tubes and tracheostomies); this particular subgroup of hospital-acquired pneumonia is known as ventilator-associated pneumonia (VAP) [4, 17, 18, 19]. A second group of patients with pneumonia in the ICU are those with severe episodes admitted from the community or hospital wards. An excellent study has demonstrated that the two groups have different causes and outcomes and require different treatment strategy [20]. Severe community-acquired pneumonia in ICU admissions has been carefully analyzed in recent years, and the cause and prognostic factors of this infection have been well established [21, 22, 23, 24, 25, 26]. On the other hand, information is lacking on episodes of severe hospital-acquired pneumonia (SHAP) requiring ICU admission.

The aims of this study were to determine the characteristics, prognostic factors, and outcome of hospitalized patients admitted to the ICU for SHAP at two different institutions.

Material and methods

Patients

This observational study was carried out in two medical-surgical ICUs, at the Hospital Sabadell (HS) and the Hospital Universitari Germans Trías i Pujol of Badalona (HGTP), with 16 and 20 beds, respectively. All 96 adults (73 men, 23 women; mean age 62±16 years) admitted during a 7-year period (January 1993–January 2000) with suspected hospital-acquired pneumonia requiring admission in the participating ICUs were included prospectively in the study. There were 56 episodes diagnosed in HS and 40 in HGTP, representing an incidence of 9.9 and 10.8 episodes per 1,000 ICU admissions, respectively. The patients' mean score on the Acute Physiology and Chronic Health Evaluation (APACHE) II at admission was 19±7. The most frequent causes of hospitalization were chronic respiratory diseases (27%), surgical conditions (17%), and cancer (17%). The incidence of chronic obstructive pulmonary disease (COPD) was significantly higher in admissions to HS than to HGTP. The underlying clinical conditions and risk factors prior to the development of pneumonia are shown in Table 1. Eleven patients presented leukopenia, six of which had neutropenia (<500 leukocytes/mm3), and only one patient had human immunodeficiency virus infection. The institutional ethics committees approved the study.

Table 1. Clinical characteristics of 96 patients with severe hospital-acquired pneumonia
Full size table

Definitions

Clinical suspicion of hospital-acquired pneumonia was based on a definition reported elsewhere [6]. It was essentially the same criteria as that developed for severe community-acquired pneumonia [27]. Crude mortality included all ICU deaths in the cohort of patients with SHAP on admission. Antimicrobial therapy was deemed appropriate when at least one antibiotic administered within 48 h of diagnosis of pneumonia was active in vitro on all organisms found associated with pneumonia; two active antimicrobials were required when Pseudomonas aeruginosa was involved. This was determined by an independent monitoring committee.

For each case the following data were collected: demographic characteristics, diagnosis on admission to hospital, surgical procedure prior to diagnosis of pneumonia, previous antibiotic treatment, APACHE II score at ICU admission, hospital stay prior to diagnosis of hospital-acquired pneumonia and prior to ICU admission, length of ICU stay, and chronic underlying diseases as described previously [28, 29, 30]. Diagnosis of COPD was based on a postbronchodilator forced expiratory volume in 1 s less than 80% of the predicted value in combination with a ratio of forced expiratory volume in 1 s to forced vital capacity less than 70%, consistent with recent published criteria [31]. Among COPD patients a predominant bronchitis or emphysema was considered according to clinical characteristics, alterations in ventilatory function, and chest radiographic examination. A hematological malignancy was considered to be present when results of the peripheral blood examination or the bone marrow or lymph node biopsy were consistent with this diagnosis. Nonhematological malignancies were considered to be present only when a histological diagnosis was available. Patients were considered to have received immunosuppressive therapies if they had taken steroid or cytotoxic therapies. High-dose steroid therapy was considered if they had taken more than 1 mg/kg prednisone daily (or an equivalent dosage) for at least 2 weeks, and short-term low-dose steroid therapy was considered if they had taken less than 1 mg/kg prednisone daily during less than 2 week before the episode of pneumonia. Cytotoxic therapy was considered if they had taken some kind of antineoplastic treatment for at least 4 weeks before the pneumonic episode. Organ transplantation patients and patients with positive serology for HIV were also considered immunocompromised. Patients were considered to have diabetes mellitus if they required prior insulin therapy. Leukopenia was considered if the total leukocyte count was lower than 4,000/µl and granulocytopenia if the absolute granulocyte count was lower than 500/µl. In addition, the systemic response and associated complications as a consequence of SHAP in the 72 h following the diagnosis of pneumonia were collected. Definitions of septic shock and acute respiratory distress syndrome have been reported elsewhere [30, 32]. Acute renal dysfunction was defined by a serum creatinine level higher than 2 mg/dl or a 50% reduction in previous creatinine clearance.

The cause of pneumonia was established when the causal pathogen was isolated from blood, pleural fluid, lower respiratory tract secretions, serological analysis, or by biopsy or postmortem lung tissue study. The diagnostic procedures were left to the discretion of the attending physicians. Definite or probable pulmonary aspergillosis was defined as described elsewhere [33]. If the patients were intubated, specimens for bacteriological studies, protected specimen brush (PSB) and bronchoalveolar lavage (BAL), were obtained under bronchoscopy guidance using a previously described technique [34], according with a systemic algorithm. When a bronchoscopic technique was not carried out, a tracheal aspirate was obtained. Quantitative cultures were considered positive when more than 1,000 CFU/ml were recovered from the PSB, more than 10,000 CFU/ml from the BAL, and more than 100,000 CFU/ml from the tracheal aspirate. In nonintubated patients lower respiratory secretions were obtained by expectoration. In these cases only good sputum samples (more than 25 neutrophils and less than 10 epithelial cells per microscopic field, ×100) were selected for the Gram stain and culture [35]. Legionella pneumophila pneumonia was diagnosed by isolation from respiratory secretions, by demonstration of a fourfold or greater rise in the reciprocal immunofluorescence antibody titer to greater than or equal to 128 against L. pneumophila serogroup 1 between paired acute- and convalescent-phase serum specimens or by demonstration of L. pneumophila serogroup 1 antigen in urine by enzyme-linked immunosorbent assay. Cytomegalovirus (CMV) infection was diagnosed by detection of CMV antigen pp65 in blood.

Statistical analysis

Descriptive analyses are presented performed. Continuous variables were expressed as mean (±SD). Associations of categorical variables with mortality are assessed by the χ2 test. Student's t test was used for continuous variables. In the multivariate analysis of patients' prognosis variables were selected for entry into the logistic regression model if at least ten of the patients exhibited the characteristic and if the variables were significantly associated with mortality at a p value less than 0.1 in the univariate analysis. Logistic regression was used for the estimation of coefficients and their standard errors. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using standard methods [36, 37].

Results

Clinical features of severe hospital-acquired pneumonia

The clinical condition of the SHAP episode which caused ICU admission was respiratory failure in 86 patients (90%), and septic shock or severe sepsis in 8 cases (8%); two patients (2%) were admitted to the ICU for mental confusion. Moreover, 80 patients required mechanical ventilation (83%), for whom emergency endotracheal intubation prior to ICU admission was needed in 12 (12%). The remaining 16 patients (17%) were treated with noninvasive mechanical ventilation or oxygen therapy using a conventional face mask. The mean ratio of PaO2/FIO2 at admission in the ICU was 155±85 mmHg, and 67% of patients had been treated previously with antibiotics. The mean hospital stay prior to pneumonia diagnosis and ICU admission was 20.2±19.8 and 20.6±19.7 days, respectively. The mean ICU stay was 15.7±21.7 days.

Microbiology of severe hospital-acquired pneumonia

Definitive microbiological diagnosis was obtained in 67 of the 96 patients (70%) admitted to the ICU with clinical criteria of SHAP. A total of 75 micro-organisms were isolated from 60 monomicrobial and 7 polymicrobial episodes. Gram-negative micro-organisms were isolated in 51% of diagnosed episodes, Gram-positive in 33%, fungi in 19%, and virus in 3% of cases. The organisms most frequently isolated from patients with SHAP were P. aeruginosa (24%), followed by Aspergillus spp. (17%), Streptococcus pneumoniae (15%), Staphylococcus aureus (12%), and L. pneumophila (12%) (Table 2).

Table 2. Microbiological findings in 67 episodes of microbiological diagnosed SHAP; most patients presented simultaneously positive results with more than one diagnostic technique [PB protected specimen brush or bronchoalveolar lavage, B blood culture, S sputum, Pl pleura, H histological sample (necropsy or biopsy), SR serology, UAg urinary antigen, BAg CMV antigen in blood]
Full size table

Among 13 episodes of pneumonia due to Aspergillus spp. eight cases were classified as definite pneumonia by histological diagnosis (one by biopsy and seven by postmortem pathology findings) and five cases as probable pneumonia by microbiological diagnosis in BAL samples plus thoracic computed tomography. Eleven patients with aspergillosis (84%) had COPD, nine patients (69%) were taking steroids, and only three patients had diagnoses of hematological malignancies. Among the nine patients with nosocomial legionellosis five were taking steroids, two had COPD, and two were patients with malignancies. The diagnosis of legionellosis was obtained in one case by demonstration of L. pneumophila serogroup 1 antigen in urine plus serological assay, and in the remaining eight patients by isolation from respiratory secretions plus serological assay.

Polymicrobial episodes were diagnosed in seven patients, and L. pneumophila was the micro-organism involved in five cases. With the exception of a comatose patient secondary to stroke (L. pneumophila associated with S. pneumoniae and H. influenzae), polymicrobial infections of L. pneumophila were diagnosed in immunocompromised patients: two patients with steroid therapy (associated with P. aeruginosa and Pneumocystis carinii infections, respectively), one transplanted patient (associated with CMV infection) and a leukemic patient (associated with Aspergillus spp. infection). In the remaining two polymicrobial episodes, Haemophilus influenzae and Serratia marcescens were isolated in a postoperative period of patient with esophageal cancer and Aspergillus spp. and S. aureus in a patient with steroid therapy (Table 3). The overall incidence of Gram-positive and Gram-negative organisms was similar in the two ICUs, but the incidence of individual pathogens differed: in HS a trend towards a higher incidence of Aspergillus spp. (p=0.06) was found, while in HGTP L. pneumophila was the most frequent pathogen causing pneumonia (p=0.01). Ninety-two episodes (96%) were considered as late-onset pneumonia, following the American Thoracic Society guidelines, and only four episodes developed within the four first days of hospital admission. Pneumonia due to multiresistant micro-organisms (P. aeruginosa, A. baumannii, methicillin-resistant S. aureus and Enterobacter spp.), L. pneumophila and Aspergillus spp. were considered as late-onset pneumonia in 96% of episodes, and 77% of them had been previously treated with antibiotics.

Table 3. Microbiological findings in seven polymicrobial episodes of microbiological diagnosed SHAP [PB protected specimen brush or bronchoalveolar lavage, H histological sample (necropsy or biopsy), BAg CMV blood antigen]
Full size table

Among 67 episodes with microbiological diagnosis, in 41 cases (61%) a bronchoscopic technique such as PSB or BAL was positive, and in 30 cases (45%) the pathogen was isolated in sputum or quantitative tracheal aspirate; 13 episodes (19%) were associated with bacteremia (including two patients with a CMV infection and positive CMV antigen in blood), nine patients (13%) had presented a positive serological test, and five pleural cultures (7%) were positive (three for P. aeruginosa, and one each for S. pneumoniae and S. aureus, respectively). In eight cases caused by Aspergillus spp. a histological diagnosis was obtained. Most patients presented simultaneously positive results with more than one diagnostic technique. The empirical antibiotic treatment was modified (changed or streamlined) in 64% of cases based on microbiological findings. In 31 of 92 cases with late onset pneumonia initial therapy was considered inappropriate.

Clinical outcome and complications

Complications associated with SHAP at admission or during ICU stay were: acute respiratory distress syndrome in 34 cases (35%), shock in 49 cases (51%), acute renal failure in 34 patients (35%), and empyema in 5 cases (5%). Fifty-one patients died (mortality rate, 53%), and most patients (59%) died during the first week of ICU stay. Among episodes with causal diagnosis the highest mortality rate was associated with pulmonary aspergillosis (77%), and pneumonia due to P. aeruginosa (55%). The mortality rate was 82% in patients with shock and 43% in bacteremic patients. The episodes without microbiological diagnosis showed a mortality rate of 45%. Among late onset episodes with inappropriate initial antibiotic treatment, the mortality rate was 71%.

An univariate analysis identified five factors associated with mortality (Table 4). These variables plus the severity-of-illness at admission (APACHE II) were included in a logistic regression model, death in the ICU being the dependent variable. The final model identified the presence of septic shock (OR 14.27, 95% CI 3.53–59.88) and chronic obstructive pulmonary disease (OR 6.11, 95% CI 1.79–27.7) as independent factors associated with death.

Table 4. Prognostic factors associated with mortality (univariate analysis)
Full size table

Discussion

This study is unique in analyzing the epidemiology, cause, and outcome of patients with SHAP acquired outside the ICU. Our findings suggest that incorporating information on variations in local epidemiological data may allow more accurate empirical therapy. Knowledge of these local patterns can aid clinicians in their decision-making processes, using a patient-based approach resulting in improved quality of care.

The SENIC and NNIS [14, 15] investigations have reported that hospital-acquired pneumonia occurs primarily in patients hospitalized with medical problems or recovering from abdominal or thoracic surgery. Our results in this specific population showed that most episodes occurred in medical patients with multiple risk factors, chronic pulmonary diseases and steroid therapy being the most important underlying conditions prior to the development of pneumonia. Previous studies have identified chronic pulmonary disease as one of the most important preexisting conditions that impair host defenses and may increase the risk of hospital-acquired pneumonia fourfold, probably due to the impairment of mucociliary clearance [1, 4, 8]. Eight patients had emphysema and 36 chronic bronchitis in our study population. Steroids and other immunosuppressive agents may be prescribed to these patients and increase the risk for pneumonia by opportunistic pathogens by impairment of cell-mediated immunity, neutrophil and macrophage function [38, 39, 40, 41]. Other patient-related factors identified as risk factors for hospital-acquired pneumonia such as prolonged hospitalization, previous surgery, diabetes mellitus, and malignancies were also present in most of our patients. These findings demonstrate that most patients with SHAP present one or more previous comorbidities, which may influence the presence of specific pathogens and the severity of the pneumonia episode.

Our results show that Gram-negative microorganisms are the most frequent pathogens causing of SHAP, and that they are predominantly late-onset episodes. All patients had prior risk factors or comorbidities. Most of our patients received a combination empirical therapy consistent with the American Thoracic Society guidelines [6], but modifications according the patterns of antimicrobial resistance own of each hospital and addition of specific antimicrobial therapy according to the presence of specific micro-organisms during the study period was needed. Indeed, in 31 of 92 late onset pneumonia (34%) the initial antibiotic treatment was inappropriate, and P. aeruginosa was the most frequent microorganisms implicated in these episodes.

In our results, the overall importance of some organisms may be controversial. Legionella was discovered as cause of outbreak-related pneumonia in the hospital setting in 1978. This problem occurred when over a course of 4 years with hospital-acquired Legionnaires' disease involving over 300 patients was discovered in United States [42, 43]. A hospital water supply harbored the bacteria and was the source of infection [42, 44]. These outbreaks were, in retrospect, endemic cases of hospital-acquired Legionnaires' disease. Our findings are similar and support the concept of therapy decision making based on local epidemiological data. Moreover, our findings demonstrate that global antibiotic treatment recommendations run an enormous risk of failure, and indicated that suggest information should be compiled on local epidemiology and characteristics patterns of organisms in each specific hospital.

Incidence of bacteremia was low. Incidence of S. aureus bloodstream infection in patients with hospital-acquired pneumonia has been reported to be lower than 20% [45, 46, 47]. This may raise some concerns on the final pathogen in some cases. For example, the recovery of S. aureus from the blood in insulin-regimen diabetic with S. aureus colonization in the sputum does not verify the cause of pneumonia. Thus we cannot ruled out that some of the isolates were contaminants or simply colonized patients. This is particularly true for episodes due to polymicrobial flora. Quantitative techniques and cytological examination were performed to separate colonization from infection. Indeed, samples showing more than 10% epithelial cells (sputum) or more than 1% epithelial cells (bronchoscopic samples) were excluded as potential contaminated samples. The use of quantitative invasive diagnostic techniques such as bronchoscopy (with either PSB and/or BAL) to retrieve lower respiratory tract secretions in the initial workup of patients with suspected hospital-acquired pneumonia remains controversial. Our findings suggest that in two-thirds of episodes with causal diagnosis the microbiological results allowed us to streamline antibiotic treatment or modify it with other antimicrobial drugs because the empirical treatment was resistant. These results suggest that the use of microbiological work up in the setting of SHAP identifies the cause in most cases and allows the prescription of a more specific therapy.

The crude mortality for hospital-acquired pneumonia is reported to range from 30% to 70%, but a number of patients succumb to their underlying disease, with pneumonia being a terminal event. Some studies estimate that attributable mortality for hospital-acquired pneumonia is one-third to one-half of all deaths, and it may be higher among bacteremic patients and those infected with certain microorganisms [9, 10, 11, 12, 13]. Our results confirm that patients with SHAP present high crude mortality, and specific pathogens, such as P. aeruginosa, showed a mortality rate higher than 50%. In this specific group of patients with SHAP a severe systemic response associated with the pulmonary infection is often documented. The severity of systemic response has also been considered as an important prognostic factor in the outcome of severe community-acquired pneumonia admitted to the ICU [48, 49]. However, additional concern on whether shock was due to sepsis or to concomitant comorbidities may be argued. Moreover, recent studies [50] suggest that some patients refractory to vasopressors may benefit from low dose steroids. Our findings confirm that presence of septic shock was independently associated with a poor prognostic. Further studies should clarify the role of unsuspected adrenal failure in pneumonia patients develo** shock.

Our study has both strengths and potential limitations. For instance, results from only two teaching hospitals were included. This limits our ability to generalize the findings to hospitals without teaching affiliation. The microbiology of pneumonia reported was largely based on invasive diagnostic techniques; this may not be the case in other hospitals. The specificity of the diagnosis of pneumonia would be expected to be different in other study populations, such trauma or cancer patients or children, and the risk of potential contaminants in some cases have been discussed in detail, previously. Therefore the findings of this paper may not be applicable to all clinical settings. The contribution of pneumonia to shock in patients with prior comorbidities may be a potential confounder. Finally, the contribution of potential outbreaks may have affected the cause, and specific pathogens, such as Legionella and Aspergillus, should not be generalized to many institutions. However, Aspergillus isolates were documented over all the study period in HS, two cases in 1993 (April and December), one case in 1994 (April), two cases in 1997 (January and February), five cases in 1998 [April (2), March, September, November], and one case in 1999 (April). In HGTP the nosocomial infection by L. pneumophila was an endemic situation during all study period. Furthermore, in our study temporal distribution of cases did not suggest an epidemic situation. In spite of these limitations the implications of our findings for therapeutic decisions remain valid.

In summary, our results show that patients with SHAP admitted to the ICU present high mortality. The severity of the systemic response and previous comorbidities such as chronic obstructive pulmonary disease in conjunction with specific micro-organisms were associated with a poor prognosis. Our findings highlight the need to incorporate information of local patterns of epidemiology into the general framework of the guidelines.