We studied prospectively all patients resident of long-term care facilities, aged 65 years and older, admitted from the emergency department to our intensive care unit between January 2000 and October 2002 with suspected severe pneumonia. The study was approved by the Institutional Review Board of the University at Buffalo. An informed consent was obtained from all participants or their health care proxy.

The inclusion criteria for severe aspiration pneumonia included: (1) the development of new radiographic infiltrate compatible with pneumonia; (2) the presence of symptoms or signs suggestive of lower respiratory tract infection (one major criteria of either cough, sputum production, or fever above 38°C or below 35.5°C, plus two minor criteria of pleuritic chest pain, dyspnea, delirium, increased alveolar arterial gradient, or white blood cell count > 12,000/mm³, and/or left shift or leukopenia < 3,000/mm³) necessitating mechanical ventilation; and (3) the presence of risk factors for oropharyngeal aspiration (documented swallowing abnormality secondary to neurologic dysphagia, disruption of the gastroesophageal junction, or anatomic abnormalities of the upper aerodigestive tract) (11). Patients with severe immunosuppression (solid organ transplantation, steroid therapy of 20 mg/day or more for more than 2 weeks, known HIV, or AIDS-defining criteria), or who had received antimicrobial therapy or were hospitalized within the last 30 days were excluded.

Demographic data collected included age, sex, Activity of Daily Living score (12), comorbid illnesses, the Charlson Index (13), and prior antimicrobial therapy. In addition, the following parameters were recorded at the time of enrollment: clinical symptoms, vital signs, laboratory and radiographic data, the Acute Physiology and Chronic Health Evaluation (APACHE II) score (14), and antibiotic coverage after the initial sampling was performed. Oral examination included the Plaque Index (15) and enumeration of the number of remaining teeth. In addition, the following parameters were recorded during the patients' daily follow up: vital signs, radiographic progression, and organ failure. At the clinical endpoints of hospital discharge or death, the length of mechanical ventilation and ICU stay were recorded.

Two serial blood cultures were performed in all patients at the time of enrollment. Urinary samples were assayed for Legionella pneumophila and pneumococcus antigens. Respiratory samples were obtained within the first 4 hours of presentation to the emergency department via protected bronchoalveolar lavage and processed as previously reported (16, 17).

Septic shock was defined according to Bone and colleagues (18). The etiology of pneumonia was considered verified if one of the following criteria is met: (1) positive urinary antigen for Legionella or pneumococcus; (2) bacterial growth in protected bronchoalveolar lavage ⩾ 10³ cfu/ml; or (3) positive respiratory assay for influenza A and B antigens. According to bacteriologic results, three groups were identified: an aerobic group corresponding to patients with pneumonia attributed to aerobic microorganisms, an anaerobic group involving patients with pneumonia who had a protected bronchoalveolar lavage specimen containing at least one anaerobic microorganism, and a group of patients with unverified pneumonia in whom the diagnostic workup did not reveal the presence of a microbial pathogen.

Antimicrobial therapy was considered adequate if the causative pathogen(s) was/were susceptible to the prescribed antibiotic(s) therapy on enrollment. The initial therapy was considered effective if the clinical situation remained stable, fever decreased, and leukocytosis improved after 72 hours of treatment. In all other cases, the initial therapy was considered a failure.

Data were analyzed using the NCSS 2000 statistical software (NCSS Statistical Analysis System, Kaysville, UT). Results were expressed as means ± SD. Among patients with verified pneumonia, aerobic and anaerobic groups were compared at baseline for epidemiologic and clinical data. Continuous variables were compared using Student's t test for normally distributed variables and the Mann-Whitney test for non-normally distributed variables. Proportions were compared using the chi-square test with Yates correction or Fisher's exact test when necessary. In the same way, such data were compared between patients exhibiting verified and unverified pneumonia. A stepwise logistic regression was performed using those variables found significant in univariate analysis with in-hospital mortality as the dependent variable. All reported p values are two-tailed. The level of significance was set at 5%.

During the study period, a total of 138 patients were eligible for enrollment. Forty-three patients could not be enrolled because of the lack of informed consent. Out of the 95 who underwent bronchial samplings, 54 participants had at least one microorganism isolated. Anaerobic bacteria were recovered from 11 (20%) of the 54 subjects. Characteristics of the study population are shown in Table 1

. All 95 participants had at least one comorbid illness, with 87% having two or more comorbidities. Stroke (78%) was the most prevalent comorbidity, followed by ischemic heart disease (51%) and chronic obstructive pulmonary disease (35%). The majority of patients presented with dyspnea, fever, and delirium. On physical examination, crepitations were the main auscultatory findings (n = 62), followed by rhonchi (n = 23) and wheezing (n = 5). A “silent” chest was present only in five subjects (5%).

Clinical and radiographic presentations are summarized in Table 2

. The severity of illness and the degree of hypoxemia were not significantly different between those with aerobic and anaerobic bacteria, or between those with verified and those with unverified pneumonia. Similarly, the distribution of radiographic infiltrates was similar among the groups identified. The involvement of the right lower lobe was no more frequent in the anaerobic group than in the aerobic one. No patient with an anaerobic strain had necrotizing pneumonia with an abscess formation or empyema. A computed tomography of the chest was performed on 18 of the 95 subjects. None showed evidence of lymphadenopathy or cavitation.

The mean number of teeth present and the Plaque Index were comparable across the various groups (Table 2). The percentage of patients who were edentates was more frequent, however, in those with aerobic compared with anaerobic isolates, although the difference did not attain statistical significance (18 of 43 versus 3 of 11; p = 0.5). Seven of those 18 edentates with aerobic bacteria had dentures, compared with 2 of 3 with anaerobic strains, and 7 of the 11 with unverified pneumonia.

Positive blood cultures were positive in three (3%) patients, and consisted of Staphylococcus aureus, Klebsiella pneumoniae, and Streptococcus pneumoniae. No anaerobic bacterium was isolated from blood cultures. Thoracocentesis was performed in six cases. In all six instances, the fluid analysis revealed an exudative effusion suggestive of parapneumonic process; however, no bacterial pathogen (whether aerobic or anaerobic) was recovered from cultures of the pleural fluid.

Overall, 67 pathogens were isolated from bronchial samplings (Table 3)

; 42 had one pathogen, 11 had two, and 1 had three pathogens. Gram-negative enteric bacilli were the predominant organisms isolated (49%), followed by anaerobic bacteria (16%) and S. aureus (12%). The majority of S. aureus (six of eight) was methicillin-sensitive. Atypical infections were identified in none of the participants. Table 4 depicts the antimicrobial susceptibility patterns of anaerobic isolates. Resistance to penicillin was detected in 27%. One isolate of Prevotella spp. was resistant to both penicillin and metronidazole.

Compared with the aerobic group, the burden of comorbidities was not found to be a determinant for the isolation of anaerobic bacteria (the median Charlson Index for aerobic and anaerobic infection was 3 and 3, respectively; p = 1.0). Similarly, the presence of percutaneous gastrotomy tube or prior pneumonia episodes was not indicative of an anaerobic aspiration. However, anaerobic pathogens were more likely to be recovered from patients with impaired functional status (p < 0.04) (Figure 1)

.

Polymicrobial infection was present in 12 (22%) of the 54 patients in whom a microbial etiology was determined (Table 5)

. Aerobic Gram-negative bacteria were associated with six anaerobic isolates. No clinical or radiographic patterns characterized mixed anaerobic from mixed aerobic cases.

In patients with verified pneumonia, initial empiric antimicrobial treatment consisted of either monotherapy (n = 22) or dual combination (n = 32). As a single agent, the administered drugs were levofloxacin (n = 10), ampicillin/sulbactam (n = 6), piperacillin/tazobactam (n = 5), and carbapenem (n = 1). The main drugs prescribed in combination included ceftriaxone plus azithromycin (n = 11), ampicillin/sulbactam plus azithromycin (n = 6), levofloxacin plus clindamycin or metronidazole (n = 5), piperacillin/tazobactam plus levofloxacin or azithromycin (n = 3), and vancomycin plus ciprofloxacin (n = 2). The initial antimicrobial coverage of anaerobic isolates is shown in Table 6

. Thirty-nine of the 43 patients with aerobic isolates received adequate antimicrobial therapy, compared with 4 of the 11 with anaerobic species (p < 0.001). Only one of the aerobic cases who received inadequate therapy had effective initial therapy. Vancomycin was added in two cases because of methicillin-resistant S. aureus infection, and one had a change of antimicrobial therapy to cover for Pseudomonas aeruginosa. In contrast, the initial therapy was considered effective in six of the seven anaerobic cases with inadequate anaerobic antimicrobial coverage. The only case identified as showing signs of ineffective therapy had an ampicillin-resistant K. pneumoniae isolated along with Fusobacterium spp. Imipenem was substituted for the initial antimicrobial regimen.

When adjusted for severity of illness and underlying comorbidities, there was no significant difference in the length of mechanical ventilation or ICU stay between the aerobic and the anaerobic group (6.3 ± 3.4 days and 9.5 ± 4.4 days for the aerobic group versus 7.6 ± 4.3 and 10.9 ± 4.5 for the anaerobic group, p = 0.7 and p = 0.6; respectively). Compared with patients with verified pneumonia, those with unverified pneumonia were ventilated for longer period of time (8.6 ± 4.8 versus 6.5 ± 3.6; p = 0.01) and had a prolonged ICU stay (11.7 ± 5.1 versus 9.8 ± 4.4; p = 0.06). Fourteen patients had a “do not resuscitate” order written on admission. Overall, 35 patients died during their hospitalization (14 in the aerobic group, 4 in the anaerobic group, and 17 of those with unverified pneumonia). The cause of mortality was attributed to progressive pneumonia in 24 patients and multiorgan failure in 11. Survivors were less likely to be functionally impaired (p = 0.01), nutritionally depleted (p < 0.001), or have significant comorbidities (p = 0.002). By multivariate analysis, only the Charlson Index (β coefficient 0.92 ± 0.22; p < 0.001) and serum albumin (β coefficient 2.22 ± 0.5; p < 0.001) were found to be independently associated with poor outcome.