Impact of Blood Cultures on the Changes of Treatment in Hospitalized Patients with Community-Acquired Pneumonia

Sample Selection

We retrospectively investigated patients admitted with CAP between January 1, 2008, and December 31, 2010. CAP was defined as the presence of symptoms of lower respiratory tract infection such as cough, sputum production, and dyspnea, along with infiltrate on the chest radiography or chest computed tomography images on admission.

To be included in this study, patients had to be 18 years or older and from whom 2 sets of BCs (2 cultures bottles [one aerobic and one anaerobic] drawn at two different times) had been obtained before starting antimicrobial agents on admission. We included patients from nursing facilities. By contrast, we excluded patients on immunosuppressant therapy (steroid therapy, chemotherapy for malignant diseases, disease-modifying anti-rheumatic drug therapy, anti-cytokine therapy), and with human immunodeficiency virus infection (HIV) as defined by the Centers for Disease Control and Prevention [4]. We also excluded patients with history of admission or antimicrobial agent use in the past 30 days.

Data Collection

Each patient’s medical record was obtained through electronic data collection. Most of the clinical variables were derived from the PSI of Fine et al. [5]. Additionally, we gathered information about the patients’ comorbidities, which were defined on the basis of documented histories from their admission summaries. We only considered the results of the initial 2 sets of BCs obtained on admission, and not the results of BCs obtained after admission.

Definitive diagnosis of illnesses after admission was made based on the results of cultures and clinical evaluation by treating doctors. The presence of bacterial endocarditis was determined based on the DUKE criteria [6]. In patient characteristics, altered mental status is not a part of the definition of “cerebrovascular disease.” Specimens were considered contaminated if only bacteria from normal skin flora were detected and treating doctors evaluated the result as contamination.

A change in antimicrobial agent management was defined as any change in the antimicrobial agent itself, or the addition or termination of agents. The rationales for change were obtained from the documentation on the patient charts recorded by the treating doctors.

The choice of initial antimicrobial agents for CAP was mainly based on the Japanese Respiratory Society Guidelines for the Management of Community-Acquired Pneumonia in Adults (the second edition, 2005). In the guidelines, empiric use of penicillin derivatives with the beta-lactamase inhibitors, Piperacillin, and cephalosporins is recommended for inpatient settings without respiratory illnesses. In case of patients with respiratory diseases such as chronic obstructive pulmonary disease, use of carbapenem or fluoroquinolones is warranted. Furthermore, in suspicion of atypical pneumonia, adding tetracyclines, macrolides, or fluoroquinolones is also recommended.

Statistical Analysis

Statistical calculations were performed using InStat Statistical Software Package Version 3.01 (GraphPad Software Inc., CA). Variables are presented as mean ± standard deviation unless otherwise stated. We used χ² analysis and the t-test for comparison between 2 groups. P < 0.05 was considered statistically significant.

Patients’ Characteristics

Of the 793 patients admitted with initial diagnosis of CAP, 394 were excluded (immunosuppressant use, 51; HIV infection, 4; admission 30 days prior, 141; antimicrobial agents use 30 days prior, 192; and BCs not performed, 6). The details are illustrated in Fig. (1). The remaining 399 patients (232 men and 167 women) had a mean age of 78.4 ± 14.6 years. In total, 97 patients (24%) had been admitted from nursing homes. Clinical variables as well as comorbid illnesses have also been summarized (Table 1). The Streptococcus pneumoniae Urine Antigen Test (BinaxNOW S. pneumoniae; Alere, MA) was positive in 55 out of 344 patients, while the Legionella Urine Antigen Test (BinaxNOW Legionella; Alere, MA) was positive in 2 out of 316 patients. The most common vital sign abnormality at admission was respiratory failure (PaO₂ < 60 mmHg or SpO₂ < 90%), followed by altered mental status. High blood urea nitrogen (BUN) level was the most common laboratory abnormality. Furthermore, the most frequent comorbid diseases were cerebrovascular disease (23%) and congestive heart failure (23%). The number of PSI grade V patients was the highest, followed by patients with PSI grade IV, and 30 (7.5%) required intensive care. The most used empiric antimicrobial agent was Ampicillin/sulbactam, followed by Ceftriaxone and Piperacillin/tazobactam.

Table 1..

Table 2..

Table 3..

Table 4..

Table 5..

The bacteria detected from the sputum cultures (Table 2) show that most common pathogen was Streptococcus pneumoniae, followed by Haemophilus influenza and Streptococcus agalactiae.

Among patients who fulfilled the inclusion criteria (n=399), 386 were made definitive diagnosis of CAP after admission (the remaining 13 were made alternative diagnosis [non-pneumonia illnesses] based on the results of cultures and the evaluation by treating doctors after admission).

The Rate of Bacteremia and Frequency of Antimicrobial Agent Change

The bacteria detected from the BCs (Table 3) show that most common pathogen was Staphylococcus epidermidis, which is commonly considered the causative bacteria of contamination. Streptococcus pneumoniae was second most common bacterium detected from the BCs, followed by Escherichia coli. Fig. (1) demonstrates that the number of pneumonia causing positive BCs was 17 (4.4 %) out of 386 patients, whereas other etiologies (alternative diagnosis) yielded positive BCs in 10 patients (bacteremia from unknown origin: 6, urinary tract infection: 3, infectious endocarditis: 1). 12 patients were evaluated as contamination by treating doctors even though the results were positive.

Among the 17 patients with positive BCs in CAP diagnosis, 8 (Pneumonia Severity Index [PSI] grade IV; 2, PSI grade V; 6) changed antimicrobial agents based on the positive BC results (Figs. 1, 2). Therefore, 2.1% out of 386 patients with definitive CAP diagnosis (0%, 0%, 0%, 1.6%, and 4.8% out of PSI grade I patients (7), grade II patients (19), grade III patients (53), grade IV patients (122), and grade V patients (124), respectively) changed antimicrobial agents based on the positive blood culture results (Fig. 2).

Specifically, 5 patients changed the antimicrobial agent itself, 2 added new agents, and 1 terminated therapy based on the BC results (Fig. 1).

The specific diagnoses, bacteria and antibiotics in the patients who changed therapy based on positive blood culture results are shown in Table 4 (the rationale [comments] for changing antimicrobial agents was based on admission summaries for each patient). By contrast, no one changed therapy based on negative BC results.

We compared the clinical variables of the patients who changed antimicrobial agents based on the positive BC results (n = 17) with the patients who did not (n = 10) (Table 5). The former group had significantly higher frequency of alternative diagnosis after admission. Also the former group had better mortality than the latter group. However, we could not show a significant difference of severity (PSI score) between two groups. The specific empiric antimicrobial agents in both groups are also demonstrated in Table 5.