Characteristics of Bacteremia Between Community-Acquired and Nosocomial Klebsiella pneumoniae Infection: Risk Factor for Mortality and the Impact of Capsular Serotypes as a Herald for Community-Acquired Infection

Background  Although several epidemiological surveys of Klebsiella clinical isolates have been performed, few studies have correlated the clinical isolate with disease.

Objective  To compare the clinical and bacteriological characteristics of Klebsiella pneumoniae bacteremia acquired as community or nosocomial infections.

Methods  We prospectively enrolled 158 consecutively hospitalized patients with K pneumoniae bacteremia. Clinical data were reviewed. Antimicrobial susceptibility testing and capsular serotyping were performed. We used the χ² test, the Fisher exact test, or the t test for statistic analysis.

Results  Underlying diabetes mellitus was more common in community-acquired than in nosocomial infection (46/94 [49%] vs 8/64 [12%]; P<.001). On the other hand, neoplastic disease (34/64 [53%] vs 13/94 [14%]; P<.001) and antibiotic resistance (P<.01) were more frequent in patients with nosocomial compared with community-acquired infections. Klebsiella pneumoniae liver abscesses, which were all community acquired, accounted for the source of 22 (23%) of 94 community-acquired K pneumoniae infections. No attributable source of infection was found for 37 (58%) of the 64 nosocomial infections vs 15 (16%) of the 94 community-acquired infections. Only 58 isolates (36.7%) could be serotyped; of these, capsular serotypes K1, K2, and K28 accounted for 37 (23.4%), 8 (5.1%), and 6 (3.8%), respectively, of all strains. However, typeable isolates were significantly more common among community-acquired than nosocomial isolates (42/94 [45%] vs 16/64 [25%]; P = .01), especially for serotype K1 (28/94 [30%] vs 9/64 [14%]; P = .02). Significant risk factors for mortality included nosocomial infection, lung infection, thrombocytopenia, leukopenia, ceftazidime resistance, inappropriate antimicrobial therapy, and septic shock.

Conclusions  Significant differences were identified between community-acquired and nosocomial K pneumoniae bacteremia. Ceftazidime resistance in nosocomial K pneumoniae bacteremia carried a high risk for mortality, and serotype K1 in K pneumoniae was more prevalent in community-acquired infection, suggesting more virulence.

KLEBSIELLA PNEUMONIAE causes suppurative infections, bacteremia, and a substantial percentage of nosocomal infections. It is the second most common cause of gram-negative bacteremia.^1-3 Strains of K pneumoniae that are resistant to multiple antibiotics, including the newer cephalosporins, have developed.⁴ Infections caused by these extended-spectrum β-lactamase–producing K pneumoniae are frequently epidemic in nature and have complicated chemotherapy significantly.^4-6 Most clinical isolates of K pneumoniae possess a well-defined polysaccharide capsule that appears to be a critical virulence factor.^7,8 Although several seroepidemiological surveys of Klebsiella clinical isolates have been performed,^9-12 few studies have correlated the clinical isolate with disease.^9,10 Some capsular serotypes are isolated at significantly higher frequency than others.^12,13 We conducted this study to compare the clinical and bacteriological characteristics of K pneumoniae bacteremia acquired as community or nosocomial infection.

From April 1, 1997, through March 31, 1999, patients with a blood culture positive for K pneumoniae at the Tri-Service General Hospital, a 1300-bed university hospital in Taipei, Taiwan, were prospectively enrolled. The clinical features of these patients and the bacteriological characteristics of these K pneumoniae strains were collected and analyzed.

Klebsiella pneumoniae bacteremia was defined as the isolation of K pneumoniae in 1 or more blood cultures in association with the clinical features of bacteremia and/or sepsis. Bacteremia was considered to be nosocomial if a blood culture that yielded positive findings was obtained later than 72 hours after admission, and if the bacteremia was not obviously associated with the clinical conditions of the patient at the time of admission. Otherwise, bacteremia was considered to be community acquired.

Previous corticosteroid treatment was defined as corticosteroid therapy for at least 2 weeks before the onset of the K pneumoniae bacteremia. Recent surgery was defined as a surgical procedure that had been performed within 1 month of the onset of K pneumoniae bacteremia. The source of bacteremia was defined clinically or by the isolation of K pneumoniae from the site of entry or both. An infiltrate had to be seen on the chest radiograph to substantiate the presence of pneumonia.

Shock was defined as a decrease in systolic blood pressure to a level of less than 90 mm Hg or a decrease of at least 40 mm Hg below baseline blood pressure, despite adequate fluid resuscitation, in conjunction with organ dysfunction and perfusion abnormalities (eg, lactic acidosis, oliguria, obtundation). Jaundice was defined as a total serum bilirubin level of at least 2.0 mg/dL (≥34.2 µmol/L).

Leukocytosis was defined as a white blood cell count of greater than 10 000/µL; leukopenia, a white blood cell count of less than 4000/µL; and thrombocytopenia, a platelet count of less than 100 000/µL.

Antimicrobial therapy was considered to be appropriate if 1 or more antimicrobial agents were active in vitro against the corresponding isolate when the agent(s) was given at an adequate dosage via an appropriate route for at least 2 days within 3 days of obtaining a blood culture that yielded positive results. Mortality was considered to be related to bacteremia if the patient was being treated for bacteremia when death occurred, unless clinical and pathological data clearly suggested otherwise.

All isolates were identified by means of the API 20E system (bioMérieux SA, Marcy I'Etoile, France). Antimicrobial susceptibilities were determined by means of the Kirby-Bauer disk-diffusion test on Mueller-Hinton agar (BBL Microbiologic System, Cockeysville, Md) for the following antibiotics: ampicillin (10 µg), amikacin (30 µg), aztreonam (30 µg), cefazolin (30 µg), cefuroxime (30 µg), ceftriaxone (30 µg), ceftazidime (30 µg), ciprofloxacin (30 µg), gentamicin (10 µg), imipenem (30 µg), piperacillin (100 µg), and a combination of trimethoprim and sulfamethoxazole (1.25 µg and 23.75 µg, respectively). Interpretations were performed according to the guidelines of the National Committee for Clinical Laboratory Standards.¹⁴

The isolates were serotyped by means of the countercurrent immunoelectrophoresis method.¹⁵ Antisera were provided by the Laboratory of Hospital Infection, Central Public Health Laboratory, London, England. In brief, strains to be serotyped were grown overnight on MacConkey agar (Difco Laboratories, Detroit, Mich), and a light suspension of bacteria was resuspended in isotonic sodium chloride solution with different antisera. Each isolate was tested against 14 antiserum pools by means of countercurrent immunoelectrophoresis and then tested again with the specific antisera of the reacting pool. Isolates were examined by means of the quellung (capsular swelling) reaction with antisera if cross-reactions occurred or if equivocal results were obtained by means of countercurrent immunoelectrophoresis. Isolates that failed to react in both methods were considered nontypeable. Klebsiella pneumoniae ATCC 9997 (K2, where K indicates capsular serotype) was used as a control strain.

The severity of illness was evaluated by means of the APACHE III (Acute Physiology and Chronic Health Evaluation III) score system.¹⁶

We performed statistical analysis using the χ² test, the Fisher exact test, or the t test. When the P value was less than .05, the difference was considered statistically significant.

During the 2-year study, 158 patients with an episode of K pneumoniae bacteremia were prospectively enrolled. Ninety-four episodes (59.5%) were community acquired, and 64 (40.5%) were nosocomial in origin (Table 1). The ages of patients ranged from 25 weeks to 90 years (mean ± SD age, 60 ± 19 years). The male-female ratio was 1.4:1.

The common underlying diseases associated with community-acquired K pneumoniae bacteremia were diabetes mellitus (n = 46 [49%]), hepatobiliary disease (n = 29 [31%]), and neoplastic diseases (n = 13 [14%]). The common underlying diseases associated with nosocomial K pneumoniae bacteremia were neoplastic diseases (n = 34 [53%]), hepatobiliary diseases (n = 14 [22%]), and diabetes mellitus (n = 8 [12%]). Underlying diabetes mellitus was more common in community-acquired than in nosocomial infection (49% vs 12%; P<.001). Neoplastic disease was more frequently noted in patients with nosocomial compared with community-acquired infection (53% vs 14%; P<.001; Table 1).

In the nosocomial K pneumoniae bacteremia group, 15 patients (23%) had indwelling urinary catheters or had undergone recent urinary manipulations. Eleven patients (17%) were receiving immunosuppressive therapy, including chemotherapy for neoplastic diseases (n = 9) and corticosteroid treatment (n = 2). Five of 11 patients with bacteremia who required respiratory assistance had pneumonia. Of the 37 patients with an unknown source of nosocomial bacteremia, 5 patients had a central venous catheter in place, 2 were receiving corticosteroid therapy, 6 were receiving chemotherapy, 2 had a central venous catheter and had undergone recent abdominal surgery, and 1 had undergone recent abdominal surgery. All of these predisposing factors were found more frequently in patients with nosocomial bacteremia (P = .01; Table 2).

For infection origin, intra-abdominal (P<.001) and urinary tract (P = .01) infections were significantly higher in community-acquired bacteremia. On the contrary, patients with an unknown origin of infection had a significantly higher incidence of nosocomial than community-acquired bacteremia (P<.001; Table 2).

Shock developed in 34 (21.5%) of 158 patients. Twenty patients (12.7%) displayed leukopenia. Of these, 10 (50.0%) had received chemotherapy for underlying hematologic neoplasms. Of the patients with leukopenia, 2 had undergone radiotherapy, 4 had cirrhosis of liver, 1 had aplastic anemia, and 3 had no underlying condition. One hundred eight patients (68.4%) showed leukocytosis. Sixty patients (38.0%) showed thrombocytopenia. Thirty patients (19.0%) had jaundice. The APACHE III score was higher in patients with nosocomial infections compared with those with community-acquired K pneumoniae bacteremia (mean ± SD score, 41 ± 24 vs 33 ± 21; P = .03). The overall mortality rate was 22.8% (36/158 patients). The mortality rate was higher in patients with nosocomial infections compared with those with community-acquired infections (36% vs 14%; P = .001) (Table 2). No significant differences between community-acquired and nosocomial bacteremia were observed in corticosteroid use, lung infection, leukocytosis, thrombocytopenia, creatinine levels of at least 2 mg/dL (≥176.8 µmol/L), ceftazidime resistance, inappropriate antibiotic therapy, metastatic foci, and shock.

Of the 158 patients, 21 (13.3%) were identified as having polymicrobial bacteremia. The concomitant organisms were Enterococcus (n = 7), Enterobacter cloacae (n = 4), Proteus mirabilis (n = 3), Escherichia coli (n = 3), methicillin-sensitive Staphylococcus aureus (n = 2), Pseudomonas aeruginosa (n = 1), Serratia marcescens (n = 1), Bacteroides species (n = 1), Streptococcus pneumoniae (n = 1), Enterobacter gergoviae (n = 1), and Citrobacter freundii (n = 1). The underlying disease in these patients included malignancies (n = 8 [38%]), cholangitis complicating hepatobiliary calculus (n = 5 [24%]), and urinary tract infection in debilitated conditions (n = 5 [24%]). In these patients with polymicrobial infections, the 8 patients with malignancy acquired the infection in the hospital, whereas the 5 patients with urinary tract infection acquired it in the community.

The antimicrobial susceptibility of the bacteremic isolates of K pneumoniae is shown in Table 3. Community-acquired isolates showed the highest frequency of resistance to trimethoprim-sulfamethoxazole (n = 22 [23%]), followed by piperacillin (n = 16 [17%]), cefazolin (n = 13 [14%]), and gentamicin (n = 10 [11%]). Resistance to third-generation cephalosporins was low (<5%). Among hospital-acquired isolates, resistance to trimethoprim sulfamethoxazole or piperacillin was most common (n = 28 [44%]), followed by cefazolin (n = 23 [36%]), gentamicin (n = 18 [28%]), cefuroxime (n = 16 [25%]), aztreonam (n = 16 [25%]), and amikacin (n = 11 [17%]). Resistance to ceftazidime and ceftriaxone was found in 14 (22%) and 11 patients (17%), respectively. The resistance of hospital-acquired isolates was significantly higher than that of the community-acquired isolates (P<.005), except imipenem. Nosocomial isolates also had a higher overall frequency of resistance to multiple antibiotics.

Of the 158 strains of K pneumoniae, 58 (36.7%) could be serotyped, and 9 different serotypes were identified (Table 4). The most common were serotypes K1, K2, and K28, and these accounted for 37 (23.4%), 8 (5.1%), and 6 (3.8%) strains, respectively (Table 4). Typeable isolates were significantly more common among the community-acquired isolates than among nosocomial isolates (42/94 [45%] vs 16/64 [25%]; P = .01). Isolation of the K1 serotype in community-acquired infection was significantly higher than in nosocomial infection (28/94 [30%] vs 9/64 [14%]; P = .02). On the contrary, nontypeable K pneumoniae was significantly higher in nosocomial infection (48/64 [75%] vs 52/94 [55%]; P = .01; Table 4).

The sources of infection and capsular serotypes of K pneumoniae are shown in Table 5. Most of the bacteremic K pneumoniae isolates were of a nontypeable serotype when the infection source was the lung, the urinary tract, or unknown. In contrast, most of the K pneumoniae isolates had an identifiable serotype when the infection source was in the abdomen. This higher rate of typeable isolates for intra-abdominal infection compared with other sources of infection was statistically significant (P = .001). Klebsiella pneumoniae liver abscesses, which were all community acquired, accounted for 23% of the source of community-acquired K pneumoniae bacteremia. Infection caused by serotypes K1 and K2 was statistically significantly higher in liver abscess compared with other sources of infection (P = .02). Serotype K1 isolate from unknown infection sources was more common in the community-acquired infections than in the nosocomial infections at a statistically significant level (7/15 [47%] vs 4/37 [11%]; P = .02). The presence of serotype K1 isolates, however, was not correlated with the presence of underlying diabetes mellitus.

Nosocomial bacteremia, ceftazidime resistance, leukopenia, and higher APACHE III scores were identified as risk factors for higher mortality (P<.005; Table 6). Although no significant difference between community-acquired and nosocomial bacteremia were observed in lung infection, thrombocytopenia, inappropriate antibiotic therapy, and septic shock, these variables were all identified as risk factors associated with significantly higher mortality (P<.05; Table 6).

Most of the patients in this study were elderly (mean ± SD age, 60 ± 19 years), with a slight male predominance (59%) in the group. The rate of hospital-acquired K pneumoniae bacteremia was 40.5%, which was lower than in studies during the past 28 years,^1,6,17-25 except for a study from Singapore.²⁶

Diabetes mellitus was the most commonly associated condition in patients with community-acquired K pneumoniae bacteremia, and the percentage of patients with diabetes (49%) was higher than in previous studies.^1,6,17-26 The association of diabetes mellitus and K pneumoniae liver abscess in Taiwan was reported previously in studies from our institution.^27,28 Neoplastic diseases (53%) were the most commonly associated condition in patients with nosocomial K pneumoniae bacteremia. Longer hospitalization, invasive procedures, chemotherapy, and the use of antibiotics are common in patients with neoplastic diseases. A higher rate of fecal colonization by K pneumoniae has been noted in patients with neoplastic diseases than in control subjects, and this has been associated with antibiotic use.²⁹

Fever occurred at a statistically significant lower rate in patients with nosocomial bacteremia than those with community-acquired bacteremia (53/64 [83%] vs 88/94 [94%]; P = .03). A higher percentage of neoplastic diseases, uremia, and corticosteroid use in patients with nosocomial bacteremia may be responsible. There was a higher percentage of leukopenia in nosocomial vs community-acquired bacteremia (14/64 [22%] vs 6/94 [6%]; P = .004), and this could have been due to a higher rate of chemotherapy for neoplastic diseases in the patients with nosocomial bacteremia. Thrombocytopenia, leukocytosis, shock, and polymicrobial causes were not significantly different between nosocomial and community-acquired bacteremia.

The nosocomial isolates were significantly more resistant to the antimicrobial agents that were tested, except for imipenem, when compared with the community-acquired isolates. More than 20% of the nosocomial strains were resistant to one of the third-generation cephalosporins, which raises a concern over an increasing prevalence of extended-spectrum β-lactamase (ESBL)–producing K pneumoniae, particularly in hospitals. The ESBL-producing K pneumoniae infections are a risk factor associated with treatment failure³⁰; therefore, aggressive infection control, including isolation of patients infected with ESBL-producing K pneumoniae and the imposition of restrictions on the use of third-generation cephalosporin, should be implemented.

A source of infection was identifiable in 67.1% of cases. The most common source was a urinary tract infection (20.9%), followed by liver abscess (13.9%). This finding is similar to those of a study performed in Singapore,²⁵ but has not been reported elsewhere for other Klebsiella bacteremia studies. In our study, of the 22 patients with liver abscess, a metastatic infection developed in 3 (one had endophthalmitis, a brain abscess, and septic pulmonary emboli; endophthalmitis developed in another; and the last one had meningitis). Metastatic infection is a characteristic feature of K pneumoniae liver abscesses in Taiwan.^27,31-33 Bacteremia with an unknown source of infection was more commonly associated with nosocomial than with community-acquired infections (58% vs 16%; P<.001). Many patients in whom nosocomial bacteremia develops have been subjected to invasive procedures such as urinary catheterization or manipulation, respiratory assistance, central venous catheterization, and recent surgery, which suggests that the procedure itself may be responsible for the bacteremia.

The seroepidemiology of Klebsiella infection has been explored in a number of studies, but none of them have clearly identified a relationship between the capsular serotype and specific clinical diseases. The aim of the present study was to correlate the relationship between the capsular serotype of bacteremic K pneumoniae and the source of infection. We found that serotypes K1 and K2 were common in isolates from patients with abdominal infections, such as liver abscesses, and with community-acquired infections from an unknown source. Klebsiella pneumoniae liver abscess is a well-known disease in Taiwan that presents as a serious infectious complication in patients with diabetes.²⁸ For nosocomial infections, nontypeable isolates were predominant when the source of infection was identified as the lung, the urinary tract, or an unknown source.

Of the 158 strains of K pneumoniae, 58 (36.7%) could be serotyped, and 9 different serotypes were identified. This differed from previous reports in which more than 90% of the isolates have been typeable and the distribution of the serotypes was very wide.^9,13 In Europe and North America, serotypes K2, K21, and K55 (accounting for 8.9%, 7.8% and 4.8%, respectively), made up 21.5% of all isolates, and this was significantly higher than the remaining serotypes in K pneumoniae bacteremic isolates. In contrast, we found serotypes K1, K2, and K28 accounted for 37 (23.4%), 8 (5.1%), and 6 (3.8%), respectively, of all strains in this study.

Of the typeable community-acquired bacteremic isolates (n = 42), 34 (81%) were serotypes K1 and K2, especially with an intra-abdominal infection or an unknown source of infection. Of nosocomial bacteremic isolates, nontypeable serotypes predominated (48/64 [75%]). Serotypes K1 and K2 have been described as the predominant typeable serotype in Taiwan.¹² Serotypes K1 and K2 showed more virulence than other serotypes in mice.³⁴ However, nontypeable capsular serotypes may possess an epidemiological advantage compared with other strains, and this may enable them to colonize and infect hospital patients more easily.^35,36 Nontypeable serotypes accounted for most of the bacteremic isolates where the urinary tract or a lung was identified as an infection source. Serotypes K1 and K2 were significantly more common in patients in whom the source of infection was liver abscess (11 of 22) than in those with other sources of infection (P = .02). For all patients with liver abscesses who developed complications involving metastatic foci, their disease was attributable to serotype K1 isolates. Whether a specific bacterial strain of K pneumoniae is associated with liver abscesses or whether other factors such as race, environment, or the presence of diabetes mellitus are involved requires further clarification. The high proportion of nontypeable strains in the study also requires further investigation.

The overall mortality in our series was 22.8%. The reported mortalities in Klebsiella bacteremia range from 20% to 62%.^1,6,17-26 In the present study, risk factors for mortality in patients with K pneumoniae bacteremia included nosocomial infection, lung infection, thrombocytopenia, leukopenia, infection with ceftazidime-resistant isolates, inappropriate antimicrobial therapy, and septic shock. In contrast to previous studies,^18,24 an increase in serum creatinine or total serum bilirubin level at the onset of bacteremia was not associated significantly with a higher mortality rate. We could not detect any statistically significant difference for mortality between any typeable serotype and the nontypeable serotypes.

Klebsiella pneumoniae bacteremia is a very important cause of morbidity and mortality in gram-negative bacteremia in Taiwan, and this is similar to the situation in most medically well-developed countries. Urinary tract infection and liver abscess were the most common infection sources for K pneumoniae bacteremia. Diabetes mellitus, neoplastic diseases, and hepatobiliary diseases were the most common underlying conditions. Of all community-acquired and nosocomial isolates, 2% and 22%, respectively, were ceftazidime resistant. Thus, strict control of the use of third-generation cephalosporins and other antibiotics should be instituted in the community and in the hospital.³⁷ Only 36.7% of K pneumoniae bacteremia isolates could be serotyped; of these, 45 (78%) were serotypes K1 and K2. This study shows a significantly higher prevalence of K1 in bacteremic K pneumoniae isolate in Taiwan, especially in patients with an infection source identified as liver abscess, biliary tract infection, or unknown. Of the nosocomial bacteremic isolates, a nontypeable serotype was predominant in 48 (75%). Further study on the capsular serotype and the virulence of the K pneumoniae is warranted. The mortality rate remains high, despite proper antimicrobial treatment and supportive measures. Factors indicating poor prognosis include the presence of a nosocomial infection, a lung infection, thrombocytopenia, leukopenia, the presence of a ceftazidime-resistant isolate, the application of an inappropriate antimicrobial therapy, and septic shock.

Accepted for publication September 6, 2001.

This study was partly supported by grant TSGH-C88-09 from Tri-Service General Hospital and a grant from the National Health Research Institute, Taipei, Taiwan.

We thank Monto Ho, MD, for his critical review of the report.

Corresponding author and reprints: Feng-Yee Chang, MD, PhD, Division of Infectious Diseases and Tropical Medicine, Tri-Service General Hospital, No. 325, Section 2, Cheng-Kung Road, Neihu, 114 Taipei, Taiwan (e-mail: fychang@ndmctsgh.edu.tw).