A prospective study of consecutive patients with community-acquired K. pneumoniae bacteremia was performed in 12 hospitals.1 The study period was January 1, 1996, to December 31, 1997. Records of patients >16 years of age with positive blood cultures for K. pneumoniae were reviewed, and a 188-item study form was completed. All items on the form were objective criteria, allowing standardization among medical centers. The study was observational in that administration of antimicrobial agents and other therapeutic management were controlled by the patient’s physician, not the investigators.

Community-acquired bacteremia was defined as a positive blood culture taken on or within 48 hours of admission. Severity of acute illness at the time of positive blood cultures was assessed by a previously validated scoring system, based on mental status, vital signs, need for mechanical ventilation, and recent cardiac arrest (Pitt bacteremia score) (35). Type of infection was defined as pneumonia, urinary tract infection, meningitis, incisional wound infection, other soft tissue infection, intraabdominal infection, and primary bloodstream infection, according to Centers for Disease Control and Prevention definitions (36). In addition, distinctive sites of K. pneumoniae bacteremia were further defined as liver abscess, meningitis, or endophthalmitis. Liver abscesses were defined by the coexistence of blood cultures positive for K. pneumoniae and evidence of an intrahepatic abscess cavity by ultrasonography or computed tomography. Meningitis was defined as culture of K. pneumoniae from cerebrospinal fluid, and endophthalmitis was defined as decreased visual acuity, pain, hypyon, or severe anterior uveitis in a patient concurrently bacteremic with K. pneumoniae. Death was defined as including deaths from all causes within 14 days of the date the first positive blood culture for K. pneumoniae was obtained.

Blood cultures of K. pneumoniae were sent by participating hospitals on nutrient agar slants to a central study laboratory in Pittsburgh, where the identity of each isolate as K. pneumoniae was confirmed by the Vitek GNI system (Biomerieux Vitek, Hazelwood, MO). Extended-spectrum beta-lactamase (ESBL) production was defined phenotypically by broth dilution as a ≥3 twofold concentration decrease in MIC for either cefotaxime or ceftazidime tested in combination with clavulanic acid compared with the MIC when tested alone. The protocol was reviewed and approved by Institutional Review Boards according to local requirements.

All data were entered into a central database (PROPHET version 5.1; BBN Systems and Technologies Corporation, Cambridge, MA). Contingency data were analyzed by two-tailed chi-square or Fisher’s exact tests, and continuous data were analyzed by Student t test or Mann-Whitney U test.

Two hundred two (44.4%) of 455 episodes of K. pneumoniae bacteremia during the study period were community-acquired cases. The percentage of cases of K. pneumoniae bacteremia that were community acquired in each study country differed strikingly: 96 (68%) of 142 in Taiwan, 25 (43%) of 68 in the United States, 28 (39%) of 71 in Australia, 40 (34%) of 116 in South Africa, 6 (22%) of 27 in Europe, and 7 (17%) of 41 in Argentina. K. pneumoniae bacteremia in Taiwan was significantly more likely to be community acquired than was bacteremia in the other countries combined (68% vs. 36%, p=0.0001).

The characteristics of patients with community-acquired K. pneumoniae bacteremia from Taiwan, South Africa, and the rest of the world were compared (Table 1). The source of bacteremia in community-acquired cases was geographically distinctive (Table 2). Pneumonia was the most common infection worldwide, accounting for 57 (28%) of 202 cases. However, 53 (93%) of 57 of all cases of community-acquired K. pneumoniae pneumonia occurred in Taiwan and South Africa.

Antibiotics had been used for >24 hours before admission in 21 (10%) of 202 patients. Prior antibiotic use was significantly lower in Taiwan (4 [4%] of 96 patients) and South Africa (2 [5%] of 40) than in the other countries (15 [23%] of 66) (p=0.0003). In four patients, the prior antibiotic was an oxyimino-containing cephalosporin (two in South Africa, one in Argentina, and one in Taiwan).

ESBL production was detected in 7 (3.5%) of 202 community-acquired strains compared with 78 (30.8%) of 253 hospital-acquired strains (p<0.00001). None of the patients with community-acquired ESBL-producing strains had recently received an oxyimino-containing cephalosporin. However, of the seven patients with community-acquired ESBL-producing K. pneumoniae bacteremia, only one (a patient with pneumonia from Africa) had no recent hospital exposure. The countries from which community-acquired ESBL-producing K. pneumoniae isolates were collected included Africa (three isolates), Turkey (two), the United States (one), and Australia (one). Twenty-five bloodstream isolates were ciprofloxacin resistant, of which 7 (28%) of 25 were community acquired. All these seven patients had serious underlying disease and frequent hospitalizations or nursing home admissions, and none had received a quinolone in the 14 days before hospital admission. Five of seven patients with community-acquired, ciprofloxacin-resistant K. pneumoniae bacteremia were from Taiwan.

When pneumonia due to Klebsiella was first described by Friedlander in 1882, he believed it to be the most common cause of bacterial pneumonia (37). Although this concept was soon refuted in favor of pneumococcus, from the 1930s through the 1960s, 10 to 50 cases of Klebsiella pneumonia were reported each year by large hospitals in the United States (2). U.S. textbooks of medicine continue to list K. pneumoniae as an important cause of community-acquired pneumonia (38,39).

In our prospective study, we found only four cases of community-acquired K. pneumoniae pneumonia in 2 years in nine large hospitals from the United States, Australia, Europe, and Argentina. The hospitals surveyed included an inner-city veterans hospital in the United States and two large inner-city public hospitals in Australia. These three centers care for large numbers of indigent and alcoholic patients.

Recently published reports from the United States, Israel, and Europe support our observations. Neither Vergis et al. (40) from the United States nor Lieberman et al. (41) from Israel found a single case of K. pneumoniae pneumonia in large multicenter studies of community-acquired pneumonia in the 1990s. Nine European studies published since 1990 show that only 14 (2.3%) of 621 patients admitted with severe community-acquired pneumonia requiring intensive-care unit admission had K. pneumoniae as the presumptive etiologic agent (42–50). In contrast, K. pneumoniae continues to be associated with community-acquired pneumonia in Africa and Asia. In our study, we observed 28 cases in Taiwan and 25 in South Africa, accounting for 29% and 62% of all cases of community-acquired K. pneumoniae bacteremia in South Africa and Taiwan, respectively (Table 2). Recent studies from Taiwan, Singapore, and South Africa corroborate these findings. In a Taiwanese study, K. pneumoniae accounted for 34% of 41 cases of community-acquired bacteremic pneumonia (51). K. pneumoniae was the cause of 15% of community-acquired pneumonia requiring intensive-care unit admission in Singapore (52). K. pneumoniae was found to be the cause of pneumonia in 32% of African patients with severe community-acquired pneumonia requiring intensive-care unit admission in Johannesburg (53) and 11% of patients requiring intensive-care unit admission in Cape Town (54).

In our study, 18% of patients with community-acquired K. pneumoniae pneumonia were alcoholics (p=0.007) (Table 3). Alcoholics in Africa and Asia may have limited access to health care (perhaps including reduced access to antibiotics) compared with those in the Americas, Europe, and Australia, and may have respiratory symptoms later. A weakness of our study is that we were not able to ascertain the duration of symptoms before each patient was hospitalized. However, a recent study of aborigines from rural northern Australia (35% of whom were alcoholics and most of whom had suboptimal access to health-care facilities) showed that none of 90 admitted to hospital with community-acquired pneumonia had K. pneumoniae infection (55). The hypothesis that Klebsiella pneumonia is related to poor primary health care for alcoholics may therefore be less plausible.

Bacteremic K. pneumoniae liver abscess occurred almost exclusively in patients from Taiwan (Table 2), consistent with a growing number of reports from Asia describing this distinctive type of infection. K. pneumoniae was the most common cause of liver abscesses in Taiwan, Singapore, and Korea in reports from 1990 to 1999 (5,9,24,31,32); similarly, numerous reports of liver abscess have recently been published from Hong Kong, Thailand, and Japan (23,25–30). In total, >900 patients with K. pneumoniae liver abscess have been reported from Asian countries in the last 10 years; in contrast, reports of only 23 patients with this condition have been published from regions outside Asia in this same period (56–63).

K. pneumoniae meningitis in adults has also been infrequently reported from North America, Europe, and Australia, in contrast to Taiwan. In our study, five cases of bacteremic K. pneumoniae meningitis were in Taiwanese patients and two in African patients (Table 2). Four (57%) of 7 patients with meningitis had prior diabetes mellitus. Meningitis caused by K. pneumoniae in the United States, Australia, and Europe is most often hospital acquired and associated with prior neurosurgical procedures or instrumentation. However, of 115 cases of K. pneumoniae meningitis reported from Taiwan (33,64,65), 84% were community acquired, and 64% of cases had concurrent Klebsiella bacteremia. Unlike pyogenic liver abscess, the clinical course was fulminant, with a death rate of 57% (33,64,65). The death rate from bacteremic K. pneumoniae meningitis in our series was 71%.

We found only one patient (an alcoholic from Taiwan) with K. pneumoniae bacteremia and endophthalmitis. K. pneumoniae endophthalmitis is also likely to be far more common in Asia than elsewhere; >50 cases have been reported in the last 10 years from Asia (6,10,11,19,27,66–70) compared with only 10 from the United States, Europe, and Australia (57,60,62,71–76). More than 50% of previously reported Asian patients with K. pneumoniae endophthalmitis have had concurrent liver abscess (6,10,11,19,27,66–70).

The reason for the geographic preponderance of these severe manifestations of K. pneumoniae infections in Asia is unknown. The geographic diversity of Klebsiella infections possibly results from interaction between bacterial variables, host variables (for example, defects in host defense caused by diabetes mellitus or alcoholism), socioeconomic factors, and possibly genetic susceptibility in different racial groups. We are studying the phenotypic and genotypic differences in K. pneumoniae causing different disease manifestations in different countries. Because no more than three hospitals from each country were included in our study, our results may not necessarily be generalizable to hospitals in other regions. In addition, other countries in the same continent (e.g., other countries in Asia or eastern Europe) were not studied but may have different clinical patterns compared with the study country.

In summary, our results challenge the classic view of serious Klebsiella infections. In the United States, Europe, Argentina, and Australia, we have observed that hospital-acquired K. pneumoniae infections predominate, with community-acquired bacteremia being caused by urinary tract infection, vascular catheter infection, and cholangitis. Classic community-acquired pneumonia is no longer an important entity in these regions. In South Africa, pneumonia (especially in alcoholics) continues to be an important community-acquired infection. In Taiwan, community-acquired pneumonia persists, and distinctive infections such as liver abscess, endophthalmitis, and meningitis have emerged as substantial public health problems.