Background. Nosocomial bloodstream infections (BSIs) are important causes of morbidity and mortality in the United States.
Methods. Data from a nationwide, concurrent surveillance study (Surveillance and Control of Pathogens of Epidemiological Importance [SCOPE]) were used to examine the secular trends in the epidemiology and microbiology of nosocomial BSIs.
Results. Our study detected 24,179 cases of nosocomial BSI in 49 US hospitals over a 7-year period from March 1995 through September 2002 (60 cases per 10,000 hospital admissions). Eighty-seven percent of BSIs were monomicrobial. Gram-positive organisms caused 65% of these BSIs, gram-negative organisms caused 25%, and fungi caused 9.5%. The crude mortality rate was 27%. The most-common organisms causing BSIs were coagulase-negative staphylococci (CoNS) (31% of isolates), Staphylococcus aureus (20%), enterococci (9%), and Candida species (9%). The mean interval between admission and infection was 13 days for infection with Escherichia coli, 16 days for S. aureus, 22 days for Candida species and Klebsiella species, 23 days for enterococci, and 26 days for Acinetobacter species. CoNS, Pseudomonas species, Enterobacter species, Serratia species, and Acinetobacter species were more likely to cause infections in patients in intensive care units (P < .001). In neutropenic patients, infections with Candida species, enterococci, and viridans group streptococci were significantly more common. The proportion of S. aureus isolates with methicillin resistance increased from 22% in 1995 to 57% in 2001 (P < .001, trend analysis). Vancomycin resistance was seen in 2% of Enterococcus faecalis isolates and in 60% of Enterococcus faecium isolates.
Conclusion. In this study, one of the largest multicenter studies performed to date, we found that the proportion of nosocomial BSIs due to antibiotic-resistant organisms is increasing in US hospitals.
Bloodstream infections (BSIs) are major causes of morbidity and mortality. On the basis of data from death certificates, these infections are the 10th leading cause of death in the United States , and the age-adjusted death rate has risen by 78% over the past 2 decades . The true incidence of nosocomial BSIs is unknown, but it is estimated that ∼250,000 cases occur annually in the United States .
Recent studies reported the incidence of BSIs to be between 1% in intensive care unit (ICU) patients  and 36% in bone-marrow transplant recipients . The crude rate mortality ranged from 12% in total hospital populations to 80% in ICU patients [6–10]. In recent series of ICU patients, crude mortality rates ranged from 35%–53% [4, 11]. The rate of mortality directly attributable to BSIs in these populations has been estimated to be 16%–40% [7, 11]. Among ICU patients with BSI, the length of stay was prolonged by 7.5–25 days and the total hospitalization time by 4.5–32 days [4, 7, 11]. Inappropriate empirical antimicrobial therapy is an important predictor of death in these patients [7, 12, 13].
Throughout the 1960s and 1970s, gram-negative organisms were most frequently isolated from patients with nosocomial BSIs. Since then, infections due to gram-positive organisms have become increasingly frequent [2, 14, 15]. In addition, antibiotic resistance rates have been rising during the past 2 decades for all predominant organisms, including Staphylococcus aureus [14, 16, 17], coagulase-negative staphylococci [18, 19], enterococci , and gram-negative pathogens [15, 21, 22]. In the face of emerging multiresistant organisms, antimicrobial prophylaxis and treatment have become increasingly difficult, and timely and accurate epidemiological information is needed for guiding appropriate empirical therapy. Analyses of Candida BSIs have shown trends to selection of non-albicans species, some of which are difficult to treat with first-generation azoles . This study was conducted to assess the epidemiological features of nosocomial BSIs in the United States, the species distribution, and the antimicrobial susceptibilities of causative pathogens.
Materials and Methods
The Surveillance and Control of Pathogens of Epidemiologic Importance (SCOPE) Project is based at Virginia Commonwealth University in Richmond, Virginia, and includes 49 hospitals of various sizes (range, 60–1200 beds) that are geographically dispersed throughout the United States (figure 1) [14, 24]. The data generated by the SCOPE Project have shown a high correlation with data from the National Nosocomial Infection Surveillance (NNIS) program of the Centers for Disease Control and Prevention (CDC; Altanta, GA) .
Study design. Clinical data were prospectively collected by local infection control practitioners using a standardized case-report form and forwarded to the coordinating center along with each microbiological isolate. The study began in 1995 and continues to the present. Patients admitted from 1 March 1995 through 30 September 2002 to 1 of the 49 hospitals participating in the SCOPE project were eligible if they met the criteria for a nosocomial BSI. A nosocomial BSI was diagnosed if 1 or more culture of blood drawn at least 48 h after admission yielded a pathogenic organism. If the bloodstream isolate was a potential skin contaminant (e.g., diphtheroids, Propionibacterium species, Bacillus species, coagulase-negative staphylococci, or micrococci), the presence of an intravascular catheter and the initiation of targeted antimicrobial therapy were required for the diagnosis, as well as at least 1 of the following findings: temperature of >38.0°C or <36.0°C, chills, and/or systolic blood pressure of <90 mm Hg. BSI episodes that represented relapses were excluded. However, second episodes that occurred during separate admissions were included as separate cases. We studied the data from all eligible episodes of BSI.
The data that were routinely collected included the patient's age, sex, location at the onset of BSI (ICU vs. non-ICU ward), clinical service at the onset of BSI, and predisposing clinical conditions, as well as the species distribution and antimicrobial susceptibility of causative pathogens and the outcome during hospitalization (i.e., crude mortality). Predisposing clinical conditions that were routinely recorded included neutropenia (defined as an absolute neutrophil count of <1000 cells/µL), receipt of peritoneal dialysis or hemodialysis, and/or presence of intravascular catheters (i.e., central lines, arterial catheters, or peripheral intravenous catheters). Sources of secondary BSI were identified by cultures of samples obtained from distant sites that yielded the same pathogen with an identical resistance pattern.
Incidence. To calculate incidence rates, we collected admission data from the participating hospitals. The 7 hospitals that participated for <12 months in the study were excluded, leaving 42 hospitals for this analysis. Incidence rates were calculated as number of BSIs per 10,000 hospital admissions.
Microbiological methods. Blood cultures were processed at the participating hospitals. The identification of blood isolates and susceptibility testing were done with the routine methods in use at the affiliated laboratories. All affiliated laboratories were College of American Pathologists—certified, and all microbiological methods used were consistent with current NCCLS recommendations. Data from all hospitals were used for analysis, and denominators for individual antimicrobial agents may vary because not all hospitals test and report all drugs. Several subprojects have evaluated samples of organisms from this study and have included validation measures, such as reidentification of species and retesting of antimicrobial susceptibilities in the Special Microbiology Laboratory at the University of Iowa College of Medicine; the Department of Epidemiology at the Virginia Commonwealth University Medical Center; and the Institute for Medical Microbiology, Immunology and Hygiene at the University of Cologne, Germany [25–29].
Statistical analysis. The results were expressed as the mean ± SD or as a proportion of the total number of patients or isolates. For continuous variables, mean values were compared using 2 sample t tests for independent samples. Differences in proportions were compared using a χ2 test or Fisher's exact test, as appropriate. Mean values are reported with SDs. All tests of significance are 2-tailed; α was set at .05. The Mann-Whitney U test was performed to test the equality of continuous variables. All statistical analyses were done using SPSS software (SPSS).
Study population and patient characteristics. During the 7.5-year study period, a total of 24,179 infections with 27,847 organisms were reported by participating hospitals. Of these, 3432 clinically significant episodes of BSI (15%) were identified in pediatric patients (⩽16 years of age). Patients had a mean age of 51 ± 26.5 years (range, 0–102 years). Forty-four percent of patients were female.
Approximately 51% (50.5%) of hospital-acquired BSIs occurred in the ICU. The clinical services most commonly diagnosing and reporting BSIs were internal medicine (38%), general surgery (20%), and pediatrics (13.5%).
The most frequent underlying conditions (recorded as diagnoses at admission) were malignancy, in 4237 patients (17.5%); cardiac conditions, in 3626 patients (15%); and gastrointestinal conditions, in 3269 patients (13.5%). However, underlying conditions were not specified for 4356 patients (14%).
Among the potential factors predisposing patients to BSI, intravascular devices were the most frequent. Central venous catheters were in place in 17,484 patients (72%), peripheral intravenous catheters were in place in 8426 patients (35%), and arterial catheters were in place in 3821 patients (16%). Urinary catheters were in place in 11,101 patients (46%). A total of 5791 patients (24%) were receiving total parenteral nutrition, and 1896 patients (8%) needed dialysis at the onset of BSI. Most of these patients underwent hemodialysis (1769 patients vs. 127 who underwent peritoneal dialysis). Ventilator support was necessary for 7844 patients (32%). Overall, 6618 patients died during hospitalization, accounting for a crude mortality rate of 27%.
The system-wide incidence of BSI was 60 cases per 10,000 hospital admissions. When stratified by pathogen, system-wide incidence rates of BSI due to the most commonly isolated pathogens varied between 1 and 16 cases per 10,000 admissions, which were the incidence rates for BSI due to Acinetobacter baumannii and CoNS, respectively (table 1). The median hospital incidence of BSI was 43 cases per 10,000 hospital admissions (range, 6–252). Incidences did not vary significantly over time (mean ± SD, 54 ± 40 cases per 10,000 admissions in 1996 vs. 52 ± 38 in 2001), and no trend could be observed.
Microbiological features. Thirteen percent of all episodes of BSI (n = 3201) were polymicrobial. Of 20,978 monomicrobial episodes, a total of 13,665 episodes (65%) were caused by gram-positive organisms and 5278 (25%) by gram-negative organisms. Fungi, mainly Candida species, were isolated in a total of 2002 episodes (9.5%). Anaerobic bacteria accounted for 1.3% of BSIs. These proportions did not change significantly during the study period. The rank order of the major pathogens (table 1) shows that CoNS accounted for nearly one-third of all nosocomial BSIs (31%), followed in rank by S. aureus (20%), enterococci (9%), and Candida species (9%). E. coli (6% of episodes) and Klebsiella species (5%), were the most common gram-negative organisms.
CoNS, Enterobacter species, Serratia species, A. baumannii, and Candida species were more likely to be isolated from patients in ICUs (P < .001), whereas S. aureus, Klebsiella species and E. coli were more common in patients in wards (P < .001). No significant differences were seen for enterococci or Pseudomonas aeruginosa (table 1).
When stratified by clinical service, the following patterns emerged: CoNS were the most frequently isolated pathogens for all services, except orthopedics and obstetrics, where S. aureus and E. coli, respectively, were more frequently isolated. S. aureus, enterococci, and Candida species usually followed in various rank orders (table 2).
When different age groups were compared, the proportion of CoNS decreased from 49% in patients <1 year to 27% in patients >65 years, whereas the proportion of S. aureus in the same patient populations increased from 9.5% to 24%, respectively. For gram-negative pathogens and Candida species, the proportions remained stable.
When patients were stratified into those with neutropenia and those without neutropenia, 2 differences emerged. BSI due to S. aureus was more common among nonneutropenic patients (21%, compared with 9% among neutropenic patients), whereas BSI due to viridans group streptococci was more common among patients with neutropenia (2%, compared with 0.5% among nonneutropenic patients). Other organisms accounted for about the same proportion (absolute difference, <2%) of BSIs in each group.
In patients with monomicrobial BSIs, the crude mortality (table 1) ranged from 21% and 22% (for CoNS and E. coli, respectively) to 39% (for P. aeruginosa and Candida species). In ICU patients, the crude mortality ranged from 26% and 34% (for CoNS and E. coli, respectively) to 48% and 47% (for P. aeruginosa and Candida species, respectively). In patients with polymicrobial BSIs, the crude mortality was 32%.
The mean time from hospital admission to onset of BSI due to the major pathogens (figure 2) ranged from 12 days (for E. coli) to 26 days (for A. baumannii). Time to infection decreased with increasing age, from 26 days, in patients <1 year of age, to 16 days, in patients >65 years of age. No significant seasonal or geographical patterns could be observed for any of the organisms when different time periods and US geographical regions (northwest, northeast, southwest, and southeast) were compared.
Primary BSI, in which no source could be determined, was seen in 12,893 patients (53%). Secondary BSI originated from intravenous catheters in 5749 patients (24%), from the urinary tract in 1580 patients (6.5%), and from the lower respiratory tract in 1539 patients (6%). However, culture samples from distant sites were infrequently collected; therefore, some points of origin might not have been detected.
Of the 1890 Candida isolates causing nosocomial BSI, C. albicans was the most common, accounting for 54% of cases of Candida BSI, followed in rank order by C. glabrata (19%), C. parapsilosis (11%), and C. tropicalis (11%) (figure 3). Crude mortality was lowest for C. albicans infection (37%) and highest for C. krusei infection (59%) (figure 3). There was an increase in the proportion of Candida species isolated from blood cultures from 8% in 1995 to 12% in 2002 (P < .001, trend analysis). Also, the proportions of C. albicans and C. parapsilosis among these isolates increased between 1995 and 2002, but the proportions of C. tropicalis and C. glabrata decreased.
Antimicrobial susceptibility. Methicillin resistance was detected in 1699 S. aureus isolates (41% of tested isolates) and in 4946 CoNS isolates (75%). The proportion of S. aureus isolates with methicillin resistance was significantly higher among ICU patients than among ward patients (44% vs. 40%; P = .004), and there was also a trend toward a higher proportion of S. aureus isolates resistant to methicillin among patients without neutropenia than among patients with neutropenia (42% vs. 32%; P = .054, figure 4). The proportion of S. aureus isolates resistant to methicillin increased from 22% in 1995 to 57% in 2001 (P < .001, trend analysis; figure 5).
Among enterococcal isolates, vancomycin resistance was found in 60% of E. faecium isolates and in 2% of E. faecalis isolates (figure 5). Among viridans group streptococci, decreased susceptibility to penicillin (MICs of ⩾0.25 mg/L) and to erythromycin was detected in 39 (35%) of 112 isolates and 44 (42%) of 104 of isolates, respectively.
Antimicrobial resistance levels for the most common gram-negative organisms causing nosocomial BSIs are shown in table 3. A relatively high proportion of E. coli isolates displayed resistance to ampicillin, piperacillin, and ampicillin-sulbactam (44%, 41%, and 39%, respectively). Resistance to trimethoprim-sulfamethoxazole was seen in 19% of isolates. Third-generation cephalosporins, aminoglycosides, and ciprofloxacin displayed activity against most isolates, as did imipenem and aztreonam (⩽5% of isolates were resistant). For Klebsiella isolates, third-generation cephalosporins, aminoglycosides, fluoroquinolones, aztreonam, and imipenem were active against >80% of isolates tested. For Enterobacter isolates, imipenem displayed the greatest activity (1% were resistant). Of the P. aeruginosa isolates, 11%, 14%, and 16% were resistant to piperacillin, imipenem, and ceftazidime, respectively. Resistance to ciprofloxacin was seen in 20% of tested isolates. Of note, the proportion of P. aeruginosa isolates resistant to ceftazidime increased from 12% in 1995 to 29% in 2001 (P < .001, trend analysis; figure 5).
In the face of increasing antimicrobial resistance, surveillance programs have become important in defining the species distribution and resistance patterns of pathogens causing BSIs, and thus are providing the basis for appropriate empirical therapy. Data from Ibrahim et al.  showed that mortality rates doubled, from 30% to 60%, when inappropriate empirical antibiotic therapy was given to ICU patients with BSIs.
There are several differences between this surveillance project and other networks that monitor a variety of infections and/or countries [14–16, 31]. The SCOPE project focuses entirely on nosocomial BSIs in the United States and constitutes the largest nongovernmental surveillance project in this area. It is a clinically oriented surveillance project, collecting data on defined infections, not on culture results alone. Moreover, the surveillance encompasses the entire hospitalized population at the sentinel hospitals, rather than just those patients in ICUs.
As in other series, the dominance of gram-positive pathogens has been documented in this study. However, our design might have led to an overestimation of BSIs due to CoNS, because even though strictly defined clinical signs were required for inclusion in the study, patients with only 1 blood culture yielding CoNS were eligible. Recent studies have shown that the proportion of contaminants is lower if 2 or more positive blood culture results (of samples drawn within a few hours of each other) were required for inclusion [9, 32, 33]. However, this should be minimized somewhat by the strict clinical definitions used by SCOPE to ascertain the clinical relevance of the positive blood culture. In addition, SCOPE data have shown a high correlation to the NNIS data .
The proportion of BSIs due to Candida species in our study differs significantly from previously reported data. A recent NNIS study  showed significant decreases in BSIs due to Candida species and C. albicans but reported a significant increase in BSIs due to C. glabrata in ICUs from 1989 through 1999. In contrast, we found a significant increase in the proportion of Candida species among isolates from blood cultures, from 8% in 1995 to 12% in 2002. Furthermore, the proportions of C. albicans and C. parapsilosis among these isolates increased from 1995 through 2002, whereas the proportions of C. tropicalis and C. glabrata decreased.
The rate of resistance to methicillin that we identified among S. aureus infections was higher than that identified in northern Europe  but comparable to rates of resistance reported in other series from the United States [15, 35]. The rate of resistance to vancomycin in E. faecium infections was comparable to those in recent studies from the United States [15, 35] but higher than those observed in recent European studies [34, 36].
In our series, the prevalence of resistance to vancomycin among E. faecium isolates was significantly higher among those recovered from neutropenic patients. Several recent studies found neutropenia to be independently associated with vancomycin-resistant enterococcal (VRE) infection when comparing patients with vancomycin-resistant and patients with vancomycin-susceptible enterococcal BSIs [37, 38]. In contrast, Lucas and co-authors , in comparing vancomycin-susceptible and vancomycin-resistant enterococcal BSIs , did not find neutropenia to be a risk factor for BSIs due to VRE.
Increasing resistance to penicillin, other β-lactam antibiotics, and macrolides has been documented in viridans group streptococci isolated from neutropenic patients with cancer [40, 41]. In our study, 34% and 42% of viridans group streptococci displayed decreased susceptibility to penicillin and macrolides, respectively. These rates are in accordance with recent data from Germany  but are lower than rates in southern Europe .
CoNS and Candida species were more likely to be isolated from patients in ICUs, thus influencing empirical antibiotic therapy, whereas S. aureus and E. coli were more commonly isolated from patients in non-ICU wards. Viridans group streptococci were more frequently seen in patients with neutropenia, whereas S. aureus was more commonly seen in nonneutropenic patients. In addition, we found enterococci, Candida species, and Serratia species more frequently in patients without neutropenia.
The finding of a higher proportion of resistance to tobramycin, compared with gentamicin, for several of the gram-negative organisms should not be taken to imply that gentamicin has better activity against these organisms. Rather, this finding is due to the variation in the antibiotics tested at the local laboratories. Thus, the true aminoglycoside susceptibility lies somewhere between the 2 values.
In conclusion, our data demonstrate one-half of all nosocomial BSIs occur in the critical-care setting, although a minority of patients receive care there. The high rates of antibiotic resistance, as well as local patterns of species distribution and drug susceptibilities in certain patient populations, should guide empirical therapy of nosocomial BSIs.
We thank the infection-control practitioners and microbiology laboratory personnel at the participating hospitals for their active participation in this project, and Donna McClish, for assistance with data management.
Financial support. SCOPE is funded in part by Pfizer Inc., Merck Inc., and Cubist Pharmaceuticals.
Conflict of interest. All authors: No conflict.
- staphylococcus aureus
- antibiotic resistance, bacterial
- viridans streptococcus
- enterococcus faecalis
- enterococcus faecium
- gram-positive bacteria
- gram-positive bacterial infections
- gram-positive cocci
- gram-positive rods
- intensive care unit
- methicillin resistance
- vancomycin resistance
- pathogenic organism
- staphylococcus, coagulase negative (organism)
- escherichia coli
- surveillance, medical
- hospital admission
- bloodstream infection