Logo of jcmPermissionsJournals.ASM.orgJournalJCM ArticleJournal InfoAuthorsReviewers
J Clin Microbiol. 2003 Aug; 41(8): 3655–3660.
PMCID: PMC179863

Epidemiology and Outcome of Nosocomial and Community-Onset Bloodstream Infection


We performed a prospective study of bloodstream infection to determine factors independently associated with mortality. Between February 1999 and July 2000, 929 consecutive episodes of bloodstream infection at two tertiary care centers were studied. An ICD-9-based Charlson Index was used to adjust for underlying illness. Crude mortality was 24% (14% for community-onset versus 34% for nosocomial bloodstream infections). Mortality attributed to the bloodstream infection was 17% overall (10% for community-onset versus 23% for nosocomial bloodstream infections). Multivariate logistic regression revealed the independent associations with in-hospital mortality to be as follows: nosocomial acquisition (odds ratio [OR] 2.6, P < 0.0001), hypotension (OR 2.6, P < 0.0001), absence of a febrile response (P = 0.003), tachypnea (OR 1.9, P = 0.001), leukopenia or leukocytosis (total white blood cell count of <4,500 or >20,000, P = 0.003), presence of a central venous catheter (OR 2.0, P = 0.0002), and presence of anaerobic organism (OR 2.5, P = 0.04). Even after adjustments were made for underlying illness and length of stay, nosocomial status of bloodstream infection was strongly associated with increased total hospital charges (P < 0.0001). Although accounting for about half of all bloodstream infections, nosocomial bloodstream infections account for most of the mortality and costs associated with bloodstream infection.

Bloodstream infections cause substantial morbidity and mortality (7, 18, 24). Increasing rates of antimicrobial resistance (1, 6-8, 21), changing patterns of antimicrobial usage (8), and the wide application of new medical technologies (e.g., indwelling catheters and other devices) may change the epidemiology and outcome of bloodstream infection. It is therefore important to continually review and update the epidemiology and outcome of bloodstream infection, including an examination of the variables most strongly associated with mortality. Understanding these variables will help to prioritize resources and plan strategies for decreasing the mortality associated with bloodstream infection.

We sought to determine the current epidemiology and outcome of bloodstream infection by prospectively evaluating consecutive patients at two large tertiary-care hospitals. Using multivariate models and controlling for underlying illness, we sought to determine which variables were most strongly and independently associated with mortality among patients with bloodstream infection. We were particularly interested in the relative contributions of nosocomial bloodstream infection to the overall mortality and costs associated with bloodstream infection.


From February 1999 to July 2000, all patients with a positive signal from the BacT/Alert blood culture system (Organon Teknika, Durham, N.C.) at the University of Iowa Hospital (UIHC; Iowa City) and the Lahey Clinic (Burlington, Mass.) were screened for enrollment. Both centers used the BacT/Alert blood culture system with FAN aerobic and standard anaerobic bottles.

Each patient enrolled in the study underwent a medical record review performed by an experienced reviewer; data were abstracted to worksheets and entered into an Access (Microsoft Corp., Seattle, Wash.) database for use in statistical analyses. Outcome measures included mortality and hospital charges. Mortality was defined as death occurring during the same hospitalization as the bloodstream infection. Additional outcome measures examined in some analyses included 30-day mortality and “attributable” mortality. Attributable mortality was defined as death occurring after bloodstream infection in patients who were deemed not to have responded to therapy (i.e., who died as a direct or indirect result of the bloodstream infection). Hospital charge data were obtained from the business office computer system. Charge data included only hospital charges and not physician charges.

Exclusion criteria.

Patients with false-positive signals (no organisms on Gram stain or subculture from bottle), patients younger than 16 years, patients not admitted to the hospital, autopsy blood cultures, and blood cultures referred from other medical centers (where the patient was not admitted to one of the study centers) were excluded from the study.


“Bloodstream infection episode” was defined as an episode of bloodstream infection from the time the first positive blood culture was obtained (designated T0). A clinical judgment was made regarding whether additional blood cultures positive for the same organism isolated at T0 represented part of the original bloodstream infection or a new bloodstream infection episode. “Community-onset” versus “nosocomial infection” were determined as follows. Each bloodstream infection episode was classified based upon Centers for Disease Control and Prevention (CDC) guidelines (9). In general, bloodstream infection episodes in which T0 was >48 h after hospital admission were considered to be nosocomial. In addition, any patient who had been admitted to either UIHC or Lahey Clinic within 48 h prior to admission or who were transferred from another hospital, a rehabilitation hospital, or a long-term care facility were considered to have a nosocomial bacteremia. Because many bacteremias that are present or incubating upon admission to the hospital are nonetheless healthcare associated, we refer to nonnosocomial bloodstream infections as community onset rather than community acquired. A “clinically significant” versus a “clinically nonsignificant (contaminant)” isolate was defined as follows. Each positive blood culture was critically assessed and categorized as either clinically significant or not significant, taking into account organism identification, clinical signs and symptoms, white blood cell (WBC) count, number of positive blood cultures out of the total number drawn, results of other cultures, pathology findings, imaging results, and clinical course. All indeterminate cases were reviewed with a physician specializing in infectious diseases prior to classification.

Underlying-illness scores.

We utilized the Charlson Index (4) to assign an underlying illness score for each patient entered in the study. The index assigned points to each patient by using the ICD-9 codes for the admission during which the bloodstream infection occurred.

Statistical analyses.

Univariate analyses were performed by using the chi-square or Fisher exact tests, as appropriate. The continuous classification variables were analyzed by using the Student t test for groups. Multivariate logistic regression models were constructed by using all independent variables that were associated with the outcome variable (P < 0.1). Variables significant at an alpha of 0.05 were retained in the final model. Stepwise variable selection was used. The C statistic and the Hosmer and Lemeshow goodness-of-fit statistic were used to assess adequacy of the model fit. Univariate analyses for the charge outcomes were performed by using one-way analysis of variance with an unbalanced design. The dependent charge variable underwent log transformation to linearize the regression model and to normalize the dependent variable. Multivariate linear regression models were constructed by using PROC GLM. R2 was used to assess the adequacy of the model fit. All statistical analyses were performed by using the SAS software program (version 8.0; SAS, Inc., Cary, N.C.).


Just over half (52%) of the patients in our study acquired their bloodstream infection in the hospital or at another healthcare facility (Table (Table1).1). Interestingly, there was no significant difference in age between those with community-onset versus nosocomial bloodstream infections (mean age, 60 years). In both community and hospital settings, bloodstream infection in males was predominant, and the most common source (when a source was documented) was an intravenous catheter. Overall, crude mortality was 24%, with much higher crude mortality among patients with nosocomial (34%) versus community-onset (14%) bloodstream infections. The attributable mortality of bloodstream infection was 17% (10% for community-onset versus 23% for nosocomial bloodstream infections). The vast majority (72%) of all deaths, and 71% of deaths attributable to bloodstream infection, occurred among patients with nosocomial bloodstream infections. The total hospital charges and length of stay were also much higher for patients with nosocomial bloodstream infections (Table (Table11).

Comparison of characteristics of community-onset versus nosocomial bloodstream infection episodes

Gram-positive pathogens caused the majority of both nosocomial and community-onset bloodstream infections (Table (Table2),2), with Staphylococcus aureus being the most common pathogen overall. However, the rank order of pathogens was different for community-onset bloodstream infections than for nosocomial bloodstream infections. Specifically, Escherichia coli was the most common cause of community-onset bloodstream infection, whereas S. aureus caused similar proportions of both community-onset (18%) and nosocomial (21%) bloodstream infections.

Summary of pathogens detected from community-onset and nosocomial bloodstream infection episodes

Table Table33 summarizes the crude mortality of bloodstream infection for each organism and organism group. The highest crude mortality rates were for bloodstream infection due to yeast (44%) and anaerobes (41%). Streptococcus pneumoniae bloodstream infection was associated with the lowest crude mortality (7%).

Summary of crude mortality by organism or organism group for both community-onset and nosocomial bloodstream infection episodes

Among patients who died, the mean and median time from T0 to death were 12.7 and 6 days, respectively (range, 0.2 to 101 days). A total of 52% of deaths occurred within the first week after T0, and 88% of deaths occurred within the first 30 days after T0. Among patients in whom death was felt to be attributable to the bloodstream infection, the mean time from T0 to death was shorter (7 days), with 70% of deaths occurring within the first week and 97% occurring within 30 days after T0.

Univariate associations with mortality among patients with bloodstream infections included sex, source, admitting service, body temperature, hypotension, WBC count, respiratory rate, immunosuppression, presence of a central venous catheter (CVC) at T0, Charlson Index score, organism group, and nosocomial acquisition of bloodstream infection (Table (Table4).4). Age (increased risk with increasing age) did not achieve a statistically significant univariate association with mortality (P = 0.12). Figure Figure11 depicts crude mortality rates stratified by selected patient characteristics at T0.

FIG. 1.
Crude mortality rates according to body temperature, SBP, respiratory rate, and total WBC count/mm3 at the time of index positive blood culture, i.e., T0 (P < 0.05 for each variable [see Table Table44]).
Univariate associations with death after bloodstream infection

Antimicrobial use and antimicrobial susceptibility test results were recorded for each patient. Excluding patients who died within 48 h of T0 (and for whom information from the laboratory was not available to guide therapy), 97% of bacteremias were treated with at least one drug to which the infecting organism(s) were susceptible. The crude mortality rate for the 30 bacteremias for which antibiotic therapy was not consistent with susceptibility results was 30% versus 19% for the other 840 bacteremias (P = 0.12).

A multivariate logistic regression model (Table (Table5)5) revealed several host factors to be strongly associated with in-hospital mortality after bloodstream infection. Anaerobes were the only organisms found to be independently associated with mortality in this model. Moreover, when 30-day mortality (rather than total in-hospital mortality) was used as the outcome measure, the presence of anaerobic organisms no longer reached statistical significance, although the rest of the model was unchanged.

Multivariate logistic regression model of factors independently associated with death after bacteremiaa

We also examined univariate associations between specific organism groups that had high crude mortality rates and acute measures associated with the systemic inflammatory response syndrome (systolic blood pressure, respiratory rate, temperature at T0). The organism group most strongly associated with any of these measures was the anaerobe category (P < 0.0001) for association with hypothermia (T < 36) and P = 0.02 for association with hypotension (systolic blood pressure [SBP] of <90 at T0).

A multivariate linear regression model was constructed to examine factors associated with total hospital charges, including length of stay, admitting service, organism type, and underlying illness score. Nosocomial acquisition of bloodstream infection was strongly associated (P < 0.0001) with increased total hospital charges (Table (Table66).

Multivariate linear regression model of factors independently associated with total hospital costs for patients with bacteremiaa


Surveillance studies have documented both an increase in antimicrobial resistance rates and a shift in organism distribution among important bloodstream pathogens both in the hospital and in community settings (1, 6-8, 21, 23). For example, in the hospital setting there has been a shift from a predominance of gram-negative organisms in the late 1970s to the present day primacy of gram-positive organisms as causes of nosocomial bloodstream infection (23). Coincident with this shift in the microbiology of bloodstream infections have come changes in the patient population. Increasing use of medical technology, the availability of life-saving treatments such as solid organ and hematopoietic stem cell transplantation, and improved intensive and supportive care have allowed for the survival of more severely ill patients, patients who are extremely vulnerable to infection. Both changes in microbiology and changes in host characteristics might be expected to influence outcome from bloodstream infection.

We present data from one of the largest recent studies of both community and nosocomial bloodstream infections. In this prospective study of over 900 episodes of bloodstream infection, we found an overall crude mortality rate of 24% and an attributable mortality rate of 17%. This is remarkably similar to the crude mortality rate of ca. 23% (18% associated directly or indirectly and 5% not associated with the bloodstream infection) reported by Weinstein et al. in a comprehensive review of 843 episodes of positive blood cultures from 1992 to 1993 (24). Their three-center study also reported an identical distribution of nosocomial versus community-onset infection (52% nosocomial) (24). Other similarities with our study, which was performed approximately 8 years later, included (i) the finding that intravenous catheters were the most common source of bloodstream infection; (ii) the finding that crude mortality was highest for yeast, anaerobic, and polymicrobial bacteremias; and (iii) the observation that vital signs (respiratory rate, temperature, and SBP) and total WBC counts at the time the blood culture was obtained were strongly and independently associated with mortality (24).

It is not surprising and is consistent with previous reports (2, 11, 20, 24) that we found certain indicators of the systemic inflammatory response syndrome or septic shock to be associated with mortality due to bloodstream infection. Of particular interest is our confirmation of the importance of a febrile response to infection, either as an indicator of underlying illness or as a protective or beneficial physiologic response. Confirmation of an elevated body temperature as protective in a model adjusted for underlying illness provides an additional reason to critically evaluate the practice of aggressive antipyretic therapy for infected patients with a physiologically elevated body temperature (14).

In contrast to previous studies (16, 18, 24), we did not find strong independent associations of specific organisms or organism groups with mortality. We found only anaerobes to be independently associated with mortality, an association that was no longer statistically significant when 30-day mortality was used as the outcome measure. Anaerobic bloodstream infection episodes were also more likely to be associated with hypothermia and hypotension at the time the blood culture was obtained. Anaerobic bacteremia has been associated in other recent studies with crude mortality rates of 25 to 38% (10, 13, 22). Given that anaerobes accounted for only 3% of bacteremias in our study, a finding similar to those of other recent reports (13, 22), it is interesting that it was the organism group with the strongest independent association with mortality. The low incidence of anaerobic bacteremia makes it more difficult to examine its independent impact on mortality in any but the largest of studies. A very large study, involving more than 4,000 episodes of bacteremia, did find anaerobic bacteremia to be independently associated with septic shock (12).

In contrast to data from multicenter nosocomial bloodstream infection surveillance programs (7-9, 21), we found S. aureus rather than coagulase-negative staphylococcus to be the most common cause of nosocomial bloodstream infection. This is due to our application of a comprehensive retrospective assessment of the clinical significance of positive blood cultures due to coagulase-negative staphylococci (and other common skin contaminants), and the inclusion of only cultures felt to be clinically significant. CDC criteria include aspects of physician behavior (e.g., antimicrobial use) to define nosocomial bloodstream infection in the setting of a single positive blood culture due to a common skin contaminant (9). As a result, many positive blood cultures for coagulase-negative staphylococci defined as nosocomial bloodstream infections by CDC criteria in fact constitute contaminant isolates (S. E. Beekmann, D. J. Diekema, E. Munson, and G. V. Doern, Abstr. 12th Annu. Meet. Soc. Healthcare Epidemiol. America, abstr. 103, 2002). We recommend more-stringent criteria to define nosocomial bloodstream infection due to coagulase-negative staphylococci and other common skin contaminants.

Antimicrobial use has been associated with the outcome of bloodstream infection (2). Our data suggest that current empirical antimicrobial use is so broad spectrum that almost all bacteremic patients receive at least one antibiotic to which their organism is susceptible in vitro. An interesting question, given the continued emergence of antimicrobial resistance, is whether clinicians are appropriately vigilant about narrowing this broad-spectrum therapy once antimicrobial susceptibility test results are available. Indeed, recently published data from our hospital demonstrated that the release of antimicrobial susceptibility test results had little impact on antimicrobial management (17). These data suggest that programs designed to optimize antimicrobial use should focus on the narrowing of therapeutic regimens at a 48- to 72-h time point after the initiation of therapy, when laboratory results are available to guide therapy.

Retrospective cohort study designs can help determine the extent to which underlying disease contributes to crude mortality from bloodstream infection in hospitalized patients (15, 26). These studies have generally demonstrated a major contribution of the bloodstream infection itself to mortality, with attributable mortality estimates ranging from 14% for coagulase-negative staphylococcal bloodstream infections (15) to 37% for Candida bloodstream infections (26). Our study, which contained no control or comparison group, used a clinical judgment to assess the relative contribution of the bloodstream infection to crude mortality. Nonetheless, our estimate of an overall attributable mortality of 23% for nosocomial bloodstream infections is very close to the assumption of a 20% attributable mortality used by Wenzel and Edmond to estimate that between 17,500 and 70,000 deaths occur annually in the United States as a result of nosocomial bloodstream infection (25).

Most studies of the epidemiology of bloodstream infection focus on nosocomial infections alone. Our study examines both community-onset and nosocomial bloodstream infections, allowing us to estimate the proportion of all bloodstream infection mortality that is associated with nosocomial versus community-onset infections. We found that 72% of the crude mortality and 71% of the attributable mortality occurred among patients with nosocomial bloodstream infection. In addition, hospital acquisition (nosocomial status) of bloodstream infection was strongly associated with mortality in our multivariate model. The nosocomial status variable almost certainly served as a marker of underlying illness variables for which it is difficult to completely adjust. However, our data suggest that there is substantial mortality attributable to nosocomial bloodstream infections; indeed, our attributable mortality estimate for nosocomial bloodstream infections was more than twice that for community-onset bloodstream infections. Our multivariate linear regression model examining factors associated with total costs also demonstrated that hospital acquisition of bloodstream infection was strongly associated with increased costs. This model included length of hospital stay.

Given the high crude and attributable mortality and significant independent association with mortality and increased costs, we conclude that nosocomial bloodstream infection is an important target for the most aggressive strategies for prevention and control (25). Any intervention that successfully decreases nosocomial bloodstream infection rates is likely to result in dramatic mortality and cost benefits. Several important interventions, including (i) adherence to guidelines on insertion and care of CVCs (3), (ii) use of antimicrobial or antiseptic impregnated catheters (5), and (iii) improved compliance with hand hygiene (19) are within our grasp and should be instituted without delay.


This study was supported in part by a research grant from Organon Teknika.


1. Archibald, L., D. Phillips, J. E. Monnet, J. E. McGowan, F. Tenover, and R. Gaynes. 1997. Antimicrobial resistance in isolates from inpatients and outpatients in the United States: increasing importance of the intensive care unit. Clin. Infect. Dis. 24:211-215. [PubMed]
2. Bryant, R. E., A. F. Hood, C. E. Hood, and M. G. Koenig. 1971. Factors affecting mortality of gram-negative rod bacteremia. Arch. Intern. Med. 127:120-128. [PubMed]
3. Centers for Disease Control and Prevention. 2002. Guidelines for the prevention of intravascular catheter-related infections. Morb. Mortal. Wkly. Rep. 51:1-26.
4. Charlson, M. E., P. Pompei, K. L. Ales, and C. R. MacKenzie. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J. Chronic Dis. 40:373-383. [PubMed]
5. Darouiche, R. O., I. Raad, S. O. Heard, J. I. Thornby, O. C. Wenker, A. Gabrielli, J. Berg, N. Khardori, H. Hanna, R. Hachem, R. L. Harris, and G. Mayhall. 1999. A comparison of two antimicrobial-impregnated central venous catheters. N. Engl. J. Med. 340:1-8. [PubMed]
6. Diekema, D. J., M. A. Pfaller, R. N. Jones, G. V. Doern, K. C. Kugler, M. L. Beach, H. S. Sader, et al. 2000. Frequency of occurrence and trends in antimicrobial susceptibility of bacterial pathogens isolated from patients with bloodstream infections in the United States, Canada, and Latin America: report from the SENTRY Antimicrobial Surveillance Program, 1998. Int. J. Antimicrob. Agents 13:257-271. [PubMed]
7. Edmond, M. B., S. E. Wallace, D. K. McClish, M. A. Pfaller, R. N. Jones, and R. P. Wenzel. 1999. Nosocomial bloodstream infections in United States hospitals: a three-year analysis. Clin. Infect. Dis. 29:239-244. [PubMed]
8. Fridkin, S. K., C. D. Steward, J. R. Edwards, E. R. Pryor, J. E. McGowan, Jr., L. K. Archibald, R. P. Gaynes, F. C. Tenover, et al. 1999. Surveillance of antimicrobial use and antimicrobial resistance in Unites States hospitals: project ICARE phase 2. Clin. Infect. Dis. 29:245-252. [PubMed]
9. Gaynes, R. P., and T. C. Horan. 1996. Surveillance of nosocomial infections, p. 1017-1031. In C. Mayhall (ed.), Hospital epidemiology and infection control. The Williams & Wilkins Co., Baltimore, Md.
10. Gomez, J., V. Banos, J. Ruiz, F. Herrero, M. Perez, L. Pretel, M. Canteras, and M. Valdes. 1993. Clinical significance of anaerobic bacteremias in a general hospital. Clin. Investig. 71:595-599. [PubMed]
11. Kreger, B. E., D. E. Craven, and W. R. McCabe. 1980. Gram-negative bacteremia: reevaluation of clinical features and treatment in 612 patients. Am. J. Med. 68:344-355. [PubMed]
12. Leibovici, L., M. Drucker, H. Konigsberger, Z. Samra, S. Harrari, S. Ashkena, and S. D. Pitlik. 1997. Septic shock in bacteremic patients: risk factors, features, and prognosis. Scand. J. Infect. Dis. 29:71-75. [PubMed]
13. Lombardi, D. P., and N. C. Engleberg. 1992. Anaerobic bacteremia: incidence, patient characteristics and clinical significance. Am. J. Med. 92:53-60. [PubMed]
14. Mackowiak, P. A. 1997. Assaulting a physiologic response. Clin. Infect. Dis. 24:1208-1213. [PubMed]
15. Martin, M. A., M. A. Pfaller, and R. P. Wenzel. 1989. Mortality and hospital stay attributable to coagulase-negative staphylococcal bacteremia. Ann. Intern. Med. 110:9-16. [PubMed]
16. Miller, P. J., and R. P. Wenzel. 1987. Etiologic organisms as independent predictors of death and morbidity associated with bloodstream infection. J. Infect. Dis. 156:471-477. [PubMed]
17. Munson, E. L., D. J. Diekema, S. E. Beekmann, K. C. Chapin, and G. V. Doern. 2003. Detection and treatment of bloodstream infection: laboratory reporting and antimicrobial management. J. Clin. Microbiol. 41:495-497. [PMC free article] [PubMed]
18. Pittet, D., N. Li, R. F. Woolson, and R. P. Wenzel. 1997. Microbiological factors influencing the outcome of nosocomial bloodstream infections: a 6-year validated, population-based model. Clin. Infect. Dis. 24:1068-1078. [PubMed]
19. Pittet, D., S. Hugonnet, S. Harbarth, P. Mourouga, V. Sauvan, S. Touveneau, and T. V. Perneger. 2000. Effectiveness of a hospital-wide programme to improve compliance with hand hygiene. Lancet 356:1307-1312. [PubMed]
20. Roberts, F. J., I. W. Geere, and A. Coldman. 1991. A three-year study of positive blood cultures, with emphasis on prognosis. Rev. Infect. Dis. 13:34-46. [PubMed]
21. Sahm, D. F., M. K. Marsilio, and G. Piazza. 1999. Antimicrobial surveillance in key bloodstream bacterial isolates: electronic surveillance with the Surveillance Network Database-USA. Clin. Infect. Dis. 29:259-263. [PubMed]
22. Salonen, J. H., E. Eerola, and O. Meurman. 1998. Clinical significance and outcome of anaerobic bacteremia. Clin. Infect. Dis. 26:1413-1417. [PubMed]
23. Schaberg, D. R., D. H. Culver, and R. P. Gaynes. 1991. Major trends in the microbial etiology of nosocomial infection. Am. J. Med. 91:72S-75S. [PubMed]
24. Weinstein, M. P., M. L. Towns, S. M. Quartey, S. Mirrett, L. G. Reimer, G. Parmigiani, and L. B. Reller. 1997. The clinical significance of positive blood cultures in the 1990's: a prospective comprehensive evaluation of the microbiology, epidemiology, and outcome of bacteremia and fungemia in adults. Clin. Infect. Dis. 21:584-602. [PubMed]
25. Wenzel, R. P., and M. B. Edmond. 2001. The impact of hospital-acquired bloodstream infections. Emerg. Infect. Dis. 7:174-177. [PMC free article] [PubMed]
26. Wey, S. B., M. Mori, M. A. Pfaller, R. F. Woolson, and R. P. Wenzel. 1988. Hospital acquired candidemia: attributable mortality and excess length of stay. Arch. Intern. Med. 148:2642-2645. [PubMed]

Articles from Journal of Clinical Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...