The Clinical and Prognostic Importance of Positive Blood Cultures in Adults

CLINICAL RESEARCH STUDY

The Clinical and Prognostic Importance of Positive Blood Cultures in Adults Brian C. Pien, MD,a,b Punidha Sundaram, MD,c Natalia Raoof, MD,c Sylvia F. Costa, MD,a,d Stanley Mirrett, MS,a Christopher W. Woods, MD,b,d,e L. Barth Reller, MD,a,d,e Melvin P. Weinstein, MDc,f a

Clinical Microbiology Laboratory, Duke University Medical Center, Durham, NC; bDurham Veterans Affairs Medical Center, NC; Department of Medicine, Robert Wood Johnson Medical School, New Brunswick, NJ; dDepartment of Medicine and eDepartment of Pathology, Duke University School of Medicine, Durham, NC; fDepartment of Pathology, Robert Wood Johnson Medical School, New Brunswick, NJ. c

ABSTRACT BACKGROUND: Bloodstream infections are a major cause of morbidity and mortality in adults. Bloodstream infections should be reassessed periodically because of increased antibiotic resistance, more patients receiving immunomodulatory therapy, improved antiretroviral therapy, and acquisition of infection in health care settings other than hospitals. METHODS: We conducted retrospective assessment by infectious disease physicians of hospitalized adults with positive blood cultures at 3 academic medical centers. RESULTS: Two thousand two hundred seventy positive blood culture episodes occurred in 1706 patients. Of 2669 isolates, 51% represented true infection, 41% contamination, and 8% unknown clinical significance. Although coagulase-negative staphylococci were most common, only 10% were clinically significant. Among 1225 true bloodstream infections, the most frequent isolates were Staphylococcus aureus, Escherichia coli, Enterococcus spp., Klebsiella pneumoniae, coagulase-negative staphylococci, Pseudomonas aeruginosa, Candida albicans, Enterobacter cloacae, and Serratia marcescens. Intravenous catheters were the most common primary source of bloodstream infection (23% of episodes). Most (81%) bloodstream infections were acquired in the hospital or other health care settings. Crude and attributable in-hospital case-fatality ratios were 20% and 12%, respectively, lower than in previous studies. Increasing age, hypotension, absence of fever, hospital acquisition, extreme white blood cell count values, and the presence of the acquired immunodeficiency syndrome, malignancy, or renal disease were significantly associated with an increased risk of in-hospital attributable death in multivariable analysis. CONCLUSIONS: The proportion of bloodstream infections due to intravenous catheters is continuing to increase. Most episodes were acquired in the hospital or other health care setting. In-hospital case-fatality ratios have decreased compared with previous studies. Several previously identified factors associated with an increased mortality remain statistically significant. © 2010 Elsevier Inc. All rights reserved. • The American Journal of Medicine (2010) 123, 819-828 KEYWORDS: Bacteremia; Bloodstream infection; Fungemia

Funding: None. Conflict of Interest: None. Authorship: All authors had access to the data and a role in writing the manuscript. This research was supported in part by a grant from the Department of Veteran Affairs Special Fellowship Program in Health Services Research. Requests for reprints should be addressed to Melvin P. Weinstein, MD, Department of Medicine, UMDNJ-Robert Wood Johnson Medical School, 1 Robert Wood Johnson Place, New Brunswick, NJ 08901. E-mail address: [email protected]

0002-9343/$ -see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2010.03.021

The clinical significance of positive blood cultures has been studied in detail over the past 3 decades.1-9 These studies have helped describe the most common etiologic microorganisms, sources of infection, and underlying conditions associated with death in true bloodstream infections. Sequential studies have documented the increasing importance of intravenous access devices as sources of bloodstream infection and reductions in case-fatality ratio (CFR) in the early 1990s compared with the mid-1970s.1-3 Since the early 1990s, more drug-resistant microorganisms have

820

The American Journal of Medicine, Vol 123, No 9, September 2010

emerged, more effective human immunodeficiency virus (prednisone ⱖ20 mg/day for 2 weeks), malignancy, HIV (HIV) therapies have been developed, and an increasing infection, hematopoietic stem cell transplantation, solid ornumber of patients are receiving immunomodulatory treatgan transplantation, cirrhosis, trauma, and recent surgery ments, including those following organ transplantation. Fur(within 2 weeks). The appropriateness of antimicrobial therthermore, a new category of acquisition has been proposed apy was assessed at 3 different time points: before the blood that distinguishes other health care culture flagged positive, after the settings from community- and hosGram stain result, and when the pital-acquired settings.10 Because final identification or susceptibilCLINICAL SIGNIFICANCE these developments might have a ity result was available. Use of substantial impact on the distribuadjunctive therapy (eg, drainage, ● More than three quarters (81%) of baction, characteristics, and outcomes catheter removal, or activated proteremias and fungemias are now acof adult bloodstream infections, the tein C infusion) and outcome (disquired in the hospital (46%) or other current study was undertaken to charged alive, attributable death, health care settings (35%). update our understanding of conand nonattributable death) also ● Intravenous catheters are now the sintemporary bacteremia and fungemia were recorded. Deaths were conand compare the findings with 2 presidered attributable to a true gle most common sources of bacteremia vious similarly designed studies.1-3 bloodstream infection if it was and fungemia (23% of episodes). felt by the physician reviewer ● Although coagulase-negative staphylothat the episode caused or signifMETHODS cocci are the most common isolates icantly contributed to death durfrom blood cultures (38% of isolates), ing hospitalization. Study Design only 10% represent true bacteremia. A multicenter retrospective study Study Definitions was performed on an open cohort An episode of bacteremia, fungeof adults who had at least one posmia, or mycobacteremia was defined by a positive blood itive blood culture at 3 academic hospitals (Robert Wood culture, which must have occurred ⬎2 days after any preJohnson University Hospital, Duke University Medical Center, and the Durham Veterans Affairs Medical Center) vious positive result unless it was clear to the investigator from January 1 through December 31, 2004. Institutional that the new positive was part of the same episode.3 Polymireview board approval was obtained at all 3 institutions. crobial episodes were defined as having ⬎1 clinically significant blood culture isolate occurring within 2 days of each other. Clinical significance was categorized as either Study Cohort true bloodstream infection, contamination, or unknown sigPatients who had blood culture isolates obtained from stannificance. These assessments were made based on the numdard bacterial, fungal, or mycobacterial blood culture autober of positive cultures, presence of a plausible source, and mated detection systems (BACTEC, BD Biosciences, clinical manifestations. Acquisition was categorized as Sparks, MD and BacT/ALERT, bioMérieux, Durham, NC) community-acquired, hospital-acquired (hospitalized ⱖ48 were identified by the clinical microbiology laboratory dahours at the time the blood culture was drawn), or other tabase at each institution. Bacterial and fungal isolates that health care-associated (eg, previous hospitalization within were referred from outside medical institutions were ex90 days, attending a hemodialysis or intravenous [IV] checluded from the study. All potential subjects were screened motherapy clinic within 30 days, home IV therapy or speby reviewing the hospital medical record. Eligible subjects cialized nursing care within 30 days, or resident of longwere aged ⱖ18 years and were hospitalized at one of the term care facility).10 The presence of infective endocarditis study institutions where access to in-depth information was determined by modified Duke criteria.11 Hypotension about the clinical management and outcome was available. was defined as having a systolic blood pressure ⱕ90 mm Hg at the time the blood culture was drawn; subsequent vasoData Collection pressor therapy or having urine output ⬍80 mL over a The medical chart of each subject was reviewed by an 4-hour period also were accepted. The appropriateness of infectious disease physician, and data were abstracted onto antimicrobial therapy was assessed based on the final antia standardized worksheet to assess the presence of 25 types microbial susceptibility results by Clinical Laboratory Stanof variables at the time the positive blood culture was dards Institute criteria, published consensus treatment reccollected. These included the medical institution, age, sex, ommendations, and pharmacologic properties.12-14 service specialty, place of acquisition, microorganism, clinical significance, source of infection, presence of endocarPrimary Outcome ditis, white blood cell count, blood pressure, body temperThe primary outcome of the study was in-hospital death ature, and underlying comorbidities. The preexisting attributable to a true bloodstream infection. Because submedical conditions of interest included diabetes mellitus, renal insufficiency (serum creatinine ⱖ2), corticosteroid use jects varied greatly in the duration of their subsequent hos-

Pien et al

Clinical and Prognostic Importance of Bloodstream Infections

pital stay, and post-discharge follow-up was not feasible for the vast majority of survivors, only in-hospital CFR was measured. The CFR was calculated as the number of subjects with a true bloodstream infection that died during hospitalization divided by the total number of true bloodstream infections.

Statistical Analysis For categorical variables and CFR, differences between proportions were tested using the ␹2 test; P-values ⱕ.05 were used as the threshold for significance and no adjustments were made for multiple comparisons. Bivariable analyses were conducted to identify associated risk factors for attributable CFR using logistic regression (Stata 9.2, StataCorp, College Station, Tex). Both 95% confidence intervals of the odds ratio estimate and P-values were presented. A multivariable logistic regression model was constructed (SAS 9.1, SAS Institute Inc., Cary, NC) with the same variables used in the multivariable analysis of the 1992 study,3 that is, increasing age, hospital acquisition, fungi or non-E. coli Enterobacteriaceae pathogens, polymicrobial infection, source (respiratory, bowel, peritoneal, or unknown), predisposing factors (malignancy, acquired immunodeficiency syndrome, or renal failure), hypotension, extremes of leukocyte count (⬍4000 or ⬎20,000/mm3), neutropenia (absolute neutrophil count ⬍1000/mm3), absence of fever (temperature ⬍38.0°C), and inappropriate antimicrobial therapy at ⱖ2 time points. Bootstrapped modeling was used to find variables that were found to have a reproducible statistically significant association in at least 60% of models. Stepwise multivariable regression was subsequently performed using the selected variables. Measures of associations were reported as adjusted odds ratios with 95% confidence intervals and P-values.

RESULTS Patient Characteristics A total of 2669 blood culture isolates from 2270 positive blood culture episodes in 1706 patients were reviewed. There were 1225 patients who developed a true bloodstream infection and had a median age of 60 years (interquartile range 50-72 years). Approximately 60% of adult patients with a true bloodstream infection were male, 59% were white, and 80% had at least one predisposing comorbid illness. The patients from the Durham Veterans Affairs Medical Center were older, with a higher proportion of diabetes mellitus compared with Duke University Medical Center and the Robert Wood Johnson University Hospital (Table 1).

Microbiology Of 2669 total blood culture isolates, 1364 (51%) represented true infection, 1101 (41%) contamination, and 204 (8%) unknown clinical significance (Table 2). The most frequently isolated microorganisms causing episodes of true

821

bloodstream infection were Staphylococcus aureus (23%), Escherichia coli (12%), Enterococcus spp. (9%), Klebsiella pneumoniae (9%), coagulase-negative staphylococci (CoNS; 8%), Pseudomonas aeruginosa (4%), Candida albicans (3%), Enterobacter cloacae (3%), Serratia marcescens (3%), and Bacteroides spp. (2%). Consistent with previous studies, ⬎90% of positive blood culture episodes due to S. aureus, E. coli and other Enterobacteriaceae, Streptococcus pneumoniae, ␤-hemolytic streptococci, P. aeruginosa, obligate anaerobic Gram-negative bacteria, Candida spp., and Mycobacterium spp. were judged to represent true bloodstream infection. Although CoNS grew from 38% of all positive blood cultures, only 10% of CoNS represented true bloodstream infection. Similarly, only 30% of viridans streptococci and very few Corynebacterium spp., Bacillus spp., Micrococcus spp., Lactobacillus spp., and Propionibacterium spp. represented true infection (Table 2). Over 15% of episodes due to Enterococcus spp., non-glucose fermenting Gram-negative bacilli (eg, Acinetobacter spp., Stenotrophomonas maltophilia, and Pseudomonas non-aeruginosa spp.), and viridans streptococci were of uncertain clinical significance. Among the 1225 true bloodstream infection episodes, 115 (9%) were polymicrobial, which was less than the 1975 study (18%; P ⬍.001)1 but similar to the 1992 study (9%).3

Source of Infection Approximately 71% of all true bloodstream infection episodes had an identifiable source, including 32% that were culture-proven and another 25% that were confirmed by localized clinical findings (Table 3). The source remained unknown in 29% of true bloodstream infections, which is similar to previous reports. IV catheters were the leading identifiable source of bloodstream infections, representing 23% of episodes. This is higher compared with 3% (14/500, P ⬍.001) in the 1975 study1 and 19% (161/843, P ⫽ .05) in the 1992 study.3 The genitourinary (12%) and respiratory (8%) tracts were the next leading identifiable sources. Gram-positive pathogens including S. aureus (30%), CoNS (16%), and Enterococcus spp. (11%), in addition to Candida spp. (15%) and K. pneumoniae (7%), were the most frequent cause of IV catheter bloodstream infections. In contrast, Enterobacteriaceae, including E. coli (44%) and K. pneumoniae (8%), in addition to Enterococcus spp. (8%) and P. aeruginosa (6%), were the most frequently isolated pathogens from genitourinary tract episodes. S. aureus (26%), S. pneumoniae (22%), and K. pneumoniae (7%) were recovered most frequently from respiratory tract episodes. Bloodstream infections from intra-abdominal foci were most frequently caused by E. coli (21%), K. pneumoniae (18%), Enterococcus spp. (13%), and obligate anaerobic Gram-negative bacilli (12%). S. aureus (86%) caused most bone- and joint-associated bloodstream infections.

822

The American Journal of Medicine, Vol 123, No 9, September 2010

Table 1 Characteristics of a Cohort of Adult Bloodstream Infection Episodes at Duke University Medical Center, the Durham VA Medical Center, and Robert Wood Johnson University Hospital During 2004 All Patients Age, median Sex Male Female Race White Nonwhite Place of acquisition Community Hospital Other health care setting Predisposing conditions (n) 0 1 2 3 4 Comorbid condition Diabetes mellitus HIV/AIDS Organ transplantation Malignancy Neutropenia Corticosteroid use Serum creatinine ⱖ2 Cirrhosis Trauma Recent surgery

%

60 years

DUMC

%

57 years

DVAMC

%

67 years

RWJUH

%

65 years

731 494

60 40

374 372

50 50

223 1

99 1

134 121

53 47

709 499

59 41

417 329

56 44

131 93

58 42

161 77

69 31

228 568 429

19 46 35

133 312 301

18 42 40

46 115 63

21 51 28

49 141 65

20 55 25

249 510 370 84 12

20 42 30 7 1

146 309 231 55 5

20 41 31 7 1

39 92 73 17 3

17 41 33 8 1

64 109 66 12 4

25 42 26 5 2

403 45 94 373 118 83 316 32 19 184

33 4 8 30 10 7 26 3 2 15

236 20 72 228 85 41 194 28 17 120

32 3 10 31 11 6 26 4 2 16

117 17 4 54 6 8 64 4 0 32

52 8 2 24 3 4 29 2 0 13

50 8 18 91 27 34 58 0 2 32

20 3 7 36 11 13 23 0 1 14

DUMC ⫽ Duke University Medical Center; DVAMC ⫽ Durham Veteran Affairs Medical Center; RWJUH ⫽ Robert Wood Johnson University Hospital; HIV ⫽ human immunodeficiency virus; AIDS ⫽ acquired immunodeficiency syndrome.

Health Care Acquisition Using the revised definition of health care acquisition setting,10 only 19% of bloodstream infections were community acquired, compared with 53% if the older definition had been used (Table 4). Hospital acquisition and acquisition in other health care settings represented 46% and 35% of episodes, respectively. Consistent with previous studies,1,3 S. pneumoniae remained mostly community acquired, whereas very few episodes due to Candida spp., CoNS, and P. aeruginosa were truly community acquired.1-3 Approximately 45% of S. aureus bacteremias were associated with exposure at other health care settings, predominantly recent hospitalization and attendance of an outpatient hemodialysis or intravenous therapy treatment center. Of the 13% of S. aureus bacteremias that were community acquired, 31% (13/42) were methicillinresistant, compared with 55% (145/263) of S. aureus bacteremias acquired in hospital or other health care settings (P ⫽ .004).

Case-fatality Ratio (CFR), Bivariable, and Multivariable Analysis The overall crude and attributable in-hospital CFRs were 20% and 12%, respectively (Table 5). The microorganisms

associated with highest attributable CFR were ␤-hemolytic streptococci (22%), Enterococcus spp. (19%), P. aeruginosa (17%), Enterobacteriaceae other than E. coli (14%), obligate anaerobic bacteria (14%), and Candida spp. (14%). In the bivariable analysis, the factors associated with attributable CFR were increasing age, hypotension, absence of fever, hospital acquisition, increasing number of predisposing comorbid conditions, cirrhosis, renal failure, corticosteroid use, extremes of white blood cell count, source (respiratory, bowel, or peritoneum), Enterococcus spp., organ transplantation service, and nonsusceptible antimicrobial therapy at ⱖ2 decision points (Table 6). IV catheters, genitourinary source, CoNS, community acquisition, and use of adjunctive therapy were associated with lower unadjusted odds ratio for attributable death. About 11% of all bloodstream infections were associated with inappropriate or nonsusceptible antimicrobial therapy at ⱖ2 decision points; such episodes were associated with a significantly increased attributable CFR in the bivariable analysis (Table 6). In the multivariable logistic regression model, the statistically significant factors associated with in-hospital death attributable to a bloodstream infection were increasing age, hypotension, absence of fever, hospital acquisition, extremes of

Pien et al Table 2

Clinical and Prognostic Importance of Bloodstream Infections

823

Microorganisms Isolated from Positive Adult Blood Cultures at DUMC, DVAMC, and RWJUH, 2004 Unknown Clinical Significance

Total Isolates

True Bloodstream Infection

Contaminant

Microorganism

n

n

%

n

%

Coagulase-negative staphylococci Staphylococcus aureus Enterococcus spp.* Viridans group streptococci Streptococcus pneumoniae ␤-hemolytic streptococci† Corynebacterium spp. Bacillus spp. Micrococcus spp. Lactobacillus spp. Other Gram-positive bacteria‡ Escherichia coli Klebsiella pneumoniae Enterobacter cloacae Serratia marcescens Proteus mirabilis Other Enterobacteriaceae Pseudomonas aeruginosa Stenotrophomonas maltophilia Acinetobacter baumanii Other Gram-negative bacteria§ Clostridium spp. Propionibacterium spp. Peptostreptococcus spp. Other Gram-positive anaerobic bacteria㛳 Bacteroides spp. Other Gram-negative anaerobic bacteria¶ Candida albicans Candida glabrata Other Candida spp.** Other fungi†† Mycobacterium spp.‡‡ All microorganisms

1005 339 203 98 26 32 86 33 14 10 13 175 118 46 42 25 62 52 11 15 22 25 35 13 4 35 8 46 32 30 7 7 2669

105 315 128 29 26 31 7 0 0 4 3 170 112 43 39 25 62 50 8 10 12 16 1 5 3 34 7 45 32 30 5 7 1364

10 93 63 30 100 97 8 0 0 40 23 97 95 93 93 100 100 96 73 67 55 64 3 38 75 97 88 98 100 100 71 100 51

828 4 23 54 0 0 76 33 14 6 9 1 1 0 0 0 0 2 0 0 5 6 33 4 1 0 0 0 0 0 1 0 1101

82 1 11 55 0 0 88 100 100 60 69 1 1 0 0 0 0 4 0 0 23 24 94 31 25 0 0 0 0 0 14 0 41

n 72 20 52 15 0 1 3 0 0 0 1 4 5 3 3 0 0 0 3 5 5 3 1 4 1 1 1 0 0 1 0 204

% 7 6 26 15 0 3 3 0 0 0 8 2 4 7 7 0 0 0 27 33 23 12 3 31 0 3 13 2 0 0 14 0 8

DUMC ⫽ Duke University Medical Center; DVAMC ⫽ Durham Veteran Affairs Medical Center; RWJUH ⫽ Robert Wood Johnson University Hospital. *Includes 101 E. faecalis, 80 E. faecium, 2 E. gallinarum, 1 E. avium, 19 Enterococcus spp. †Includes 17 Streptococcus agalactiae, 6 group G streptococci, 5 S. pyogenes, 3 group F streptococci, and 1 S. equinus. ‡Includes 3 Listeria monocytogenes, 3 Abiotrophia spp., 2 Aerococcus spp., 2 Rothia spp., 2 Gemella spp., and 1 Dermabacter hominis. §Includes 4 other Acinetobacter spp., 3 other Pseudomonas spp., 3 Neisseria spp., 2 Achromobacter xylosoxidans, 2 Haemophilus influenzae, 2 Ochrobacter anthropi, 1 Aeromonas hydrophila, 1 Burkholderia cepacia, 1 Capnocytophaga spp., 1 Haemophilus parainfluenzae, 1 Moraxella catarrhalis, and 1 Roseomonas spp. 㛳Includes 3 Eubacterium lentum, and 1 Actinomyces meyeri. ¶Includes 2 Fusobacterium spp., 2 Veillonella spp., 1 Desulfomonas pigra, 1 Prevotella spp., Porphyromonas spp., and 1 Wolinella spp. **Includes 14 C. tropicalis, 13 C. parapsilopsis, 2 C. krusei, and 1 C. lusitaniae. ††Includes 2 Cryptococcus neoformans, 2 Histoplasma capsulatum, 1 Fusarium spp., 1 Cladosporium spp., and 1 Paecilomyces spp. ‡‡Includes 3 M. mucogenicum, 2 M. avium-intracellulare complex, 1 M. tuberculosis, and 1 M. chelonae.

white blood cell count, and having either malignancy, HIV, or renal disease. A steadily increasing trend between age and increased risk of attributable death was observed after examining the relationship between the 2 variables (Figure 1). A similar relationship also was seen between increasing number of predisposing comorbid conditions and attributable mortality (Figure 2).

DISCUSSION Studies over the past 4 decades have documented both recurrent findings and evolving changes in the microbiology, epidemiology, and outcomes of adult bloodstream infections.1-9 Compared with the 2 similarly designed studies that were performed by members of our group, S. aureus and E. coli remain the most frequently isolated bloodstream pathogens.1-3 Approximately 40% of all positive blood cul-

824

The American Journal of Medicine, Vol 123, No 9, September 2010

Table 3

Enterobacteriaceae other than E. coli and Candida spp. bloodstream infections have been increasing, whereas bacteremias due to S. pneumoniae and viridans group streptococci appear to be decreasing at our institutions. Episodes of obligate anaerobic bacteremia represent about 5% of episodes, compared with 16% and 4% in the 1975 and 1992 studies, respectively.1,3 M. avium complex bacteremia also has decreased, likely due to the availability of highly active antiretroviral therapy for HIV infection and efficacy of prophylaxis for patients with CD4 counts below 100/mm3.15 Using a recently revised definition,10 only 19% of bloodstream infections were community acquired, and these were associated with a lower unadjusted odds ratio of attributable death. Although molecular strain typing was not performed, nearly one third of community-associated S. aureus bacteremias were methicillin resistant, thus confirming the growing overlap of methicillin resistance between community and health care S. aureus isolates.16-18 This observation also highlights the importance of empiric methicillin-resistant S. aureus coverage if S. aureus bacteremia is suspected, whether acquisition occurs in the community or health care setting. This is further underscored by the continued observation that inappropriate antimicrobial therapy at ⱖ2 decision points remained significantly associated with attributable mortality in the bivariable analysis. In the present study, the overall crude in-hospital CFR was lower compared with the 1975 study (42% [212/500]; P ⬍.001),2 but similar to the 1992 study (22% [190/843]; P ⫽ .2).3 The overall attributable in-hospital CFR was 12%, which also was lower than 19% (97/500; P ⬍ .001) in the 19752 and 17% (148/843; P ⬍.001) in the 1992 study.3 The

Sources of Bacteremia and Fungemia Number of Episodes Confirmed by Clinical Total Number Culture Evidence of Episodes %

Source

Intravenous catheter 58 Genitourinary 115 Respiratory 53 Bone or joint 39 Intra-abdominal abscess 29 Skin 15 Bowel or peritoneum 9 Biliary 9 Surgical wound 28 Other* 30 Unknown 0 Total 385

92 22 35 9 14 31 35 37 9 20 0 304

282 143 97 51 51 51 49 50 37 51 358 1225

23 12 8 4 4 4 4 4 3 4 29 100

*Includes 17 infected hemodialysis grafts or fistulas, 11 cases of endocarditis or pericarditis, 7 central nervous system infections, 5 infected implantable pacemaker or cardiac defibrillator devices, 3 eye infections, 2 infected left ventricular assist devices, and 1 case each of infected gastrostomy tube, bone marrow, odontogenic infection, suppurative thrombophlebitis, otitis externa, mastoiditis.

ture episodes represent contamination, mostly due to coagulase-negative staphylococci. Since the 1990s, IV catheters have become the most common identifiable source of bloodstream infection, and the relative proportion appears to be increasing. As a consequence of medical progress and extensive health care exposure, the relative frequency of

Table 4

Place of Acquisition of Microorganisms Causing Bacteremia and Fungemia Other Health Care Settings Recent Hospitalization

Hemodialysis or IV Clinic

Long-term Care Facility

Home IV Therapy

Community

Hospital

Microorganism

n

%

n

%

n

%

n

%

n

%

n

%

Staphylococcus aureus Coagulase-negative staphylococci Enterococcus spp. Viridans group streptococci Streptococcus pneumoniae ␤-hemolytic streptococci Other gram-positive bacteria Escherichia coli Non-E. coli Enterobacteriaceae Pseudomonas aeruginosa Other gram-negative bacteria Obligate anaerobic bacteria Fungi Mycobacteria All microorganisms

42 9 14 10 20 10 4 60 38 5 4 16 5 2 239

13 9 11 37 77 32 40 35 14 10 15 27 5 29 19

131 57 72 5 1 4 5 43 146 30 11 29 90 2 627

55 54 59 9 4 16 50 25 53 60 41 49 85 29 48

102 25 20 10 3 10 1 33 58 9 6 11 11 3 302

32 24 16 37 12 32 10 19 21 18 22 19 10 43 23

34 11 13 2 0 4 0 19 23 5 6 2 0 0 119

11 10 11 7 0 13 0 11 8 10 22 3 0 0 9

5 1 2 0 2 2 0 12 7 1 0 1 0 0 33

2 1 2 0 8 6 0 7 3 2 0 2 0 0 2

1 2 2 0 0 0 0 3 1 0 0 0 0 0 9

0 2 2 0 0 0 0 2 0 0 0 0 0 0 1

IV ⫽ intravenous.

Pien et al Table 5

Clinical and Prognostic Importance of Bloodstream Infections

825

Sources and Case-fatality Ratio (CFR) for Episodes of Unimicrobial Bacteremia and Fungemia, According to Microorganism Episodes

Microorganism Staphylococcus aureus

n

%

Crude CFR

Attributable CFR

n

n

%

%

Common Sources IV catheter (28%), bone or joint (14%), other (9%), respiratory (8%), surgical site (6%) IV catheter (53%) IV catheter (29%), genitourinary (13%), bowel or peritoneum (9%) Other (20%, esp. endocarditis), intraabdominal abscess (10%) Respiratory (81%), other (15%, esp. meningitis) Cutaneous (23%), IV catheter (12%), bone or joint (12%)

285

26

50

18

32

11

Coagulase-negative staphylococci Enterococcus spp.

87 85

8 8

10 28

11 33

2 16

2 19

Viridans group streptococci

20

2

2

10

1

5

Streptococcus pneumoniae

26

2

3

12

2

8

␤-hemolytic streptococci

26

2

7

27

6

23

5 148 225

0 13 20

0 16 50

0 11 22

0 14 32

0 9 14

Pseudomonas aeruginosa

42

4

9

21

5

12

Other Gram-negative bacilli Gram-positive anaerobic bacteria

20 17

2 2

2 3

10 18

1 2

5 12

Gram-negative anaerobic bacteria

32

3

7

22

5

16

37 45 4 6 1110

3 4 0 1 100

15 14 2 0 218

41 31 50 0 20

8 5 2 0 133

22 11 50 0 12

Other Gram-positive bacteria Escherichia coli Non-E. coli Enterobacteriaceae

Candida albicans Other Candida spp. Other fungi Mycobacteria All microorganisms

Genitourinary (45%), biliary (11%) IV catheter (16%), genitourinary (15%), respiratory (9%), biliary (7%) Respiratory (19%), genitourinary (19%), IV catheter (17%) Respiratory (32%), IV catheter (16%) Bowel (29%), intra-abdominal abscess (28%), cutaneous (12%) Bowel (34%), intra-abdominal abscess (25%), cutaneous (9%) IV catheter (43%), genitourinary (8%) IV catheter (49%) IV catheter (33%), respiratory (33%)

IV ⫽ intravenous.

improvement in outcome may be explained by earlier clinical recognition and initiation of appropriate antimicrobial therapy, improved ability of modern blood culture systems to detect low-grade bacteremia, or other improvements in the quality of medical care. A large US sepsis study based on National Hospital Discharge Survey between 1979 and 2000 also found that crude in-hospital mortality consistently fell from 28% to 18%, but that the population-adjusted annual incidence of sepsis increased from 83 to 240 per 100,000, and deaths due to sepsis increased from 22 to 44 per 100,000.19 This could signify a growing patient population with significant comorbid conditions that require more frequent hospitalization and, consequently, develop more frequent health care-related complications, including bloodstream infection. In the current study and in previous studies of bacteremia and fungemia done by our group1,2 as well as by others,4 the absence of fever was associated with increased mortality. The explanation for this intriguing observation is not known. It might be that lack of fever leads to delayed diagnosis because blood cultures are not obtained until later in the course of illness, resulting in delayed therapeutic intervention. However, an equally, if not more, plausible

explanation is the effect of temperature on host defenses. Indeed, studies of infection using a variety of vertebrate and mammalian models have demonstrated that infected animals either adjust body temperature upward (endotherms), seek higher ambient temperature (poikilotherms), or have higher survival rates at higher ambient temperatures (poikilotherms held at specific temperatures).19 Other studies demonstrate that neutrophil migration, T-cell proliferation, and production of interferon and other cytokines are enhanced in the presence of fever.19 Taken together, the data suggest that fever is an indicator of intact host defenses and that elevated temperatures have had adaptive value for survival in an evolutionary sense. Contrary to some previous reports, we could not find a significant association between attributable mortality from bloodstream infection and nonwhite race.20,21 We also did not detect any significant increased odds of attributable death with male sex, S. aureus, or P. aeruginosa, as in our earlier studies.2,3 The bivariable analysis, however, did show associated odds ratios of attributable death with organ transplantation, high-dose corticosteroid use, infective endocarditis, and increasing number of predisposing factors that were not found in previous studies.

826

The American Journal of Medicine, Vol 123, No 9, September 2010

Table 6 Bivariable and Multivariable Analysis of Factors Associated with Attributable Death Due to Episodes of Adult Bloodstream Infection During 2004

Demographic and clinical features Increasing age (per 1 year) Increasing age (per 10 years) Male Nonwhite race Hypotension Absence of fever Community acquisition Hospital acquisition Other health care acquisition White blood cell count ⬍4000 or ⬎20,000/mm3 Absolute neutrophil count ⬍1000/mm3 Nonsusceptible ABX therapy at ⱖ2 time points Adjunctive therapy Infective endocarditis Comorbid conditions Increasing number of conditions Diabetes mellitus HIV infection, non-AIDS AIDS Any HIV infection Hematopoietic stem cell transplantation Solid organ transplantation Any organ transplantation Acute leukemia or lymphoma Solid malignancy Any malignancy Cirrhosis Serum creatinine ⬎2 mg/dL Trauma Recent surgery Corticosteroid use Malignancy, AIDS, or serum creatinine ⬎2 mg/dL Microorganism Staphylococcus aureus Methicillin-resistant S. aureus Coagulase-negative staphylococci Enterococcus spp. Streptococcus pneumoniae Viridans streptococci ␤-hemolytic streptococci E. coli Non-E. coli Enterobacteriaceae P. aeruginosa Other nonfermentative gram-negative bacilli Obligate anaerobic bacteria Candida spp. Mycobacterium spp. Fungi or non-E. coli Enterobacteriaceae Unimicrobial bloodstream infection Polymicrobial bloodstream infection Source Intravascular catheter Respiratory

Unadjusted Odds Ratio

95% CI

P Value

1.02

1.01-1.03

.002

1.07 0.91 5.69 2.65 0.38 1.68 0.93 2.03 1.41 1.88 0.58 2.14

0.75-1.52 0.64-1.29 3.98-8.16 1.87-3.75 0.21-0.69 1.19-2.38 0.65-1.33 1.43-2.88 0.85-2.33 1.17-3.03 0.41-0.58 1.04-4.42

.7 .6 ⬍.001 ⬍.001 .001 .003 .7 ⬍.001 .2 .009 .002 .04

1.43 1.31 1.61 1.25 1.35 1.62 1.33 1.41 0.80 0.87 0.82 6.97 2.02 0.40 0.75 1.81 1.42

1.18-1.73 0.92-1.86 0.54-3.77 0.48-3.29 0.59-3.07 0.54-4.86 0.68-2.58 0.79-2.52 0.46-1.39 0.55-1.39 0.56-1.20 3.40-14.29 1.41-2.89 0.05-3.00 0.45-1.27 1.02-3.22 1.00-2.02

⬍.001 .1 .5 .6 .5 .4 .4 .2 .4 .6 .3 ⬍.001 ⬍.001 .4 .3 .04 .05

0.91 1.17 0.15 1.77 0.60 0.34 2.11 0.71 1.28 1.55

0.61-1.36 0.58-2.35 0.04-0.63 1.04-3.01 0.14-2.55 0.04-2.54 0.84-5.31 0.04-1.25 0.85-1.91 0.71-3.39

.6 .7 .01 .03 .5 .3 .1 .2 .2 .3

1.21 1.19 1.20 1.36 0.84 1.19

0.56-2.62 0.64-2.20 0.14-10.08 0.94-1.95 0.48-1.47 0.68-2.07

.6 .6 .9 .1 .5 .5

0.61 2.16

0.38-0.96 1.29-3.62

.03 .004

Adjusted Odds Ratio

95% CI

P Value

1.03 1.41

1.02-1.05 1.26-1.58

⬍.001 ⬍.001

4.46 3.38

3.14-6.32 2.46-4.65

⬍.001 ⬍.001

3.50

2.50-4.89

⬍.001

1.59

1.13-2.22

.007

1.74

1.25-2.40

⬍.001

Pien et al Table 6

Clinical and Prognostic Importance of Bloodstream Infections

827

Continued

Genitourinary Surgical wound infection Bowel or peritoneum Intra-abdominal abscess Biliary Bone or joint Skin Other Unknown source Respiratory, bowel, peritoneum, or unknown Hospital service General medicine General surgery Cardiothoracic surgery Other surgery Obstetrics and gynecology Solid organ transplant Hematopoietic stem cell transplant Any transplant

Unadjusted Odds Ratio

95% CI

P Value

0.24 2.05 2.93 0.28 0.79 0.96 1.58 0.99 1.22 1.97

0.09-0.59 0.92-4.57 1.51-5.69 0.07-1.18 0.31-2.04 0.41-2.29 0.75-3.32 0.44-2.22 0.84-1.75 1.40-2.77

.002 .08 .002 .08 .6 .9 .2 1.0 .3 ⬍.001

1.14 0.65 1.53 0.35 0.31 1.91 1.74 1.89

0.77-1.70 0.36-1.18 0.51-4.57 0.11-1.14 0.04-2.31 0.90-4.06 0.65-4.70 1.02-3.48

.5 .2 .4 .08 .2 .09 .3 .04

Adjusted Odds Ratio

95% CI

P Value

CI ⫽ confidence interval; ABX ⫽ antibiotic; HIV ⫽ human immunodeficiency virus; AIDS ⫽ acquired immunodeficiency syndrome.

In the multivariable analysis, increasing age was significantly associated with increased odds ratio for attributable in-hospital death and confirms a large observational longitudinal study using hospital discharge data in the US.22 The presence of hypotension, absence of fever, hospital acquisition, extreme values of white blood cell count, and the presence of malignancy, the acquired immunodeficiency syndrome, or elevated serum creatinine ⱖ2 also were associated with significant adjusted odds ratio for attributable CFR and were consistent with results from the 1992 study.3 Health care acquisition also has been statistically significantly associated with CFR in other studies.10,23,24

Figure 1 Relationship between age and risk of attributable death due to an adult bloodstream infection.

This study of more than 1100 adult patients with bacteremia and fungemia used methodology similar to 2 previous studies performed by our group and provides insight into the contemporary microbiology, epidemiology, and outcomes of adult bloodstream infections as well as comparisons over 3 decades. Although subjective assessments were made to determine whether a positive blood culture patient isolate represented true bloodstream infection and whether death was attributable to the episode, these studies avoided reliance on hospital discharge diagnosis coding and crude mortality. Moreover, the clinical definitions and variables stud-

Figure 2 Relationship between the number of comorbid conditions and risk of attributable death due to an adult bloodstream infection.

828

The American Journal of Medicine, Vol 123, No 9, September 2010

ied were consistent over the 3 decades of comparative studies. Hospital discharge coding may be biased toward increased diagnosis of sepsis and bloodstream infection in order to improve reimbursement.19 This study lacked a validated comorbidity index, which might have confounded our results. However, we collected data on eight of the 19 conditions of the modified Charlson score that might help account for a substantial proportion of any confounding.25 In summary, our findings not only delineate the current reality of adult bloodstream infections but also call attention to new challenges. For example, traditional infection control practices aimed at decreasing hospital-acquired infections need to be extended to all health care facilities because health care-associated infections including bacteremia and fungemia occur in diverse settings and not only during inpatient stays. Also, the importance of early diagnosis and intervention with effective antimicrobial therapy will likely be crucial to achieve further decreases in mortality associated with bloodstream infection.

8.

9. 10.

11.

12.

13.

14. 15.

ACKNOWLEDGMENTS We thank Lauren Lindblad and Shein Chung-Chow of the Duke Clinical Research Institute for their assistance with the multivariable logistic regression model. We appreciate all the efforts of the Duke University Medical Center, Robert Wood Johnson University Hospital, and Durham Veteran’s Affairs Medical Center Clinical Microbiology Laboratory staff.

References 1. Weinstein MP, Reller LB, Murphy JR, Lichtenstein KA. The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. I. Laboratory and epidemiologic observations. Rev Infect Dis. 1983;5:35-53. 2. Weinstein MP, Murphy JR, Reller LB, Lichtenstein KA. The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations, with special reference to factors influencing prognosis. Rev Infect Dis. 1983;5:54-70. 3. Weinstein MP, Towns ML, Quartey SM, et al. The clinical significance of positive blood cultures in the 1990s: a prospective comprehensive evaluation of the microbiology, epidemiology, and outcome of bacteremia and fungemia in adults. Clin Infect Dis. 1997;24:584-602. 4. Diekema DJ, Beekmann SE, Chapin KC, et al. Epidemiology and outcome of nosocomial and community-onset bloodstream infection. J Clin Microbiol. 2003;41:3655-3660. 5. Lark RL, Chenoweth C, Saint S, et al. Four year prospective evaluation of nosocomial bacteremia: epidemiology, microbiology, and patient outcome. Diagn Microbiol Infect Dis. 2000;38:131-140. 6. Lark RL, Saint S, Chenoweth C, et al. Four-year prospective evaluation of community-acquired bacteremia: epidemiology, microbiology, and patient outcome. Diagn Microbiol Infect Dis. 2001;41(1-2):15-22. 7. Wisplinghoff H, Bischoff T, Tallent SM, et al. Nosocomial bloodstream infections in US hospitals: analysis of 24,179 cases from a

16.

17.

18.

19.

20.

21.

22. 23.

24.

25.

prospective nationwide surveillance study. Clin Infect Dis. 2004;39: 309-317. Edmond MB, Wallace SE, McClish DK, et al. Nosocomial bloodstream infections in United States hospitals: a three-year analysis. Clin Infect Dis. 1999;29:239-244. Cockerill FR 3rd, Wilson JW, Vetter EA, et al. Optimal testing parameters for blood cultures. Clin Infect Dis. 2004;38:1724-1730. Friedman ND, Kaye KS, Stout JE, et al. Health care-associated bloodstream infections in adults: a reason to change the accepted definition of community-acquired infections. Ann Intern Med. 2002;137:791797. Li JS, Sexton DJ, Mick N, et al. Proposed modifications to the Duke criteria for the diagnosis of infective endocarditis. Clin Infect Dis. 2000;30:633-638. National Committee for Clinical Laboratory Standards. Performance Standards for Antimicrobial Susceptibility Testing, 14th informational supplement (M100-S14). Wayne, PA: National Committee for Clinical Laboratory Standards; 2004. Gilbert DN, Moellering RC, Eliopoulos GM, et al. The Sanford Guide to Antimicrobial Therapy, 34th edition. Sperryville, VA: Antimicrobial Therapy, Inc.; 2004. Mandell GL, Bennett JE, Dolin R. Principles and Practice of Infectious Diseases, 5th edition. Philadelphia: Elsevier; 2000. Kaplan JE, Benson C, Holmes KH, et al. Guidelines for prevention and treatment of opportunistic infections in HIV-infected adults and adolescents: recommendations from CDC, the National Institutes of Health, and the HIV Medicine Association of the Infectious Diseases Society of America. MMWR Recomm Rep. 2009;58(RR-4):1-207; quiz CE1-4. Moran GJ, Krishnadasan A, Gorwitz RJ, et al. Methicillin-resistant S. aureus infections among patients in the emergency department. N Engl J Med. 2006;355:666-674. King MD, Humphrey BJ, Wang YF, et al. Emergence of communityacquired methicillin-resistant Staphylococcus aureus USA 300 clone as the predominant cause of skin and soft-tissue infections. Ann Intern Med. 2006;144:309-317. Seybold U, Kourbatova EV, Johnson JG, et al. Emergence of community-associated methicillin-resistant Staphylococcus aureus USA300 genotype as a major cause of health care-associated blood stream infections. Clin Infect Dis. 2006;42:647-656. Kluger MJ, Kozak W, Conn CA, et al. The adaptive value of fever. In: Mackowiak P, ed. Fever: Basic Mechanisms and Management. Philadelphia: Lippincott-Raven Publishers; 1997:255-266. Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003;348:1546-1554. Esper AM, Moss M, Lewis CA, et al. The role of infection and comorbidity: Factors that influence disparities in sepsis. Crit Care Med. 2006;34:2576-2582. Martin GS, Mannino DM, Moss M. The effect of age on the development and outcome of adult sepsis. Crit Care Med. 2006;34:15-21. Shorr AF, Tabak YP, Killian AD, et al. Healthcare-associated bloodstream infection: a distinct entity? Insights from a large U.S. database. Crit Care Med. 2006;34:2588-2595. Crane SJ, Uslan DZ, Baddour LM. Bloodstream infections in a geriatric cohort: a population-based study. Am J Med. 2007;120:10781083. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373-383.