Risk Factors for Mortality of Bacteremic Patients in the Emergency Department

Abstract

Objectives:  Patients with bacteremia have a high mortality and generally require urgent treatment. The authors conducted a study to describe bacteremic patients in emergency departments (EDs) and to identify risk factors for mortality.

Methods:  Bacteremic patients in EDs were identified retrospectively at a university hospital from January 2007 to December 2007. Demographic characteristics, underlying illness, clinical conditions, microbiology, and the source of bacteremia were collected and analyzed for their association with 28-day mortality.

Results:  During the study period, 621 cases (50.2% male) were included, with a mean (±SD) age of 62.8 (±17.4) years. The most common underlying disease was diabetes mellitus (39.3%). Escherichia coli (39.2%) was the most frequently isolated pathogen. The most common source of bacteremia was urinary tract infection (41.2%), followed by primary bacteremia (13.2%). The overall 28-day mortality rate was 12.6%. Multivariate stepwise logistic regression analysis showed age > 60 years (odds ratio [OR] = 2.52, 95% confidence interval [CI] = 1.29 to 4.92, p = 0.007), malignancy (OR = 2.66, 95% CI = 1.44 to 4.91, p = 0.002), liver cirrhosis (OR = 2.08, 95% CI = 1.02 to 4.26, p = 0.044), alcohol use (OR = 5.73, 95% CI = 2.10 to 15.63, p = 0.001), polymicrobial bacteremia (OR = 3.99, 95% CI = 1.75 to 9.10, p = 0.001), anemia (OR = 2.33, 95% CI = 1.34 to 4.03, p = 0.003), and sepsis (OR = 1.94, 95% CI = 1.16 to 3.37, p = 0.019) were independent risk factors for 28-day mortality.

Conclusions:  Bacteremic patients in the ED have a high mortality, particularly with these risk factors. It is important for physicians to recognize the factors that potentially contribute to mortality of bacteremic patients in the ED.

Fever is one of the leading reasons for visiting emergency departments (EDs), and it is the chief complaint at 5% of all visits in the United States.¹ Blood culture is an important diagnostic tool to evaluate certain febrile patients in the ED. It is estimated that blood cultures are ordered in 2.8% of all ED visits, and about 3.1 million blood cultures are ordered annually in U.S. EDs.¹ Although indications for the use of blood cultures are available,^2,3 there are no generally acceptable guidelines for blood culture for patients with sepsis in the ED.

Patients who present with bacteremia in the ED have a high mortality and generally require urgent admission for further treatment with antimicrobial agents.⁴ Therefore, identification of bacteremic patients at risk for mortality is a critical issue in the ED. Furthermore, to guide empirical antimicrobial treatment, recognizing the most common pathogens responsible for bacteremia and the most frequent sites of infection causing bacteremia are crucial. However, few studies discuss the manifestations of bacteremia in the ED or evaluate the risk factors for their mortality.^5,6 Therefore, we performed this study with the following objectives: 1) describe the clinical characteristics, 2) identify the most frequently isolated microorganisms, 3) define the most common sources of bacteremia, 4) describe survival rates, and 5) determine the independent predictive factors for the 28-day mortality of bacteremic patients in the ED.

Methods

Study Design

We conducted a retrospective observational study at E-Da Hospital, a 1,000-bed university-affiliated hospital in southern Taiwan. This study was approved by the institutional review board of E-Da Hospital.

Study Setting and Population

The annual number of ED visits is about 60,000 at E-Da Hospital. Adult patients aged >16 years visiting the ED with positive blood cultures meeting the inclusion criteria from January 2007 through December 2007 were included in this study.

Study Protocol

The medical records of included patients were reviewed by two authors. If any discrepancy was found, the medical records were inspected again by these two authors together. Demographic information of age, sex, underlying illness, clinical condition, microbiology, source of bacteremia, and outcome were collected from the medical records. Clinical data were analyzed by case episodes. Episodes were considered distinct if separated by at least 7 days and if the cause for visiting the ED was different.

Two pairs of standard aerobic and anaerobic blood cultures were ordered by physicians if infection was suspected. If patients received antibiotics within 24 hours, BACTEC resin-containing blood culture bottles were used (Becton Dickinson Diagnostic Instrument System, Sparks, MD). Blood culture samples were processed by the BACTEC 9240 system. All positive cultures were Gram-stained and subcultured on blood agar and eosin-methylene blue agar plates for further identification. Biochemical tests and/or automatic identification systems were used for final identification. Susceptibilities to antimicrobial agents were determined by the standard disk diffusion method and extended-spectrum β-lactamase (ESBL) production was determined following the Clinical and Laboratory Standards Institute criteria (CLSI, formerly National Committee for Clinical Laboratory Standards).⁷ Blood cultures were regarded as contaminated if they grew skin flora, including coagulase-negative staphylococci, micrococci species, diphtheroids, Bacillus species, and Propionibacterium acnes in only one bottle of blood cultures.⁸ These case episodes were excluded from our study. Monomicrobial and polymicrobial infections were defined as one species, and two or more species were isolated, respectively. If the blood cultures yielded fungi or mycobacteria, the case episodes were excluded from our study.

The source of bacteremia was determined clinically on the basis of the presence of an active infection site coincident with bacteremia or isolation of the organism from other clinical specimens prior to or on the same date as the onset of bacteremia. If the source of bacteremia could not be attributed to any known source, it was classified as primary bacteremia. For infective endocarditis, modified Duke’s criteria were applied.⁹ For catheter-related bloodstream infection, semiquantitative tip culture (>15 colonies) and/or high clinical suspicion were adopted.¹⁰ Lower respiratory tract infection included pneumonia, chronic obstructive pulmonary disease with bacterial infection, lung abscess, and pulmonary empyema. Biliary tract infection included cholangitis, cholecystitis, and gallbladder empyema. Intraabdominal infection included peritonitis, intraabdominal abscess, and gastrointestinal infections, but not liver abscess. Intravascular infection included mycotic aneurysm and arteriovenous graft infection. According to the American College of Chest Physicians and Society of Critical Care Medicine (ACCP/SCCM) Consensus Conference,¹¹ sepsis was defined as presence of an infectious process associated with two or more of the following criteria: body temperature >38 or <36°C, heart rate >90 beats/min, respiratory rate >20/min or partial pressure of carbon dioxide (PaCO₂) <32 mm Hg, and white blood cell count >12 × 10⁹ or <4 × 10⁹/L, or >10% immature forms. Immunosuppressant therapy was defined as the use of cytotoxic agents or corticosteroids (>30 mg prednisolone daily or equivalent for 1 week or more). Anemia was defined as hemoglobin <10 g/dL. Neutropenia was defined as an absolute neutrophil count <0.5 × 10⁹/L. Thrombocytopenia was defined as a platelet count <100 × 10⁹/L.

Outcome Measures

The primary outcome was 28-day mortality. If patients were discharged within 28 days after admission and were not followed-up at our hospital, telephone contact was made to collect the required information. If patients were lost to follow-up, they were excluded from our study.

Data Analysis

The results were analyzed using the commercially available SPSS software package (Version 14.0, SPSS Inc., Chicago, IL) to test the difference between the surviving and expired groups. Categorical variables were analyzed using the chi-square test or Fisher exact tests, as appropriate. Continuous variables were analyzed using Student’s t-test. Univariate odds ratios (ORs) were computed by Mantel-Haenszel test. To identify risk factors for 28-day mortality, a multivariate regression model was built. All exploratory variables that could contribute to the 28-day mortality were added sequentially in a forward stepwise fashion (likelihood ratio). ORs, 95% confidence intervals (CI), and p-values were calculated for each risk factor. Kaplan-Meier curves were used for the 28-day survival analysis. All p-values were two-tailed, and a p-value of <0.05 was considered statistically significant.

Results

During the study period, there were 2,554 positive blood cultures among 62,715 ED visits. After excluding mycobacteria, fungi, contamination, and identical pathogens, 678 pathogens were isolated in 634 case episodes. Thirteen case episodes (2.1%) were lost to follow-up. A total of 664 pathogens isolated in 621 case episodes were included in our study. There were 312 males (50.2%) and 309 females (49.8%), with a mean ± SD age of 62.8 ± 17.4 years (range = 18–96 years). Diabetes mellitus was found in 39.3% of patients, followed by hypertension, malignancy, and liver cirrhosis (Table 1). Patients were found to have polymicrobial bacteremia in 6.3% of cases and to have Gram-negative bacteremia in 23.3%.

Table 1.    Demographic Characteristics, Underlying Illnesses, and Clinical Conditions of 621 Bacteremic Case Episodes in the ED

Characteristics	No. (%)
HIV = human immunodeficiency virus; COPD = chronic obstructive pulmonary disease. *Isolates n = 664. †Hemoglobin < 10 g/dL. ‡Absolute neutrophil count < 0.5 × 10^9/L. §Case episodes n = 612. ¶Platelet count < 100 × 10^9/L.
Age (yr), mean ± SD	62.8 ± 17.3
Sex, male	312 (50.2)
Underlying illness
Diabetes mellitus	244 (39.3)
End-stage renal disease	28 (4.5)
Malignancy	124 (20.0)
Liver cirrhosis	87 (14.0)
Autoimmune disease	10 (1.6)
HIV infection	5 (0.8)
COPD	25 (4.0)
Coronary artery disease	20 (3.2)
Hypertension	226 (36.4)
Congestive heart failure	38 (6.1)
Immunosuppressant therapy	39 (6.3)
Intravenous drug abuse	17 (2.7)
Cerebrovascular event	71 (11.4)
Bedsore	19 (3.1)
Alcohol use	44 (7.1)
Clinical condition
Monomicrobial bacteremia	582 (93.7)
Polymicrobial bacteremia	39 (6.3)
Gram-positive bacteremia*	155 (23.3)
Gram-negative bacteremia*	509 (76.7)
Anemia†	167 (26.9)
Neutropenia‡§	9 (1.5)
Thrombocytopenia¶	147 (23.7)
Sepsis	225 (36.2)

Univariate analysis of risk factors for 28-day mortality is shown in Table 2. There were no significant differences in mortality by age or sex. However, mortality rates were different in the presence versus absence of malignancy, liver cirrhosis, immunosuppressant therapy, alcohol use, polymicrobial bacteremia, anemia, thrombocytopenia, and sepsis.

Table 2.    Demographic Characteristics, Underlying Illnesses, and Clinical Conditions for 28-day Mortality of 621 Bacteremic Case Episodes in the ED

Factors	28-day Mortality		OR (95% CI)	p-value
	With Factor	Without Factor
HIV = human immunodeficiency virus; COPD = chronic obstructive pulmonary disease. *Hemoglobin < 10 g/dL. †Absolute neutrophil count < 0.5 × 10^9/L. ‡Case episodes n = 612. §Platelet count < 100 × 10^9/L.
Age > 60 yr	49/367 (13.4)	29/254 (11.4)	1.20 (0.73–1.95)	0.475
Sex, male	47/312 (15.1)	31/309 (10.0)	1.59 (0.98–2.58)	0.059
Underlying illness
Diabetes mellitus	27/244 (11.1)	51/377 (13.5)	0.80 (0.48–1.31)	0.366
End-stage renal disease	4/28 (14.3)	74/593 (12.5)	1.17 (0.40–3.46)	0.769
Malignancy	28/124 (22.6)	50/497 (10.1)	2.61 (1.56–4.35)	<0.001
Liver cirrhosis	28/87 (32.2)	50/534 (9.4)	4.59 (2.69–7.85)	<0.001
Autoimmune disease	2/10 (20.0)	76/611 (12.4)	1.76 (0.37–8.44)	0.364
HIV infection	1/5 (20.0)	77/616 (12.5)	1.75 (0.19–15.86)	0.490
COPD	5/25 (20.0)	73/596 (12.2)	1.79 (0.65–4.92)	0.227
Immunosuppressant therapy	10/39 (25.6)	68/582 (11.7)	2.61 (1.22–5.58)	0.021
Alcohol use	17/44 (38.6)	61/577 (10.6)	5.33 (2.75–10.33)	<0.001
Clinical condition
Polymicrobial bacteremia	13/39 (33.3)	65/582 (11.2)	3.98 (1.95–8.12)	<0.001
Anemia*	36/167 (21.6)	42/454 (9.3)	2.70 (1.66–4.39)	<0.001
Neutropenia†‡	3/9 (33.3)	72/603 (11.9)	3.69 (0.90–15.07)	0.086
Thrombocytopenia§	37/147 (25.2)	41/474 (8.6)	3.55 (2.17–5.81)	<0.001
Sepsis	39/225 (17.3)	39/396 (9.8)	1.92 (1.19–3.10)	0.007

Figures 1 and 2 show the sources and species of bacteremia, respectively. The most common source of bacteremia was urinary tract infection, and the most frequent isolate was Escherichia coli. Patients with liver abscess, urinary tract infection, and bone or joint infection had the lowest mortality rates (3.7, 4.7, and 4.8%, respectively), and patients with respiratory tract infection (40.4%) and intraabdominal infection (44.1%) had the highest mortality rates. The ORs for mortality regarding different sources of bacteremia are shown in Table 3. There was no significant difference in mortality rate between Gram-positive and Gram-negative bacteremia (p = 0.727; Table 4). No significant differences were discovered among the Gram-positive pathogens. However, a statistically higher 28-day mortality rate was discovered in patients with Klebsiella spp., Pseudomonas spp., and Aeromonas spp. bacteremia when compared with E. coli.

Figure 1.  Sources of infection among bacteremic patients in the ED. UTI = urinary tract infection; SSTI = skin and soft tissue infection; LRI = lower respiratory tract infection; BTI = biliary tract infection; ENT = ear, nose and throat; CNS = central nervous system.

Figure 2.  Isolated pathogens (n=664) of bacteremic ED patients. (a) ESBL-producing E. coli, 12/260 (4.6%); (b) ESBL-producing K. pneumoniae, 3/115 (2.6%); (c) oxacillin-resistant S. aureus (ORSA), 29/77 (37.7%); (d) ESBL-producing P. mirabilis, 2/19 (10.5%). ESBL = extended-spectrum β-lactamase.

Table 3.    Sites of Infection Associated with 28-day Mortality Among 621 Bacteremic Case Episodes in the ED

Sites of Infection	Mortality (%)	OR (95% CI)	p-value
Urinary tract	12/256 (4.7)	1 (reference)
Lower respiratory tract	21/52 (40.4)	13.77 (6.18–30.71)	<0.001
Biliary tract	4/51 (7.8)	1.73 (0.54–5.60)	0.360
Skin and soft tissue	7/58 (12.1)	2.79 (1.05–7.43)	0.040
Intraabdomen	15/34 (44.1)	16.05 (6.59–39.14)	<0.001
Bone/joint	1/21 (4.8)	1.02 (0.13–8.22)	0.988
Liver abscess	1/27 (3.7)	0.78 (0.10–6.26)	0.817
Infective endocarditis	1/15 (6.7)	1.45 (0.18–11.98)	0.729
Catheter-related	2/12 (16.7)	4.07 (0.80–20.65)	0.091
Central nervous system	0/3 (0)	—	—
Ear, nose, throat/mouth	0/5 (0)	—	—
Intravascular infection	1/5 (20.0)	5.08 (0.53–49.04)	0.160
Primary bacteremia	13/82 (15.9)	3.83 (1.67–8.78)	0.001

Table 4.    Pathogens Associated with 28-day Mortality Among 582 Case Episodes With Monomicrobial Bacteremia in the ED

Pathogens	Mortality (%)	OR (95% CI)	p-value
*Compared with Gram-negative pathogens.
Gram-positive*	17/142 (12.0)	1.11 (0.62–2.00)	0.727
Staphylococcus spp.	10/77 (13.0)	1 (reference)	—
Streptococcus spp.	6/55 (10.9)	0.82 (0.28–2.41)	0.719
Enterococcus spp.	0/7 (0)	—	—
Others	1/3 (33.3)	—	—
Gram-negative	48/440 (10.9)
Escherichia coli	15/235 (6.4)	1 (reference)	—
Klebsiella spp.	14/96 (14.6)	2.50 (1.16–5.42)	0.020
Pseudomonas spp.	4/19 (21.1)	3.91 (1.15–13.26)	0.029
Proteus mirabilis	0/15 (0)	—	—
Enterobacter spp.	1/12 (8.3)	1.33 (0.16–11.03)	0.790
Aeromonas spp.	3/12 (25.0)	4.89 (1.20–19.97)	0.027
Salmonella spp.	1/10 (10.0)	1.63 (0.19–13.73)	0.653
Acinetobacter spp.	2/8 (25.0)	4.89 (0.91–26.33)	0.065
Citrobacter spp.	1/6 (16.7)	2.93 (0.32–26.74)	0.340
Others	7/27 (25.9)	—	—

The overall mortality rates of all bacteremic cases in the ED were as follows: 3-day mortality, 4.0%; 7-day mortality, 6.8%; 14-day mortality, 10.3%; 21-day mortality, 11.4%; and 28-day mortality, 12.6%. The Kaplan-Meier survival curve is shown in Figure 3. More than half of the deaths (53.8%) occurred during the first week after the onset of bacteremia.

Figure 3.  Kaplan-Meier survival curve of bacteremic ED patients.

We tested the demographic data, underlying illnesses, and clinical conditions to identify independent risks for the 28-day mortality in a multivariate stepwise logistic regression model and found that age > 60 years, malignancy, liver cirrhosis, alcohol use, polymicrobial bacteremia, anemia, and sepsis were independent predicting factors (Table 5).

Table 5.    Multivariate Forward Stepwise Logistic Regression Model of Risk Factors for 28-day Mortality of Bacteremic Case Episodes in the ED*

Factors	OR (95% CI)	p-value
*Case episodes n = 612. †Hemoglobin < 10 g/dL.
Age > 60 yr	2.52 (1.29–4.92)	0.007
Malignancy	2.66 (1.44–4.91)	0.002
Liver cirrhosis	2.08 (1.02–4.26)	0.044
Alcohol use	5.73 (2.10–15.63)	0.001
Polymicrobial bacteremia	3.99 (1.75–9.10)	0.001
Anemia†	2.33 (1.34–4.03)	0.003
Sepsis	1.94 (1.16–3.37)	0.019

Discussion

Although bacteremia is a critical condition in the ED, there are few studies discussing bacteremia in adults visiting EDs. Javaloyas et al.⁴ reported that Gram-negative bacteria accounted for 65.0% of community-acquired bacteremia, and E. coli was the most frequent isolate (45.0%). Recently, Groeneveld¹² evaluated the risk factors for community-acquired infections in febrile medical patients. The most frequent source of bacteremia was the respiratory system, and Enterobacteriaceae (including E. coli) were the most frequent isolates. These findings are similar to those of our study. However, neither of these studies were ED population-based, and they did not thoroughly examine and analyze the underlying illness, microbiology, and prognosis of bacteremic patients.

Patients with underlying malignancy are in an immunocompromised state and easily infected by many microorganisms. Mortality appears to increase in cancer patients complicated with bacteremia, particularly when a clinical site of infection exists.^13,14 Liver cirrhosis and chronic alcohol use are also known to be associated with impaired immunity. In both cases, patients are frequently infected with highly virulent pathogens and consequently experience increased morbidity and mortality.^15,16 This defect in the immune system could explain the higher mortality rate in patients with these underlying diseases.

Polymicrobial infection is frequently found in immunocompromised patients or in intraabdominal and complicated soft tissue infection.¹⁴ As a combination of several antimicrobial agents is usually needed to treat patients with polymicrobial infections, it is important to recognize the risks for polymicrobial bacteremia. However, no comprehensive study has been performed to evaluate the risks for polymicrobial infection in the ED. Our study showed polymicrobial infection was an independent factor for mortality in bacteremic patients. Further study investigating polymicrobial bacteremia in the ED is necessary.

The lower respiratory tract is one of the most frequent infection-related causes of death.¹⁷ Probable pathogens were found in 5%–14% of pretreatment blood cultures in a large series of nonselected patients with community-acquired pneumonia.¹⁸ Patients with community-acquired pneumonia complicated with bacteremia were reported to have a higher mortality rate (27.6%) than those with no bacteremia (12.6%).¹⁹ Our study also showed high mortality rate (40.4%) in bacteremic patients with lower respiratory tract infection.

With respect to the bacteremia etiology, we observed a significantly lower mortality rate with E. coli bacteremia and a higher mortality rate with Klebsiella spp., Pseudomonas spp., and Aeromonas spp. bacteremia. This result is also similar to another study.¹² Moreover, Pedersen et al.²⁰ reported a 30-day mortality rate of 13% to 15% in patients with primary bacteremia. The incidence (13.2%) and the mortality rate (15.9%) of primary bacteremia in our study were close to those of this previous report.

Limitations

Data used in the analysis were collected from medical records. There may be inconsistencies among the completeness of these data concerning the chronic illnesses and the sources of bacteremia. All of the univariate and multivariate analyses were carried out in an exploratory fashion, rather than in pursuit of a specific hypothesis. Some risk factors may not have been detected or explored in our study. Further prospective studies with specific hypotheses are necessary to accurately discover these risk factors.

Conclusions

In this general survey of bacteremic patients in the ED, we found that old age, malignancy, liver cirrhosis, alcohol use, polymicrobial bacteremia, anemia, and sepsis were independent predicting factors for 28-day mortality in bacteremic ED patients. The mortality rate differed between patients with urinary tract infection and other sources of infections. We also found that patients with E. coli bacteremia had a lower mortality rate than did those with sepsis due to Klebsiella, Pseudomonas, and Aeromonas spp. It is important for physicians to recognize these factors that potentially contribute to mortality in bacteremic ED patients.