a b s t r a c t

Objectives: The objectives were to investigate the clinical characteristics of community-onset bacteremia in cirrhotic adults visiting the emergency department (ED), as well as the clinical impact of empirical antibiotics on their outcome.

Methods: Cirrhotic adults with community-onset bacteremia who visited the ED from January 2005 to December 2009 were included retrospectively. Clinical data and outcome were collected from the medical chart. The in vitro susceptibility was measured by the broth microdilution method.

Results: Of the 246 bacteremic episodes in Cirrhotic patients, the major sources of bacteremia included intraAbdominal infections (111, 45.1%), primary bacteremia (43, 17.5%), urinary tract infection (39, 15.9%), and

soft tissue infection (22, 8.9%). Of the 258 bacteremic pathogens identified, Escherichia coli (83 isolates, 33.7%) and Klebsiella pneumoniae (61, 23.6%) were the most common microorganisms. In the multivariate analysis, delayed appropriate antibiotic therapy (N 72 hours; odds ratio [OR], 4.29; P = .003), serum creatinine greater than 1.5 mg/dL at the ED (OR, 3.12; P = .005), severe sepsis (OR, 3.61; P = .01), Pittsburgh bacteremia score of at least 4 (OR, 2.66; P = .04), bacteremia due to pneumonia (OR, 5.44; P = .02), and a comorbidity of diabetes mellitus (OR, 3.54; P = .004) were independently associated with the 28-day mortality.

Conclusions: Focusing on cirrhotic adults with community-onset bacteremia, we emphasized that the cirrhosis severity is one of the critical factors when choosing empirical antimicrobial therapy and that the strategy of empirical therapy is warranted for cirrhotic adults with severe decompensation (Child’s C group). For critically ill patients, especially in those with Child’s C group, only piperacillin/tazobactam, ertapenem, or imipenem treat- ment was warranted because of susceptibility rate of greater than 90%.

(C) 2014

Introduction

The association between liver cirrhosis and bacteremia or inva- sive bacterial infections due to alterations in enteric flora growth

[1] and multifactorial impairment of the immune system [2,3] is well known. Bacteremia is a serious, life-threatening condition that is associated with high morbidity and mortality [4]. Previous reports found that an increased mortality rate was present among cirrhotic

? Funding: This study was partially funded by grant (NCKUH-10305018) from National Cheng Kung University Hospital and, in part, supported by a research grant from the Ministry of Science and Technology Taiwan (NSC102-2314-B-006-079).

?? Competing interests: none.

* Correspondence to: C.-C. Lee, MD, MSc, Department of Internal Medicine, National Cheng Kung University Hospital, No. 138, Sheng Li Rd, 704, Tainan, Taiwan. Tel.: +886 6 2353535×3596; fax: +886 6 2752038.

?? Correspondence to: W.-C. Ko, MD, Department of Internal Medicine, National Cheng Kung University Hospital, No. 138, Sheng Li Rd, 70403, Tainan, Taiwan. Tel.: +886 6 2353535×3596; fax: +886 6 2752038.

E-mail addresses: [email protected] (C.-C. Lee), [email protected] (W.-C. Ko).

patients with bacteremia, accounting for up to 55% of the in-hospital mortality rate [5,6].

Chronic hepatitis B virus (HBV) infection can lead to chronic and progressive liver disease and is also the major risk factor for the development of cirrhosis. Hepatitis C virus (HCV) infection is one of the most important causes of liver cirrhosis with a high worldwide health impact. In the Asia-Pacific region and sub-Saharan Africa, infection with HBV or HCV is particularly prevalent [7,8]. Therefore, clinicians frequently encounter cirrhotic patients in endemic areas of HBV or HCV infection; and ED clinicians are often the first health care professional to make the diagnosis of sepsis and to prescribe empirical antibiotics. In general, early administration of appropriate empirical antibiotics has repeatedly been shown to decrease mortality in patients with community-acquired and nosocomial Bloodstream infections [9,10]. Although there are several previous studies focusing on specific pathogens in cirrhotic patients, such as Clostridium or Aeromonas bacteremia [11,12], because of the lack of the updated distri- bution of bacteremia-causing microorganism and susceptibility among those with community-onset bacteremia, these reports offer limited

http://dx.doi.org/10.1016/j.ajem.2014.11.024

0735-6757/(C) 2014

information to clinicians who manage cirrhotic patients. Moreover, since 2009, the Clinical and Laboratory Standards Institute (CLSI) guidelines have suggested new cephalosporin interpretive criteria for Enterobacte- riaceae [13]. As previously reported, Enterobacteriaceae have always been a vast proportion of the bacteremia-causing microorganism in the community [14,15]. However, no study dealing with updated surveillance of in vitro susceptibility of cirrhotic patients with community-onset bacteremia has been reported. Therefore, to delineate the clinical features and offer optimal management of community-onset bacteremia in cirrhotic adults at the ED, this study analyzed the clinical characteristics, bacterial species, in vitro suscepti- bility data, and the clinical impact and appropriateness of empirical antibiotic treatment.

Materials and methods

Study design and population

A retrospective, observational cohort study was conducted between January 2005 and December 2009 at a medical center of approximately 1200 beds in southern Taiwan. There were approximately 65000 annual visits to the emergency department (ED) during the study period. The local institutional review board approved this study. The medical records of adults visiting the ED with positive blood cultures were reviewed, and cirrhotic adults were included in the study. The exclusion criteria were defined as the following: patients lacking hospitalization information or who were transferred before the ED visit, patients with hospital- onset bacteremia, patients with fungemia or mycobacteremia, patients with contaminated cultures from blood sampling, patients with bac- teremia before ED arrival, and patients lacking 28-day mortality after the onset of bacteremia. For eligible adults, demographic informa- tion, clinical characteristics, Pittsburgh bacteremia severity scores, comorbidities, laboratory data during the ED stay, duration and type of antimicrobial agents, microbiological results, sources of bacteremia, and further hospitalization data were retrieved from the medical re- cords. Their medical records were reviewed for the above clinical infor- mation by 2 of the authors. If any discrepancy was found, the medical records would be inspected jointly by the 2 authors.

As previously described [15], multiple bacteremic episodes in a patient were considered distinct events if they were separated by at least 7 days or if different causes were noted for the respective ED visits. The primary outcome was the 28-day mortality.

Microbiological studies

At the ED, nurses performed the blood sampling for at least 2 sets of blood cultures. Each set of blood samples consisted of an aerobic culture bottle and an anaerobic culture bottle, with approximately 5 to 8 mL of blood for each bottle. Following the collection, the culture bottles were transported immediately to the clinical laboratory department, loaded into the BACTEC 9240 system (Becton Dickinson and Company, Franklin Lakes, NJ), and incubated for 5 days or until the instrument detected bacterial growth. Biochemical tests and automatic identification systems were used for species identification. The in vitro susceptibility was determined by the broth microdilution method. To compare the alteration in the susceptibility from the old criteria to the new criteria, the results were interpreted following both the old breakpoints (M100-S18) recommended by CLSI in 2008 [16] and the updated breakpoints (M100-S24) recommended by CLSI in 2014 [13].

Definitions

Bacteremia was defined as microorganism growth in the blood culture after the exclusion of contamination during sampling. Blood cultures with the growth of potential contaminating patho- gens, such as coagulase-negative staphylococci, Propionebacterium

acnes, Peptostreptococcus, or Micrococcus or Bacillus species, were con- sidered to be contaminated in accordance with the previously described criteria [17]. Polymicrobial bacteremia was defined as the isolation of more than 1 microbial species from a single bacteremic episode. The antibiotic therapy was considered appropriate if the patient was treated by at least 1 drug to which the pathogen was susceptible in vitro or if the patient was not treated with antibiotic, and otherwise the treatment was considered inappropriate. A delayed appropriate antibiotic therapy was defined as the delayed use of appropriate drugs until more than 72 hours after bacteremia onset, whereas early appropriate antibiotic therapy was defined as the use of appropriate drugs less than or equal to 72 hours after bacteremia onset. Community-onset bacteremia, as previously described [18], indicated that the place of bacteremia onset was in the community and therefore included health care-associated bacteremia occurring in long-term care facilities and community- acquired bacteremia.

The severity of bloodstream infections at the time of blood sampling

was assessed by the Pittsburgh bacteremia score, a validated scoring system based on vital signs, mental status, mechanical ventilation, and the presence of cardiac arrest [19]. Cirrhosis of the liver was diagnosed by a quantitative scoring system derived from the appearance of the liver surface, liver parenchymal texture, intrahepatic blood vessel size, and splenic size in high-resolution real-time ultrasound [20]. To assess the severity of cirrhosis, patients were categorized as class A (5-6 points), B (7-9 points), or C (10-15 points) according to the Child- Turcotte-Pugh classification, as shown in Table 1 [21].

Underlying malignancy included both hematologic malignancies and solid tumors, whereas the definitions of comorbidities followed those previously described [22]. The sources of bacteremia were catego- rized as urinary tract infections, lower respiratory tract infections, intraabdominal infections, skin and soft tissue infections, or primary bacteremia, according to the definitions of the Centers for Disease Control and Prevention [23]. For abnormality of laboratory data, the cutoff values for C-reactive protein (100 mg/L), blood urea nitrogen (20 mg/dL), and serum creatinine (1.5 mg/dL) were defined as in the previous description [14].

Data analysis

The statistical analyses were performed using the Statistical Package for the Social Sciences for Windows (SPSS, Chicago, IL), Version 15.0. The continuous variables were expressed as the means +- standard deviations (SDs) and compared by Student t test. Categorical variables, expressed as the numbers and percentages, were compared by the ?² or Fisher exact test. Kaplan-Meier curves were utilized for the 28-day survival analysis. All significant variables (P values of b.05) in the univariate analysis were incorporated into a hierarchical logistic regres- sion model. A P value of b .05 was considered to be statistically significant.

Results

Demographics and clinical characteristics of the study cohort

During the 5-year period, medical records of 1933 patients with de- tected bacteremia at the ED were reviewed. Based on the inclusion and exclusion criteria, 246 bacteremic episodes with community-onset

Table 1

Child-Pugh-Turcotte classification

Variable 1 point 2 points 3 points

Encephalopathy Absent Medically controlled Poorly controlled

Ascites Absent Medically controlled Poorly controlled

Bilirubin (mg/dL) b 2.0 2.0-3.0 N 3.0

Albumin (g/dL) N 3.5 2.8-3.5 N 2.8

International normalized ratio b 1.7 1.7-2.2 N 2.2 Modified from the previous description by Schuppan and Afdhal [21].

Fig. 1. Patient flowchart.

bacteremia were enrolled, as shown in Fig. 1. Of these episodes, no patient revisited the ED within 7 days after the bacteremic onset, and none had the same causes of bacteremia for the respective ED visits within the 28 days after bacteremia. Thus, 246 cirrhotic adults were included in the present study. Their mean age was 62.7 years, and 167 patients (67.9%) were male. Following Child’s classification, most (103, 41.9%) of the patients belonged to class C, 88 (35.8%) belonged to class B, and 55 (22.4%) belonged to class A. There were 114 (46.3%) patients with HBV infection, 84 (34.1%) with HCV infection, and 10 (4.1%) with HBV and HCV coinfection. Major comorbidities included malignancy (97 patients, 39.4%), hypertension (53, 21.5%),

diabetes mellitus (41, 16.7%), Chronic renal insufficiency (37, 15.0%),

neurologic disorders (26, 10.6%), chronic pulmonary disease (13,

5.3%), congestive heart failure (11, 4.5%), and autoimmune diseases

(7, 2.8%), urologic disease (7, 2.8%), coronary artery diseases (6, 2.4%), human immunodeficiency virus infection (6, 2.4%), and psychological disease (6, 2.4%). Three major sources of bacteremia included intraabdominal infections (111 patients, 45.1%), primary bacteremia (43, 17.5%), and urinary tract infections (39, 15.9%). Of note, Spontaneous bacterial peritonitis accounted for the majority (64, 57.7%) of the intraabdominal infections.

The vast majority (230 patients, 93.4%) visited the ED from the com- munity, 8 (3.3%) were transferred from long-term health care facilities, and 8 (3.3%) from other hospitals. Most (188, 76.4%) patients were

Fig. 2. Species distribution of 258 bacteremic pathogens in cirrhotic adults with community-onset bacteremia, categorized as patients with 28-day mortality and survivors. The percentage data were the proportion of all indicated patients; and asterisk indicates P b .05, calculated by the ?² test.

biological data and antimicrob”>subsequently admitted to the wards; and 38 (15.4%), to the Intensive care units . Nineteen (7.7%) patients were discharged from the ED, and 1 (0.4%) was transferred to another hospital. Eight patients died during their ED stay, 45 patients died during hospitalization, and 5 died in the 28 days following discharge from the ED. Thus, the 28-day mortality was 23.8% (58 patients).

microbiological data and antimicrobial susceptibility

Because there were 12 episodes of polymicrobial bacteremia, a total of 258 isolates were obtained from 246 eligible patients. Their species distribution is shown in Fig. 2. Gram-negative aerobes predominated (196 isolates, 76.0%). Escherichia coli (83, 33.7%) and Klebsiella species (61, 23.6%) were the 2 major gram-negative aerobes, and Streptococcus species (32, 12.4%) were the major gram-positive aerobe. Only 6 anaerobes, including Clostridium (2 isolates), Bacteroides (2), and Prevotella species (2), were noted.

Based on the new (M100-S24) and old (M100-S18) criteria issued by the CLSI, the in vitro susceptibilities for the commonly used antibiotics in the ED are shown in Table 2 including the following: piperacillin, piperacillin/tazobactam, ampicillin/sulbactam, cefazolin, cefuroxime, ceftazidime, ceftriaxone, cefepime, ertapenem, imipenem, and levofloxacin. Of the antibiotics with altered interpretive criteria (eg, cefazolin, ceftazidime, ceftriaxone, cefepime, ertapenem, and imipenem), the decreased susceptibilities of all of the bacteremia- causing microorganisms, gram-negative bacilli, and Enterobacteriaceae were discovered. Of note, methicillin-resistant S aureus accounted for 33.3% (7/21) of the S aureus isolates; and extended-spectrum ?-lactamase production was evident in 12.2% (10/82) of the E coli and 3.3% (2/61) of the K pneumoniae isolates.

Clinical characteristics, empirical antibiotics, pathogens, and clinical outcomes

The association between several variables and the 28-day mortality are presented in Table 3. Bacteremia due to respiratory tract infections, leukocytosis, blood urea nitrogen greater than 20 mg/dL, and serum creatinine greater than 1.5 mg/dL at the ED were significantly associated with the 28-day mortality. In addition to empirically inappropriate antibiotic therapy at the ED, a delayed appropriate antibiotic therapy, an initial syndrome with severe sepsis or septic shock, a Pittsburgh bacteremia score of at least 4 points, and the requirement of ICU care or hospitalization were associated with the 28-day mortality (Table 4). In

contrast, Streptococcus bacteremia heralded a better outcome (Fig. 2). In the multivariate regression, the following factors at the ED were indepen- dently linked to the 28-day mortality: delayed appropriate antibiotic therapy, Pittsburgh bacteremia score of at least 4 points, initial syndrome with severe sepsis, bacteremia due to respiratory tract infections, under- lying diabetes mellitus, and serum creatinine greater than 1.5 mg/dL (Table 5). Of note, delayed appropriate antibiotic therapy (odds ratio [OR], 4.29; 95% confidence interval [CI], 1.65-11.16) and bacteremic pneumonia (OR, 5.44; 95% CI, 1.25-23.71) were the most powerful risk factors. In a further analysis of Kaplan-Meier survival curve, a significant difference (P = .002) between treatment with delayed appropriate antibiotics and treatment with early appropriate antibiotics was demon- strated in cirrhotic patients with severe decompensation (class C of Child-Pugh score); this difference was not significant (P = .11) in cirrhotic patients with mild-to-moderate hepatic decompensation (class A and B), as shown in Fig. 3.

Discussion

Several investigations have demonstrated the clinical impact of delayed administration of appropriate antibiotics to patients with severe infections, such as bacteremia [14], or severe sepsis [24]. However, to our knowledge, there is no study focusing on cirrhotic adults with bacteremia and the impact of early appropriate therapy on the patient’s outcome in the English literature. In the multivariate regression analysis, we found that delayed appropriate antibiotic therapy was the most significant prognostic factor in the cirrhotic adults with community-onset bacteremia. In further analysis, its respective impact on the outcome of the patients with Child’s classes A plus B and patients with class C was evident in the Kaplan-Meier survival curve. Therefore, we emphasize that the cirrhosis severity is one of the critical factors to be considered when choosing empirical antimicrobial therapy and that the strategy of empirical therapy for bloodstream infections is warranted for cirrhotic adults with Child’s C group.

Because of its low simplicity and fairly good predictive value, the Child-Pugh-Turcotte classification is widely used for prediction of long-term outcome [21,25]. For 1-Year survival rate, patients with Child’s A, B, and C cirrhosis are 100%, 80%, and 45%, respectively [25]. Dissimilar to these reports, the different impact of delayed appropriate antibiotic on short-term mortality of this classification was firstly emphasized in the present study. Thus, for ED clinicians in their approach to a cirrhotic adult with a suspected bacteremia, several laboratory data (serum albumin, bilirubin, and international

Table 2

In vitro susceptibility of 11 drugs for bacteremic microorganisms from cirrhotic adults with community-onset bacteremia, according to the old^a or new^b criteria issued by the CLSI Drugs Susceptibility rate (%)

All N = 258 Gram-negative

bacilli n = 196

Enterobacteriaceae

n = 186

E coli n= 83 K pneumoniae

n= 61

Old New Old New Old New Old New Old New

Levofloxacin	82.9 -c 86.2 -c 87.6 -c 84.1 -c 91.8 -c
Ampicillin/sulbactam	70.9 -c 67.3 -c 66.7 -c 67.1 -c 72.1 -c
Piperacillin	80.2 -c 79.6 -c 79.0 -c 70.7 -c 90.2 -c
Piperacillin/tazobactam	91.9 -c 93.9 -c 93.5 -c 93.9 -c 95.1 -c

Cefazolin	74.8	66.7	75.0	64.3	76.9	64.3	81.7	68.3	85.2	77.0
Ceftazidime	80.6	78.3	89.3	86.2	91.4	86.2	89.0	85.4	96.7	93.4

Cefuroxime 80.2 -^c 83.2 -^c 84.9 -^c 86.6 -^c 83.6 -^c

Ceftriaxone	82.2	81.8	84.7	84.2	87.1	86.6	82.9	82.9	90.2	88.5
Cefepime	90.3	87.2	95.4	91.3	95.3	93.0	98.8	93.9	98.4	93.4
Ertapenem	93.8	91.9	96.4	93.9	98.4	95.7	100	100	96.7	90.2
Imipenem	95.3	93.4	99.0	96.4	98.9	96.2	100	100	98.4	93.4
a Based on CLSI issued in 2008.

^b Based on CLSI issued in 2014.

^c The interpretive criteria did not change.

Table 3

Clinical and microbiological characters associated with 28-day mortality among cirrhotic adults with community-onset bacteremia at the ED

Characteristics	Case no. (%)		OR (95% CI)	P
	Nonsurvivors, n = 58	Survivors, n = 188
Age (y), mean +- standard deviation	63.4 +- 15.9	62.4 +- 13.7	-		.64
Male sex	40 (69.0)	127 (67.6)	1.07 (0.57-2.01)		.84
Polymicrobial bacteremia	5 (8.6)	7 (3.7)	2.44 (0.74-8.00)		.16
LTHF residents	4 (6.9)	4 (2.1)	3.41 (0.83-14.08)		.09
Major source of bacteremia
Intraabdominal infection	25 (43.1)	86 (45.7)	0.90 (0.50-1.63)		.72
Liver abscess	0 (0)	8 (4.3)	-		.11
Biliary tract infection	2 (3.4)	13 (6.9)	0.48 (0.11-2.20)		.53
Spontaneous bacterial peritonitis	17 (29.3)	47 (25.0)	1.24 (0.65-2.39)		.51
Urinary tract infection	9 (15.5)	30 (16.0)	0.97 (0.43-2.18)	1.00
Respiratory tract infection	8 (13.8)	8 (4.3)	3.60 (1.29-10.07)	.03
Primary bacteremia	7 (12.1)	36 (19.1)	0.58 (0.24-1.38)	.22
Skin and soft tissue infection	7 (12.1)	15 (8.0)	1.58 (0.61-4.09)	.34
Bone/joint infection	1 (1.7)	6 (2.1)	0.53 (0.06-4.51)	1.00
Major comorbidity
Malignancy	26 (44.8)	71 (37.8)	1.34 (0.74-2.43)	.34
Diabetes mellitus	14 (24.1)	27 (14.4)	1.90 (0.92-3.92)	.08
Hypertension	12 (20.7)	41 (21.8)	0.94 (0.45-1.93)	.86
Chronic renal insufficiency	10 (17.2)	27 (14.4)	1.24 (0.56-2.75)	.59
Neurological disorder	7 (12.1)	19 (10.1)	1.22 (0.49-3.07)	.67
Chronic pulmonary diseases	4 (6.9)	9 (4.8)	1.47 (0.44-4.97)	.51
Coronary artery diseases	3 (5.2)	3 (1.6)	3.36 (0.66-17.14)	.15
Congestive heart failure Laboratory data at the EDa Leukocyte N 9000/mm3	2 (3.4) 36 (62.1)	9 (4.8) 85 (45.2)	0.71 (0.15-3.39) 1.98 (1.09-3.63)	1.00 .03
Platelet b100000/mm3	37 (63.8)	128 (68.1)	0.83 (0.45-1.53)	.54
C-reactive protein N 100 mg/L	13/49 (26.5)	29/164 (17.7)	1.68 (0.79-3.56)	.17
Blood urea nitrogen N 20 mg/dL	44/56 (78.6)	74/176 (42.0)	5.05 (2.50-10.23)	b.001
Serum creatinine N 1.5 mg/dL	33 (56.9)	39 (20.7)	5.04 (2.69-9.45)	b.001
LTHF = long-term health care facility.

^a Not all patients had the indicated laboratory data.

normalized ratio) and physical examination (Conscious level and existence of ascites) should be rapidly measured to evaluate the degree of liver decompensation.

The increase in mortality and morbidity due to infections in cirrhotic patients is well recognized and is associated with several abnormalities of the defense mechanisms [2,26]. There are several investigations of the clinical characteristics and outcomes of bacteremia in cirrhotic patients [5,27]. In accordance with these studies, which have reported a bacteremia-related fatality from 18% to 54.8%, we found a high crude mortality (23.8%) in the current study. The distribution of bacteremic isolates in our population was similar to a previous report of cirrhotic patients with community-onset bacteremia in Taiwan [5]. With regard to the predominance of Enterobacteriaceae, E coli and K pneumoniae were 2 major gram-negative aerobes; in addition, the Aeromonas and Vibrio species also remained important.

Although the clinical characteristics and outcomes of bacteremia in cirrhotic patients have been well discussed in previous reports [5,27,28], these reports are out of date and their information (eg, bacteremic isolate and in vitro susceptibility) is not sufficiently timely to assist ED clinicians. Moreover, the current documentation from the CLSI suggests using the new interpretive criteria for susceptibility testing for Enterobacteriaceae. The predominance of Enterobacteriaceae in cirrhotic patients with bacteremia provided ample reason to update the susceptibility data. Therefore, the changes of susceptibilities according to the previous and updated criteria were also used to establish the guideline of empirical therapy for bloodstream infections in the current study. Based on our data, the antibiotics with changed interpretive criteria (eg, cefazolin, ceftazidime, ceftriaxone, cefepime, ertapenem, and imipenem) had increased resistance in bacteremia-causing microor- ganisms, gram-negative bacilli, and Enterobacteriaceae. Therefore,

Table 4

Comparisons of disease severity and antimicrobial therapy among surviving and fatal cirrhotic adults with community-onset bacteremia at the ED

Characteristics	Case number (%)		Odds ratio (95% C.I.)	P value
	Nonsurvivors, n = 58	Survivors, n = 188
Child-Pugh score, class C	32 (55.2)	71 (37.8)	2.03 (1.12-3.68)	.02
Initial presentation at the ED
Severe sepsis	50 (86.2)	84 (44.7)	7.74 (3.48-17.22)	b.001
Septic shock	35 (60.3)	32 (17.0)	7.42 (3.88-14.20)	b.001
Severity-of-illness marker at the ED Pittsburgh bacteremia score >=4 points	25 (43.1)	20 (10.6)	6.36 (3.17-12.77)	b.001
Hospitalization	48 (82.8)	178 (94.7)	0.27 (0.11-0.69)	.01
Length of stay (d), mean +- SD	9.5 +- 8.6	16.5 +- 17.9	-	b.001
ICU admission	21 (36.2)	17 (9.0)	5.71 (2.75-11.87)	b.001
Length of stay (d), mean +- SD	3.3 +- 6.9	0.6 +- 2.8	-	.006
Antibiotic therapy Inappropriateness during the ED stay	19 (32.8)	39 (20.7)	1.86 (0.97-3.57)	.06
No antibiotic prescribed in the ED	1 (1.7)	12 (6.4)	0.26 (0.03-2.02)	.31
Delayed appropriate antibiotic therapya	15 (25.9)	19 (10.1)	3.10 (1.46-6.60)	.002

^a Receipt of appropriate antibiotics at more than 72 hours after bacteremia onset.

Table 5

Multivariate analysis of the risk factors for 28-day mortality among cirrhotic adults with community-onset bacteremia at the ED

Characteristics	OR (95% CI)	P
Delayed appropriate antibiotic therapya	4.29 (1.65-11.16)	.003
Comorbidity with diabetes mellitus	3.54 (1.48-8.45)	.004
Serum creatinine N 1.5 mg/dL at the ED	3.12 (1.40-6.94)	.005
Pittsburgh bacteremia score >=4 points at the ED	2.66 (1.06-6.68)	.04
Severe sepsis at the ED Bacteremia	3.61 (1.36-9.57)	.01

Due to respiratory tract infections 5.44 (1.25-23.71) .02

Due to Streptococcus species 0.27 (0.05-1.49) .13

^a Receipt of appropriate antibiotics at more than 72 hours after bacteremia onset.

focusing on critically ill patients, especially in those with severe decom- pensation (class C of Child-Pugh score), only piperacillin/tazobactam, ertapenem, or imipenem treatment was warranted because their sus- ceptibility rate was more than 90%. Of note, the cefazolin susceptibility of all bacteremic isolates decreased from 74.8% in the old criteria to 66.7% in the new criteria; thus, its empirical role remains debatable.

In addition to inappropriate empirical therapy, other prognostic factors were identified for cirrhotic patients with bacteremia by the multivariate analysis. The 2 most significant factors related to the severity of bacteremic onset were the high Pittsburgh bacteremia score and severe sepsis. The value of Pittsburgh bacteremia score in evaluating the severity of illness in bacteremia due to E coli [29], K pneumoniae [19,29], or streptococci [30] has been reported. These microorganisms were major bacteremic pathogens in our population; and therefore, it is not surprising to find that the Pittsburgh bacteremia score was an important predictor of mortality in our study. An additional prognostic factor was high serum creatinine at the onset of bacteremia; this finding was compatible with the results of our previous ED-based study [31]. Another factor was airway infection as the source of bacteremia, and this was compatible with the coexistence of bacteremia and pneumonia with high risk of short-term mortality established in a previous finding

[32].A significant host factor was underlying diabetes mellitus, an immu- nocompromised status, which is reasonable to produce a poor prognosis.

Limitation

We interpreted our data in light of several limitations inherent in the study design. First, this study is a retrospective, observational cohort study; and data used in the analysis were collected from medical re- cords, which carried the potential for recall bias and selection bias. Therefore, the strength of our study could be influenced by these biases. Second, during the study period, there were a total of 63106 patients with blood culture sampling; and the bacterial growth of blood cultures was noted in 3934 (4.9%) patients. The majority (1933 patients) had true bacteremia; 1135 had contaminated blood cultures. Thus, the abso- lute contamination rate of blood cultures was 1.8% (1135 of 63,106); and true bacteremic rate was 3.1% (1933 of 63,106). These incidences were similar to our previous report [33]. Third, this study population was composed of cirrhotic adults with detected bacteremia and was not focused on all cirrhotic patients with bacterial infections. In addi- tion, the clinical presentation of bacteremia in cirrhotic patients was dif- ferent from that in the general population; and the predictor of bacteremia in cirrhotic patients was poor in the literature [21]. Thus,a further prospective study with specific hypothesis is necessary to accu- rately discover these predictors. Finally, the study was conducted in a single ED in a tertiary hospital; and therefore, especially the susceptibil- ity data, our findings may not be generalizable to other communities. Nevertheless, our finding was sufficient to offer useful clinical informa- tion for the establishment of antibiotic guidelines in the Taiwan ED.

Conclusion

For cirrhotic adults with community-onset bacteremia in the ED, the distribution of bacteremia-causing microorganisms, in vitro updated susceptibilities, and resources of bacteremia were demonstrated. A delayed appropriate antibiotic therapy was the most important prog- nostic factor in the cirrhotic adults with community-onset bacteremia,

Fig. 3. The Kaplan-Meier survival analysis of cirrhotic adults with mild-to-moderate hepatic decompensation (class A and B of Child-Pugh score) and with severe decompensation (class C) treated by delayed appropriate antibiotics (ie, length of delayed appropriateness more than 72 hours after onset of bacteremia) or early appropriate antibiotics (ie, use of appropriate drugs at less than or equal to 72 hours after bacteremia onset).

especially for patients with Child-Pugh class C. Focusing on those with critical illness, piperacillin/tazobactam, ertapenem, or imipenem treat- ment was warranted in the era of increasing Antimicrobial resistance. And notably, the empirical role of cefazolin remained debatable because of the susceptibility rate of less than 70%. To avoid a delayed appropriate antibiotic therapy, guidelines for empirical antibiotic therapy based on our findings should be established.

References

1. Bauer TM, Schwacha H, Steinbruckner B, Brinkmann FE, Ditzen AK, Aponte JJ, et al. Small intestinal bacterial overgrowth in human cirrhosis is associated with systemic endotoxemia. Am J Gastroenterol 2002;97:2364-70.
2. Fierer J, Finley F. Deficient serum bactericidal activity against Escherichia coli in patients with cirrhosis of the liver. J Clin Invest 1979;63:912-21.
3. Rimola A, Soto R, Bory F, Arroyo V, Piera C, Rodes J. Reticuloendothelial system phagocytic activity in cirrhosis and its relation to bacterial infections and prognosis. Hepatology 1984;4:53-8.
4. Bates DW, Pruess KE, Lee TH. How bad are bacteremia and sepsis? Outcomes in a cohort with suspected bacteremia. Arch Intern Med 1995;155:593-8.
5. Kuo CH, Changchien CS, Yang CY, Sheen IS, Liaw YF. Bacteremia in patients with cirrhosis of the liver. Liver 1991;11:334-9.
6. Thulstrup AM, Sorensen HT, Schonheyder HC, Moller JK, Tage-Jensen U. Population- based study of the risk and short-term prognosis for bacteremia in patients with liver cirrhosis. Clin Infect Dis 2000;31:1357-61.
7. Lee WM. Hepatitis B, virus infection. N Engl J Med 1997;337:1733-45.
8. Lauer GM, Walker BD. Hepatitis C virus infection. N Engl J Med 2001;345:41-52.
9. Leibovici L, Shraga I, Drucker M, Konigsberger H, Samra Z, Pitlik SD. The benefit of appropriate empirical antibiotic treatment in patients with bloodstream infection. J Intern Med 1998;244:379-86.
10. Salomao R, Castelo Filho A, Pignatari AC, Wey SB. Nosocomial and community acquired bacteremia: variables associated with outcomes. Rev Paul Med 1993;111:456-61.
11. Chen YM, Lee HC, Chang CM, Chuang YC, Ko WC. Clostridium bacteremia: emphasis on the poor prognosis in cirrhotic patients. J Microbiol Immunol Infect 2001;34:113-8.
12. Lau SM, Peng MY, Chang FY. Outcomes of Aeromonas bacteremia in patients with different types of underlying disease. J Microbiol Immunol Infect 2000;33:241-7.
13. Clinical and Laboratory Standards Institute. Performance standards for antimicrobial disk susceptibility tests, 24th informational supplement. Document M100-S24. Wayne, PA: Clinical and Laboratory Standards Institute; 2014.
14. Lee CC, Lee CH, Chuang MC, Hong MY, Hsu HC, Ko WC. Impact of inappropriate empirical antibiotic therapy on outcome of bacteremic adults visiting the ED. Am J Emerg Med 2012;30:1447-56.
15. Lin JN, Tsai YS, Lai CH, Chen YH, Tsai SS, Lin HL, et al. risk factors for mortality of bac- teremic patients in the emergency department. Acad Emerg Med 2009;16:749-55.
16. Clinical and Laboratory Standards Institute. Performance standards for antimicrobial disk susceptibility tests, 18th informational supplement. Document M100-S18. Wayne, PA: CLSI; 2008.
17. Lee CC, Lin WJ, Shih HI, Wu CJ, Chen PL, Lee HC, et al. Clinical significance of potential contaminants in blood cultures among patients in a medical center. J Microbiol Immunol Infect 2007;40:438-44.
18. Cheong HS, Kang CI, Wi YM, Kim ES, Lee JS, Ko KS, et al. Clinical significance and predictors of community-onset Pseudomonas aeruginosa bacteremia. Am J Med 2008;121:709-14.
19. Paterson DL, Ko WC, Von Gottberg A, Mohapatra S, Casellas JM, Goossens H, et al. International prospective study of Klebsiella pneumoniae bacteremia: implications of extended-spectrum beta-lactamase production in Nosocomial infections. Ann Intern Med 2004;140:26-32.
20. Lin DY, Sheen IS, Chiu CT, Lin SM, Kuo YC, Liaw YF. Ultrasonographic changes of early liver cirrhosis in chronic hepatitis B: a longitudinal study. J Clin Ultrasound 1993;21: 303-8.
21. Schuppan D, Afdhal NH. Liver cirrhosis. Lancet 2008;371:838-51.
22. Schellevis FG, van der Velden J, van de Lisdonk E, van Eijk JT, van Weel C. Comorbidity of chronic diseases in General practice. J Clin Epidemiol 1993;46:469-73.
23. Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988;16:128-40.
24. Nguyen HB, Rivers EP, Abrahamian FM, Moran GJ, Abraham E, Trzeciak S, et al. Severe sepsis and septic shock: review of the literature and emergency department management guidelines. Ann Emerg Med 2006;48:28-54.
25. Infante-Rivard C, Esnaola S, Villeneuve JP. Clinical and statistical validity of conven- tional prognostic factors in predicting short-term survival among cirrhotics. Hepatology 1987;7:660-4.
26. Hassner A, Kletter Y, Shlag D, Yedvab M, Aronson M, Shibolet S. Impaired monocyte function in liver cirrhosis. Br Med J (Clin Res Ed) 1981;282:1262-3.
27. Almdal T, Skinhoj P, Friis H. Bacteremia in patients suffering from cirrhosis. Infection

    1986;14:68-70.

    Jones EA, Sherlock S, Crowley N. Bacteraemia in association with hepatocellular and hepatobiliary disease. Postgrad Med J 1967;43(Suppl 43):7-11.

28. Lee CI, Lee NY, Yan JJ, Lee HC, Ko NY, Chang CM, et al. Extended-spectrum beta- lactamase-producing phenotype signifies a poor prognosis for patients with cefpodoxime-resistant Escherichia coli or Klebsiella pneumoniae bacteremia. J Microbiol Immunol Infect 2009;42:303-9.
29. Yu VL, Chiou CC, Feldman C, Ortqvist A, Rello J, Morris AJ, et al. An international prospective study of pneumococcal bacteremia: correlation with in vitro resistance, antibiotics administered, and clinical outcome. Clin Infect Dis 2003;37:230-7.
30. Lee CC, Chang CM, Hong MY, Hsu HC, Ko WC. Different impact of the appropriate- ness of empirical antibiotics for bacteremia among younger adults and the elderly in the ED. Am J Emerg Med 2013;31:282-90.
31. Metersky ML, Ma A, Bratzler DW, Houck PM. Predicting bacteremia in patients with community-acquired pneumonia. Am J Respir Crit Care Med 2004;169:342-7.
32. Lee CC, Lee NY, Chuang MC, Chen PL, Chang CM, Ko WC. The impact of overcrowding on the bacterial contamination of blood cultures in the ED. Am J Emerg Med 2012; 30:839-45.