Seven-Year Study of Bacteremia in Hospitalized Patients on Chronic Hemodialysis in a Single Tertiary Hospital

Abstract

Sixty-seven hospitalized bacteremic patients on chronic hemodialysis were matched with 134 bacteremic controls. Mortality (25% and 28% respectively) was unrelated to dialysis. Predictors of mortality were: advanced age, decreased functional capacity, and malignancies. Most common organisms recovered were Staphylococcus aureus (30%), coagulase negative Staphylococcus (15%), and polymicrobial agents (15%).

• Bacteremia
• Chronic hemodialysis

Introduction

Infection is known to be a major cause of morbidity and mortality in patients on chronic hemodialysis (CHD).[[4]], [[11]], [[14]] Previous studies suggested that hemodialysis vascular access is the major risk factor for bacteremia in CHD population. The majority of these bacteremias are caused by staphylococci and are associated with high rate of mortality.

The aims of the present study were to describe the presentation, pathogens, course, and fatal outcome of bacteremia in hospitalized patients on chronic hemodialysis (HPCH); and to define risk factors for death in these patients.

Methods

We used prospective data from the Beilinson's bacteremia computerized database,[[1]], [[15]] from March 1, 1988 to August 1, 1995, to capture all cases of bacteremia and fungemia in HPCH defined as dialysis that was started prior to the present hospital stay (not including outpatients receiving dialysis at the hospital). Each case was matched with two bacteremic hospitalized patients not on hemodialysis treatment.

Rabin Medical Center, Beilinson Campus is a 900-bed university hospital in central Israel that serves an urban population of about 300,000 as a first line facility. It has a large dialysis unit and any active unit for transplantation of solid organs. Episodes of bacteremia were detected by daily surveillance of laboratory records. An episode was included in the database if blood culture was positive and not contaminated in the presence of clinical or laboratory evidence of infection.[[1]], [[15]] Aerobic and anaerobic blood cultures were obtained in BACTEC bottles (6B aerobic and 7D anaerobic tryptic soy broth; Johnston, USA). The bottles were tested on the radiometric BACTEC-460 system (Johnston). The median number of sets obtained from a patient was two, and the range was 1–4. Susceptibility to antibiotics was tested by the standard disk diffusion technique. Organisms that are commonly recovered from the environment or the skin (namely coagulase negative staphylococci and aerobic gram-positive rods) were judged as contaminants, unless the clinical findings, the results of cultures of material from other body sites, and the number of positive sets (two or more) indicated a high probability of blood-stream infection. During hospitalization and afterwards, demographic data, information on underlying disorders, presentation of the acute disease, the infecting microorganism, treatment, source of disease, and outcome of hospital stay were gathered and recorded. Episodes of bacteremia or fungemia were considered hospital-acquired if it occurred >48 h after admission to the hospital. Septic shock was defined as systolic blood pressure less than 90 mmHg, plus at least one of the following: clouded perception, hypoxemia (pO₂<70 mmHg), or findings compatible with disseminated intravascular coagulopathy. The definition did not include the duration of hypotension or response to fluid resuscitation.

Functional capacity was divided into four categories: 3-full activity; 2-curtailed activity, does not leave home; 1-needs assistance for most daily activities; and 0-bedridden. An episode of bacteremia occurring 48 h after admission was defined as nosocomial. Neutropenia was defined as a granulocyte count of <500 per/mm³. Two control patients were matched for every HD patient in the study group. They were matched from the computerized database according to the following criteria: age (±5 years), sex, department of hospitalization, date of hospitalization (±6 months), and the presence or absence of solid malignancies, diabetes mellitus or congestive heart failure, source of infection, infective microorganism, functional capacity group, hospital or community acquired infection, and neutropenia. If more than two matches were found, each control was chosen at random. Both study cases and control had blood cultures taken during the episode of bacteremia. All subjects received empirical antibiotic treatment that was considered appropriate if the infecting microorganism was subsequently found to be susceptible in vitro to the drug administered, and if that antibiotic was given intravenously.

Statistical Analysis

For multivariate analysis, variables were added on the basis of the largest increase in the test statistic for the whole mode until the addition was nonsignificant. The risk factors for mortality are related to the hemodialysis treatment, infectious disease, the underlying disorder and the interaction between them. Because most of the continuous variables were not normally distributed, their values were presented as median and range and the Wilcoxon rank sum (nonparametric) test was used for comparisons between two classes. The following variables were tested for association of hemodialysis with mortality in the study group: age, sex, place of acquisition, functional capacity, evidence of malignancy, atherosclerotic heart disease, congestive heart failure, chronic lung disease, serum albumin, creatinine and blood urea nitrogen levels, septic shock, DIC, ARDS, liver failure and coma; source of infection, infecting microorganism and appropriate empiric antibiotic treatment. For analysis of continuous variables values were divided into quartiles.

Results

Sixty-seven episodes of bacteremia and fungemia were identified in hospitalized patients receiving long-term hemodialysis. The comparison of the cases and matched paired controls for demographic data and underlying disorders is shown in Table 1. Thirty percent of the episodes in both groups were hospital-acquired. There was a significantly higher rate of nursing home residents in the control group (20% vs. 3%, p = 0.08).

Table 1. Univariate comparison of demographic data and underlying disorders in the study and control groups

	Hemodialysis study group	Nonhemodialysis control group	p value
Number of episodes	67	134
Number of patients	67	134
Aged (yrs)
Median	63	64	0.5
Range	8–89	6–95
Male (%)	62	62
Hospital acquired infections (%)	30	30
Nursing home residents (%)	3	20	0.08
Functional capacity (%)
0	6	6
1	2	2
2	43	43
3	49	49
Underlying disorders (%)
Ischemic heart disease	16	16	0.5
Congestive heart failure	9	9
Diabetes mellitus	25	16
Cerebrovascular accident	4	5	0.8
Malignancy	6	5
Dementia	0	4	0.1
Cirrhosis	0	1	0.5
Decubitus ulcer	4	4
System lupus erythematosus	4	0	0.03
Multi-trauma	3	2	0.8
S/P splenectomy	4	3	0.6

The distribution of the bacteremic infections in the study group was as follows: vascular access site—24%; respiratory tract—9%; urinary tract—7%; neutropenic fever—5%; endovasculitis—3%; others—9%; and unknown—33%. There was no difference in median body temperature, leukocyte count, or potassium levels on the day of blood culture collection between the groups (Table 2). As expected, blood urea nitrogen and creatinine levels were significantly higher in the hemodialysis patients, and plasma albumin levels were lower. The majority of infections involved gram-positive aerobic organisms (n = 39, 58.2%), followed by gram-negative aerobes (n = 16, 23.9%) and yeast (n = 2, 3%). In 10 cases (14.9%), two organisms were identified. The microbiological data are given in Table 3. The most common pathogen isolated in the study group, accounting for 30% of the patients with bacteremic episodes, was Staphylococcus aureus, 29% of the patients with Staphylococcus aureus bacteremia had methicillin-resistant Staphylococcus aureus (MRSA).

Table 2. Cinical and laboratory data in the study and control groups on the day of blood culture

	Hemodialysis study group median (range)		Nonhemodialysis control group median (range)		p value
Temperature (°C)	38.5	(36–40.6)	38.5	(36–41)	0.2
Leukocytes (10^9/L)	11.9	(0.6–36.8)	11.2	(0.3–288)	0.7
Blood urea nitrogen (mg/dL)	70	(28–218)	23	(1–182)	0.0001
Creatinine (mg/dL)	8.1	(2.5–14.8)	1.3	(0.4–8.3)	0.0001
Potassium (mEq/L)	4.4	(2.5–14.8)	4.1	(3–5.9)	0.1
Albumin (mg/dL)	3.1	(2.3–3.9)	3.3	(2.1–4.9)	0.004
Hospital stay (days)	6	(0–48)	7	(1–70)	0.1

Table 3. Microbiological findings in 67 cases of bacteremia in hospitalized patients on chronic hemodialysis

Causative organism	No. of cases	(%)
Staphylococcus aureus	20	(29.9)
Coagulase-negative Staphylococcus sp.	10	(14.9)
Streptococcus sp.	9	(13.4)
Pseudomonas aeruginosa	8	(11.9)
Eschericia coli	5	(7.5)
Klebsiella	3	(4.5)
Candida albicans	2	(3.0)
Polymicrobial	10	(14.9)

Antibiotic therapy was started on day 1 of bacteremia in 90% of the hemodialysis patients vs. 73% in the control patients (p = 0.05). Forty percent of the patients on hemodialysis received appropriate empiric antibiotic treatment compared to only 27% of the controls (p = 0.04). The frequency of septic shock, adult respiratory failure, diffuse intravascular coagulation, liver failure, and coma was similar in the two groups.

Twenty-five percent of the study patients and 28% of the control patients died during hospitalization (p = 0.7). Univariate analysis revealed that the variables associated with fatal outcome in the hemodialysis group were advanced age (p<0.004) and solid malignancies (p<0.0001). Elevated blood urea nitrogen levels and low serum albumin levels were associated with increased mortality risk in the control patients, but not in the hemodialysis patients (Table 4).

Table 4. Univariate predictors of a fatal outcome in the study and control groups

	Hemodialysis study group (n = 67)					Nonhemodialysis control group (n = 134)
	Mean (range)					Mean (range)
	Survived (n = 50)		Died (n = 17)		p value	Survived (n = 96)		Died (n = 30)		p value
Age (years)	61.5	(20–82)	70	(8–89)	0.0004	62	(6–95)	66	(6–82)	0.4
BUNa (mg/dL)	69	(30–179)	82	(28–218)	0.7	18	(1–109)	37	(4–182)	0.0001
Creatine (mg/dL)	8.1	(2.5–12.2)	5.5	(3.4–14.8)	0.5	1.1	(0.4–6.6)	1.8	(0.5–8.3)	0.3
Albumin (mg/dL)	3	(2.3–3.9)	3.2	(2.7–3.9)	0.6	3.5	(2.1–4.6)	3	(2.1–4.9)	0.001

^aBUN = blood urea nitrogen.

Table 5 shows the variables found to be significant independent predictors of mortality on logistic regression analysis. There included advanced age, decreased functional capacity, and malignancies in the study group, and unknown source of infection and low levels of serum albumin in the control group.

Table 5. Significant and independent predictors of mortality on stepwise logistic regression analysis in the study and control groups

	Multivariate adjusted odds ratios (95%CI)
	Hemodialysis study group mean (range)		Nonhemodialysis control group mean (range)
Male sex	2.3	(0.6–8.0)	1.0	(0.5–2.2)
Concomitant malignancy	17.3	(1.7–174.3)	2.3	(0.9–6.0)
Inappropriate antibiotic treatment	1.3	(0.4–4.5)	1.3	(0.5–3.3)
Age (≥median)	4.9	(1.4–17.1)	1.5	(0.7–3.2)
Low functional capacity	11.4	(1.1–125)	2.2	(0.7–6.8)
Unknown source of bacteremia	3.2	(1.0–10)	3.2	(1.5–7.2)
Low serum albumin levels	1.0	(0.5–2.3)	6.5	(2.0–20.8)
Neutropenia	1.6	(0.3–7.1)	2.8	(0.3–23.8)
Pseudomonas infection	2.8	(0.6–11.8)	2.5	(0.8–7.9)

Discussion

The present comparison of case patients and matched controls shows that hemodialysis is not a risk factor for mortality after an episode of bacteremia or fungemia. Mortality during hospital stay was similar in the two groups: 25% in the hemodialysis patients and 28% in the patients not receiving hemodialysis (p = 0.7). The risk of a fatal outcome increase with advanced age only in the study group. The median age of the hemodialysis patients who survived the bacteremic episodes was 61.5 years compared to 70 years in those who did not (p = 0.0004). However, age was not a risk factor in the control group (Table 3). Today, because age per se is no longer a contraindication for dialysis, we expect that in a number of years, the majority of dialysis patients will be 65 years old or more, with adequate survival and good subjective quality of life. However, clinicians should be aware that the high-risk conditions associated with age may place these patients at higher risk of mortality in the presence of infection.

Malignant disease was significantly and independently associated with increased mortality in the study group but not the control group. One possible explanation, other than cancer-induced systemic debilitation, is that hemodialysis patients with diagnosed malignancies may not be treated aggressively for septic conditions because of anticipated poor outcome.

Bacteremia and fungemia of unknown origin accounted for one-third of all episodes of infection in the study patients. Infection may arise from the usual sources, though the source may not be identified,[[16]], [[17]] or a high inoculum of microorganisms may be injection accidentally into the bloodstream.[[5]] High-efficiency and high-flux hemodialysis is becoming increasingly popular in patients with chronic renal failure because they reduce the time of treatment. Both methods require bicarbonate dialysate, often prepared from concentrates that support bacterial growth with endotoxin production. Gordon et al.,[[9]] in a study conducted at three dialysis centers, found that bacterial counts significantly exceeded the microbiologic standards for dialysate. Other studies suggested that outbreaks of gram-negative bacteremia might have been due to contaminated O-rings and removable headers whereby bacteria gained access to the bloodstream.[[3]], [[7]], [[10]]

In the present study, 45% of the bacteremic episodes in the study group were caused by staphylococci. This is compatible with other studies showing that patients undergoing hemodialysis are at high risk of developing staphylococcal bacteremia and its complications.[[6]] Thus, empirical antibiotic treatment of suspected bacteremia of unknown origin or associated with intravascular access (24% of patients) must include an antistaphylococcal drug.[[19]] Twenty nine percent of Staphylococcus aureus isolated were methicillin-resistant Staphylococcus aureus (MRSA). This finding suggests that patients treated with chronic hemodialysis may benefit from the newly developed Staphylococcus aureus conjugated vaccine, which has been found to be protective against Staphylococcus aureus bacteremia for approximately 40 weeks in patients receiving hemodialysis.[[6a]] MRSA infection may be adequately treated with an initially with a single dose of 1 g intravenous vancomycin,[[2]] yet after dialysis, plasma concentrations of vancomycin return to pre-dialysis value.[[8]]

Polymicrobial bacteremia accounted for 14% of the episodes in the study group. Hospitalized patients have been found to be at higher risk of polymicrobial bacteremia due to invasion via the urinary tract, intraabdominal foci, and skin and soft tissues.[[20]] The mortality rate related to sepsis is reportedly twofold higher in polymicrobial than in unimicrobial bacteremia.[[22]] The site and type of polymicrobial bacteremia can suggest a source of sepsis as well as additional diagnostic and therapeutic measures.[[21]]

Candida albicans was isolated in 3% of episodes. All patients in both groups with candidemia died. Over the past decade, the incidence of hospital-acquired bloodstream infections caused by Candida sp. has risen.[[23]] The incidence is dramatically higher in high-risk, critical-care units. Candida infection of the fistula used for hemodialysis is exceedingly rare, with only four cases reported in the English literature.[[18]] Antifungal therapy should be given, but complete removal of the arteriovenous fistula may be necessary for long-term cure.

Low serum albumin level is a known marker of increased mortality in the general population. It is associated with even higher mortality in patients with sepsis and severe acute toxic state.[[12]], [[13]] In the present study, low serum albumin was a significant predictor of mortality in the control group but not in the hemodialysis group.

In summary, the present case-control study yielded three factors that were predictive of a fatal outcome after bacteremia or fungemia in patients on hemodialysis: old age, underlying malignancy, and functional capacity.

As fatality was unrelated to the hemodialysis treatment per se, we suggest that the incidence of bacterial and fungal infection in patients on hemodialysis may be reduced by a more appropriate supportive therapy especially in the high risk groups of patients receiving chronic hemodialysis treatment, elderly patients with underlying malignancies and low functional capacity.

Acknowledgment

The authors wish to thank Gloria Ginzach, Marian Propp, and Phyllis Curchack Kornspan for their editorial and secretarial assistance.