Clinical Characteristics, Microbiology and Outcomes of Bloodstream Infections in Patients Undergoing Haemodialysis—An Australian Network Cohort Study

ABSTRACT

Aim

This study reviewed the morbidity and mortality of bloodstream infections (BSIs) in patients with kidney failure receiving haemodialysis (HD).

Methods

We conducted a retrospective cohort study reviewing clinical characteristics, microbiology and outcomes of BSIs in patients receiving HD from 1 December 2017 to 1 December 2023 using data from an integrated network of dialysis sites. Descriptive and regression analyses were conducted, as well as Kaplan–Meier analysis for survival.

Results

Of 1461 patients who received HD over the study period, 171 positive blood cultures occurred in 137 patients. Sixty blood cultures (35%) were considered non‐significant contaminants, leaving 111 significant BSIs in 97 patients. Patients with BSIs had a median age of 65 years (interquartile range [IQR]: 50–76); 61% were males, with 48% having diabetes mellitus and 57% cardiovascular disease. A high proportion of BSI episodes (66/111, 59%) were determined to be HD access‐associated, primarily in patients with central venous catheters (55/111, 50%). The most frequent non‐HD access‐associated sources were intra‐abdominal/biliary (9%) and urinary (8%). Staphylococcus aureus (25%) was the most isolated organism, followed by Staphylococcus epidermis (12%). Polymicrobial BSI occurred in 10% of episodes. Median hospitalisation length was 9 days (IQR: 6–15). Mortality at 90 days was high (24%), and on multivariate analysis was associated with increasing age (odds ratio: 1.06 per year [IQR: 1.01–1.10]).

Conclusion

BSIs remain a major issue in patients receiving HD, contributing to prolonged hospitalisations and high mortality rates. HD access‐associated BSIs account for a substantial proportion of BSIs and remain a pertinent target for intervention.

1. Introduction

People with kidney failure receiving haemodialysis (HD) as kidney replacement therapy are at increased risk for serious infections. Mortality rates from infection in people on HD have been described as up to 82 times that of the general population [1]; however, appear to be decreasing over time [2]. Bloodstream infections (BSIs) in particular are much more common in people receiving HD than in non‐dialysing individuals. Those undergoing HD have 100 times the risk of BSIs with Staphylococcus aureus [3], an organism associated with mortality rates of 10%–30% [4]. In addition, BSIs with multidrug‐resistant bacteria, including methicillin‐resistant S. aureus (MRSA) and extended‐spectrum beta‐lactamase (ESBL)‐producing organisms, are more common in those receiving HD [5].

Increased susceptibility to severe infections in people with kidney failure receiving HD is likely multifactorial. Dysregulated immunity, underlying comorbidities and disruption of the skin barrier from indwelling prosthetic material or needling may all contribute. Mode of HD access is a well‐documented risk factor for BSI, with access through a central venous catheter (CVC) increasing BSI rates by 10–20 times compared to arteriovenous fistula (AVF) or arteriovenous graft (AVG) access [6, 7]. CVC‐associated BSIs have been well studied with significant interest in preventative strategies [6, 8]. However, whilst~60% of incident patients in Australia start HD with a CVC, only ~20% of prevalent HD patients continue with this modality, with the remainder using an AVF or AVG [9]. Understanding the microbiology and outcomes of BSIs in patients receiving HD through all access modalities is therefore important to identify those at increased risk of complex or recurrent infections and to optimise empiric antimicrobial therapies. We performed a retrospective study to comprehensively review demographics, microbiology and clinical outcomes of patients with a BSI episode and evaluate risk factors for BSIs in our local HD population.

2. Methods

This single‐centre, retrospective cohort study was conducted using data from the largest integrated network of dialysis sites in Australia. The network cares for patients with kidney failure throughout the state of Victoria, providing HD to around 500 patients in remote, rural and metropolitan settings through a combination of inpatient HD, 22 satellite dialysis centres and home HD [10]. The primary coordinating centre is a 1084‐bed public tertiary hospital which serves a predominately metropolitan population of more than 550 000 people [11]. A bespoke, nephrology database (Nephworks) records demographics and treatment details of all patients registered to the nephrology network, including a complete list of those receiving HD. This includes regional and rural patients who may have attended the metropolitan centre for hospital admission, vascular access or kidney transplantation outpatient review, as well as local metropolitan dialysis patients.

To identify episodes of BSI in patients receiving HD, patient identification numbers of adults ≥ 18 years who underwent HD from 1 December 2017 to 1 December 2023 were extracted from the Nephworks database (Figure 1). Basic demographics and HD modality information were also extracted. Identification numbers of those receiving HD were cross‐referenced against all positive blood cultures from the tertiary hospital microbiology laboratory during the same time period. Blood cultures are performed using the BD Bactec system (Becton Dickinson, United States). Positive blood cultures are subcultured on non‐selective (horse blood agar, chocolate agar) and selective media (MacConkey agar) and growth identified using the BD Bruker MALDI biotyper. Duplicate blood cultures of the same organism within a 3‐month period were excluded to account for episodes of persistent or relapsed infection. A retrospective chart review of each positive culture episode was conducted. Those not receiving HD at the time of BSI and those receiving ‘acute HD’, defined as HD required for acute kidney injury and not continued beyond the duration of a single hospital admission, were excluded. BSIs with organisms determined to be a contaminant (i.e., not of clinical relevance) by the treating team were not further analysed. For each remaining episode of BSI, demographics and clinical parameters were extracted from hospital electronic medical records.

FIGURE 1.

Study methodology. Abbreviations: BSI, bloodstream infection; HD, haemodialysis.

The Nephworks network includes a comprehensive database of metropolitan, regional and rural patients on established HD across Victoria, whereas blood cultures obtained from the microbiology laboratory would be taken predominantly from metropolitan patients. Given that non‐metropolitan HD patients within the network may present to regional or rural hospital sites for their initial blood cultures and may be ‘missed’, incidence calculations included only BSIs and HD years in metropolitan patients registered to in‐centre and seven associated satellite HD sites.

Source of infection was determined by recorded diagnosis in the patient's clinical notes as defined by the treating team. HD access‐associated infection was defined as a BSI where the source of infection was clinically determined to be related to a tunnelled or non‐tunnelled CVC, AVF or AVG. Metastatic infection was defined as a secondary site of infection subsequent to BSI, for example, septic emboli, joint infection, or endocarditis. Source control was any invasive intervention undertaken to reduce the burden of infection, such as removal of infected indwelling prosthetic material (including CVCs or grafts) or drainage. Immunosuppression included those living with human immunodeficiency virus, receiving steroid dosing equivalent to more than 20 mg prednisolone daily for ≥ 4 weeks, or recent chemotherapy or immunomodulation (including for transplantation, malignancy or autoimmunity) at the time of BSI.

Collated data was analysed using descriptive statistics. Categorical variables were analysed with chi‐square (χ ²) and Mann–Whitney U test (two‐tailed) for continuous variables. Univariable and multivariable logistic regression analysis was conducted to evaluate associations with 90‐day mortality in R (version 4.2.2, R Core Team 2024). Multivariable models incorporated variables associated with the dependent variable on bi‐variable analysis (p < 0.05). Model fit was evaluated using the Hosmer–Lemeshow goodness‐of‐fit test, with p value > 0.05 indicating adequate fit. Kaplan–Meier curves were used to visualise time to BSI in metropolitan patients who commenced HD after the study start date (Python 3.12, lifelines package, 2024 [12]), with patients censored at the time of death, withdrawal of HD or the study endpoint. Log rank testing was used to determine differences in time to mortality from the onset of first BSI compared to patients who commenced HD after the study start date but did not develop a BSI.

Local ethics approval was obtained through the Melbourne Health Human Research and Ethics Committee (HREC) Quality Assurance pathway (QA2024044). As this was a retrospective observational study, individual patient consent was not required.

3. Results

In the nephrology network, 1461 patients received HD during the study period, totalling 3623 HD years. When matched to hospital laboratory reporting systems, 171 positive blood cultures occurred in 137 patients concurrently receiving HD. Sixty (35%) positive blood cultures were considered non‐significant contaminants after review of clinical notes, leaving 111 BSI events in 97 patients for analysis. Metropolitan patients accounted for 447 patients (31%) in the network, with a cumulative total of 754.8 patient years on HD. Sixty‐seven episodes of BSI (60%) occurred in 59 metropolitan patients (13%), with an estimated incidence rate of 8.9 BSI per 100 patient years on HD (95% confidence interval [CI]: 7.4–11.2).

3.1. Patient Demographics

In the 97 patients who experienced BSI, median age was 65 years (interquartile range [IQR]: 50–76) with the most frequently recorded ethnicity being Australian or European (41/97 and 17/97, respectively, 60%) (Table 1). Co‐morbidities were common. Diabetes mellitus affected 48% of patients, and 57% had documented cardiovascular disease. Basic demographics of HD patients who did not have a known BSI were similar to those who did; however, there was a higher proportion of metropolitan patients in the BSI group (61% vs. 31%, p < 0.01).

TABLE 1.

Patient Demographics.

	HD patients with BSI (n = 97)	HD patients without BSI (n = 1364)
Male sex, n (%)	59 (61)	907 (66)
Median age, years [IQR]	65.1 [50–76]	67.8 [54–77]
Metropolitan area patient, n (%)	59 (61)	388 (28)
Ethnicity, n (%)
African	3 (3)	31 (2)
Asian	9 (9)	102 (7)
Aboriginal Australian	0 (0)	40 (3)
Australian	41 (42)	698 (51)
European	17 (18)	170 (12)
Middle Eastern	8 (8)	43 (3)
Māori/Pasifika	1 (1)	40 (3)
North/South American	0 (0)	10 (1)
Unknown	18 (19)	230 (17)
Comorbidities, n (%)
Kidney transplant prior to BSI	21 (22)	N/A
Immunosuppression	29 (30)	N/A
Diabetes mellitus	43 (48)	N/A
Cardiovascular disease	55 (57)	N/A
Median body mass index, kg/m2 (n = 91) [IQR]	28 [23–31]	N/A
Haemodialysis access
Primary access at time of BSI
Arteriovenous fistula/graft	43 (44)	—
Central venous catheter	54 (56)	—
> 1 access	32 (33)	—
First HD access, n (%)
Arteriovenous fistula/graft	48 (49)	720 (57)
Central venous catheter	49 (49)	553 (43)
Median age at first HD, years [IQR]	59 [45–75]	62 [48–72]
Median time on HD prior to first BSI, years [IQR]	1.7 [0.4–3.6]	—

Abbreviations: BSI, bloodstream infection; HD, haemodialysis; IQR, interquartile range; N/A, not available.

On review of all BSI episodes (n = 111), primary access was an AVF or AVG in 44%, and CVC in the remainder (56%). One third of patients had more than one HD access (36/111, 32%) at the time of BSI, with the majority (32/36) having an AVF as a second (inactive) access. A total of 65 patients (59%) had a CVC in situ at the time of infection. Total median time on HD prior to BSI was 1.7 years (IQR: 0.4–3.6) for all patients. Median time on HD for metropolitan HD patients who commenced HD after the start date of the study was 1.6 years (IQR: 0.7–2.8), and 12% of BSI episodes in this group occurred in the first 3 months of commencing HD (Figure 2a).

FIGURE 2.

Kaplan–Meier analysis for time to bloodstream infection (BSI) and survival. (a) Kaplan–Meier plot of time to BSI in local metropolitan patients who commenced haemodialysis (HD) after the start date of the study (1 December 2017). Event is date of BSI (patients able to have multiple BSIs [n = 2]). Patients censored at death, withdrawal of HD or study end date (1 December 2023). (b) Kaplan–Meier plot comparing time to death during the study period in all HD patients from onset of BSI (orange) or start date of the study/commencement of HD if no BSI (blue). Non‐BSI group includes those who commenced haemodialysis (HD) from 1 December 2017 to 1 December 2023.

3.2. Clinical Outcomes

Median time in hospital during an episode of BSI was 9 days (IQR: 6–15) (Table 2). Twenty patients (18%) developed a metastatic infection, including infective endocarditis in 11 (10%). By 90 days, 27 patients (24%) had died. On univariable logistic regression, 90‐day mortality was associated with increased age (odds ratio [OR] 1.06 for each additional year, 95% CI 1.02–1.10, p = 0.002), cardiovascular disease (OR: 3.6, 95% CI 1.28–11.8, p = 0.02) and longer time on HD (OR: 1.19 per HD year, 95% CI 1.03–1.38, p = 0.02) (Table 3). Undergoing a source control procedure was protective (OR: 0.32, 95% CI 0.12–0.84, p = 0.02). On multivariable analysis, only increased age was associated with 90‐day mortality (OR: 1.05, 95% CI 1.01–1.10, p = 0.02).

TABLE 2.

Outcome of bloodstream infection in haemodialysis patients.

Median length of stay, days [IQR]	9 (6–15)
Metastatic complications of BSI, n (%)	20 (18)
Intensive care unit (ICU) admission, n (%)	21 (19)
Median ICU length of stay, days [IQR]	2 (1.8–4.5)
30‐day mortality, n (%)	18 (16)
90‐day mortality, n (%)	27 (24)

Access and interventions	n (%)
Source control procedures	61 (55)
Central venous catheter removal	48 (43)
Other methods a	13 (12)
Median duration antimicrobial, days [IQR]	16.5 [14–28]
Staphylococcus aureus	28 [15–40]
Coagulase‐negative Staphylococcus b	14 [10–28]
Gram‐negative organisms b	14 [10–19]
Polymicrobial infection	28 [15–28]
Lifelong antibiotic suppression	4 (4)

^a Other source control procedures included: endoscopic retrograde cholangiopancreatography (n = 4), drainage of collection (n = 3, prostatic, renal and peri‐arteriovenous graft [AVG]), AVG excision (n = 2), arteriovenous fistula pseudoaneurysm excision (n = 1), aortic valve replacement (n = 1), cholecystectomy (n = 1), amputation (n = 1).

^b Single culture only.

TABLE 3.

Regression analysis of 90‐day mortality with bloodstream infection (BSI).

	Univariate analysis		Multivariate analysis
	OR (95% CI)	p	OR (95% CI)	p
Age (per year)	1.06 (1.02–1.10)	0.002*	1.05 (1.01–1.10)	0.02*
Male sex	1.28 (0.42–4.04)	0.65	—
Haemodialysis duration (per year)	1.19 (1.03–1.38)	0.02*	1.10 (0.93–1.30)	0.27
Body mass index (per kg/m2)	0.92 (0.83–1.0)	0.06	—
Diabetes mellitus	0.95 (0.37–2.45)	0.93	—
Cardiovascular disease	3.6 (1.28–11.8)	0.02*	2.51 (0.72–10.32)	0.17
Previous kidney transplant	1.01 (0.30–2.99)	0.99	—
Central venous catheter access	0.66 (0.25–1.70)	0.39	—
Haemodialysis‐associated BSI	0.56 (0.21–1.44)	0.23	—
S. aureus BSI	1.27 (0.43–3.47)	0.65	—
Source control procedure	0.32 (0.12–0.84)	0.02*	0.46 (0.13–1.50)	0.20

^* = p < 0.05.

On review of patients who commenced HD after the study start date (n = 860), the proportion of patients who had died was higher amongst those who had experienced a BSI (19/48, 40%) than those who did not (210/812, 26%, p = 0.04). Time from first BSI to mortality occurred steeply in the first few months, with mortality then following a similar rate in the no BSI group (Figure 2b, p = 0.5).

3.3. Microbiology of BSIs

Twelve patients (12%) had recurrent BSI episodes during the study period; one patient had four episodes, and the remainder had two. Mean duration between BSI episodes was 0.95 years (IQR: 0.6–1.5). Eleven BSI episodes (10%) were polymicrobial, and 37 separate organisms were identified. The most common was S. aureus (31/124, 25%) followed by Staphylococcus epidermidis (15/124, 12%) (Table 4). Multi‐drug resistance was uncommon, with only two MRSA and two vancomycin‐resistant Enterococci (VRE) cultured. The most frequently clinician‐identified source of BSI was CVC‐associated (55/111, 50%) or unknown (13%) followed by biliary/intra‐abdominal (9%).

TABLE 4.

Microbiological and treatment aspects of bloodstream infections in patients with haemodialysis.

Source of bloodstream infection	n (%)
Haemodialysis access‐associated BSI, n (%)	66 (59)
Catheter‐associated	55 (50)
Unknown	14 (13)
Skin and soft tissue infection
Fistula/graft‐associated	9 (8)
Diabetic foot ulcer	3 (3)
Other	2 (2)
Biliary/intra‐abdominal	10 (9)
Urinary	9 (8)
Respiratory	5 (5)
Dental	2 (2)
Infective endocarditis	2 (2)

Organism cultured (n = 124)	n (%)
Contaminant organism a	82
Polymicrobial infection	11 (10)
Staphylococcus aureus	31 (25)
Enterobacterales (not including Escherichia coli )	23 (19)
Coagulase‐negative Staphylococcus	21 (17)
Escherichia coli	12 (10)
Streptococcus species	9 (6)
Enterococcus species	7 (6)
Stenotrophomonas maltophilia	3 (2)
Candida	1 (1)
Other anaerobic	6 (5)
Other gram‐negative	9 (7)
Other gram‐positive	2 (2)

Note: Enterobacterales: Citrobacter (1), Enterobacter (1), Klebsiella (11), Morganella (1), Pantoea (3), Proteus (2), Serratia (4). Coagulase‐negative Staphylococcus: S. capitus (2), S. epidermidis (15), S. haemolyticus (1), S. lugdunesis (2), S. warneri (1). Streptococcus species: Group B Strep (2), S. milleri (2), S. pneumonia (2), Virdans strep (3). Enterococcus species:  E. faecalis (4), E. faecium (1). Anaerobes: Clostridium (1), Fusobacterium (2), Bacillus cereus (1), Parabacteroides (1). Other gram‐negative: Acinetobacter (1), Achromobacter (2), Campylobacter (1), Chryseobacterium (1), Cupriavidus (1), Moraxella (1), Sphingomonas (2). Other gram‐positive: Corynybacterium (1), Listeria (1).

Abbreviations: BSI, bloodstream infection; IQR, interquartile range.

^a Polymicrobial infection was defined as ≥ 2 organisms cultured.

A high proportion of BSIs in HD patients were HD access‐associated (66/111, 59%). The majority of HD access‐associated BSIs were CVC‐associated (55/66, 83%). S. aureus was more common in HD access‐associated BSIs (41% vs. 9%, p < 0.01) occurring in 20/55 (36%) CVC‐associated BSIs and 6/9 (67%) with AVF/AVG infection. CVC‐associated BSIs occurred after a shorter mean duration on HD (15 months, IQR: 3–30) than AVF/AVG‐associated (57 months, IQR: 21–72) and non‐HD associated BSIs (37 months, IQR: 10–63). Metastatic infection occurred in 12/55 (22%) patients with CVC‐associated BSI. Fifty‐one patients with HD access‐associated BSIs underwent targeted source control procedures (46%). This included 47 CVC removals, AVG excision (n = 2), AVG collection drainage (n = 1) and excision of an AVF pseudoaneurysm (n = 1). Of the six patients with CVC‐associated BSIs who did not undergo line removal, three were deceased at 90 days. Reasons for not undergoing line removal included limited life expectancy (n = 2), no alternative access sites (n = 2) or patient preference (n = 2).

4. Discussion

In this retrospective cohort study, patients receiving HD who developed BSIs experienced substantial rates of morbidity and mortality, with almost one in four patients dying within 90 days of a clinically significant positive blood culture. There was a high rate of complications from BSIs such as metastatic spread of infection and endocarditis, with patients requiring prolonged hospitalisation and invasive source control procedures. It is particularly notable with these high morbidity and mortality rates, two‐thirds of BSI episodes were HD access‐associated and therefore potentially preventable.

Estimated BSI incidence was 8.9 BSI per 100 patient years on HD for metropolitan patients, with 13% developing a clinically significant BSI during the study period. We calculated incidence only amongst those registered to metropolitan HD centres to account for potentially ‘missed’ regional and rural patients, as described. Our data reflect these potentially unaccounted for BSIs, with non‐metropolitan patients representing 69% of the network HD cohort, but only 40% of total BSIs. This method may still have underestimated incidence if metropolitan HD patients presented to a different urban hospital or private laboratory for initial blood cultures. Despite these limitations, our findings are comparable to an Australian surveillance network study in the state of Victoria from 2008 to 2015 reporting 394 BSIs over 79 803 patient months (5.9 BSI per 100 patient years) [6] and a population‐based Danish cohort study which reported 13.7 episodes of first BSI per 100 person‐years [13]. We did not note any clear trend in absolute numbers of BSIs over the 6‐year study period, and infection prevention measures, including the use of dialysis catheter locks and dressings, have remained consistent in our centre.

Regarding risk factors for BSIs, patients in our cohort were multimorbid. High prevalence of diabetes mellitus and immunosuppression was apparent and may be indicative of the aetiology precipitating kidney failure. However, these co‐morbidities also predispose patients to serious infections. It is notable that a proportion of patients were commenced on lifelong antibiotics or were unable to undergo source control procedures, suggesting underlying inadequate physiological reserve for invasive procedures. On univariable analysis, increasing age, cardiovascular disease and longer duration on HD were associated with increased 90‐day mortality following BSI. Increasing age was the only associated variable on multivariable analysis.

Around half of BSI episodes were CVC‐associated, a potentially modifiable risk factor. These BSIs occurred earlier after commencement of HD. AVF formation is a preferred access modality to reduce HD access‐associated infection [14] although patients may be dialysed through a CVC for multifactorial reasons such as short‐term bridging therapy in acute HD, previous AVF complications, or cosmetic appearance [15]. Onethird of patients with BSIs had more than one vascular access for HD in situ, this may reflect early HD awaiting fistula maturation or complexities in vascular access. In patients who dialyse with a CVC, optimising line care to prevent infection is paramount. Although multifaceted quality improvement programs have recently not demonstrated efficacy in the Australian setting [8], further innovation is clearly required.

From a microbiological perspective, the overall predominance of gram‐positive organisms, particularly S. aureus , is consistent with similar studies [6]. We noted a lower‐than‐expected prevalence of multidrug‐resistant organisms, particularly MRSA, which accounted for only two S. aureus isolates (6%), much lower than the 2022 reported incidence of 15% for all Australian S. aureus BSIs [16]. Despite the low prevalence of MRSA, vancomycin is often administered empirically in HD patients at our institution given that the mortality of BSIs is high and nephrotoxicity is not a significant problem in this population. We also noted high rates of contaminant organisms, with 35% of positive cultures considered not clinically significant by the treating team. Contaminant cultures can result in unnecessary hospitalisation, antimicrobial use and repeat testing with substantial additional costs to both the patient and healthcare system. This remains an area to target for optimisation.

Strengths of this study include a comprehensive clinical review of over 100 BSIs in a cohort of HD patients, describing patient demographics, microbiological features and outcomes. To our knowledge, this is the first descriptive clinical review of all BSI infections in HD patients, not limited to catheter or other HD access‐associated BSIs, in an Australian setting. This study, however, does have limitations. Given the retrospective nature of data collection, variables such as co‐morbidities and infection source relied on documentation from the treating team and was not standardised. The study is from a single network, and the patient cohort and microbiology may not be generalisable to other centres. Our incidence calculations likely estimate the lower bounds of incidence given the reasons described above. We were unable to calculate catheter days as a denominator for incidence as this data is not routinely collected, although this would have been highly relevant for incidence rate reporting.

5. Conclusion

This retrospective cohort study has demonstrated HD access‐associated BSIs remain a major issue for patients on HD, accounting for almost two‐thirds of BSIs. BSIs were associated with prolonged hospitalisations, invasive interventions, and a mortality rate of 24% within 90 days of the positive blood culture. Considering many of these infections are potentially preventable, further research is required to examine novel prevention and infection control strategies in this vulnerable dialysis patient population.

Author Contributions

Aliya N. Bryce: conceptualisation, methodology, investigation, formal analysis, writing – original draft. Brett Sobey: software, formal analysis. Paul Kinsella: investigation, writing – review and editing. Steven Y. C. Tong: writing – review and editing, supervision. Nigel D. Toussaint: conceptualisation, methodology, formal analysis, writing – review and editing, supervision.

Conflicts of Interest

The authors declare no conflicts of interest.

Bryce A. N., Sobey B., Kinsella P., Tong S. Y. C., and Toussaint N. D., “Clinical Characteristics, Microbiology and Outcomes of Bloodstream Infections in Patients Undergoing Haemodialysis—An Australian Network Cohort Study,” Nephrology 30, no. 9 (2025): e70113, 10.1111/nep.70113.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.