Intensive Care Unit–onset Bloodstream Infections Represent a Distinct Category of Hospital–onset Infections: A Multicentre, Retrospective Cohort Study. Queensland Critical Care Network (QCCRN)

Abstract

Background:

The location of onset of bloodstream infections (BSIs) associated with intensive care unit (ICU) admission may influence their clinical and epidemiological characteristics.

Methods:

A multicentre, retrospective cohort study was conducted in Queensland, Australia, and BSIs associated with ICU admission were identified and classified as community-onset, hospital-onset, or ICU-onset if first isolated within, after 48 hours but within 48 hours of ICU admission, or after 48 hours following ICU admission, respectively.

Results:

We included 3,540 episodes of ICU-associated BSI, with 1,693 classified as community-onset, 663 hospital-onset, and 1,184 ICU-onset. Compared with hospital-onset BSIs, patients with ICU-onset BSIs were younger, had fewer comorbidities, had lower APACHE II scores, and were more likely male. Patients with ICU-onset BSI were more likely to be surgical admissions and have a primary cardiovascular or neurological diagnosis. The distribution of infective agents varied significantly among community-, hospital-, and ICU-onset BSI groups. The all-cause 30-day case-fatality rates for first-episode community-onset, hospital-onset, and ICU-onset BSIs were 17.1%, 21.7%, and 23.5%, respectively (p < 0.001).

Conclusion:

With different epidemiological features and causal pathogens, ICU-onset BSI represents a distinct BSI group arising in hospitalized patients.

Introduction

Bloodstream infections (BSIs) associated with admission to intensive care units (ICUs) represent 16% of the overall burden of infections in the ICU and are consistently associated with substantial mortality (1). They are both important causes and complications of admission to ICUs (2, 3, 4, 5). Traditionally, hospital-treated infections have been classified as either community or hospital (nosocomial) acquired (6). Because determining the actual place of acquisition is fraught with difficulty, classification of infections based on the location of onset has gained widespread acceptance (7). Community-onset infections are those where the infection is first detected prior to or within 48 hours of hospital admission, and hospital-onset infections are those first cultured 48 hours or more after admission or within 48 hours of discharge (8,9).

Hospital-onset BSIs are associated with a major burden of illness among patients admitted to ICUs worldwide (2,10). They have different characteristics, are caused by various pathogens, and are associated with higher mortality than community-onset BSIs (11). These infections may develop among patients admitted to hospital wards and precipitate the need for ICU admission. Those subsequently occurring as complications within the critical care setting may be classified as ICU-onset. However, whether defining the subgroup of hospital-onset infections as being of ICU-onset or not has not been previously systematically studied. Therefore, this study's objective was to examine the occurrence, clinical determinants, and outcome of ICU-associated BSIs with a comparative analysis of community-onset, hospital-onset, and ICU-onset infections.

Methods

Design

Retrospective, multicentre cohort. Manuscript prepared following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement (12).

Study population

The study population was comprised of adult admissions to study ICUs in Queensland, Australia (13). Study ICUs included 12 publicly funded, closed-model ICUs that collectively represent a majority of state ICU capacity, including all state-wide referral centres for cardiothoracic, neurosurgical, obstetric, transplant, burns, and trauma patients. Of the included ICUs, five were classified as tertiary, three as metropolitan, and four as regional. All patients aged 18 years and older admitted between January 1, 2015, and December 31, 2021, to study ICUs for 48 hours and longer were included.

Study protocol

The human research ethics committee at the Royal Brisbane and Women's Hospital approved this study and granted a waiver of individual consent. Study subjects were identified and information was obtained by linkage of local, state-wide, and national electronic sources. Patient admission, demographic, physiology, and laboratory data were retrieved using the eCritical MetaVision (iMDsoft, Boston, MA, USA) clinical information system at each centre. Further registration, severity of illness, and management details were obtained from the Australia and Zealand Intensive Care Society (ANZICS) Centre for Outcome and Resource Evaluation (CORE) Adult Patient Database (APD) (14). Once the cohort of ICU admissions was established, further linkages to the state-wide Queensland Hospital Admitted Patient Data Collection (QHAPDC) and Register of Deaths were performed (15).

The Charlson Comorbidity Index was determined using validated algorithms using International Classification of Disease version 10 Australian Modification (ICD-10AM) discharge codes from the QHAPDC (16). Severity of illness was scored using the Acute Physiology and Chronic Health Evaluation (APACHE) II score, and standard ANZICS APD definitions were applied unless otherwise specified (14). A primary admitting diagnosis by body system was assigned based on APACHE diagnostic codes.

All positive blood culture tests obtained during hospitalization were initially included. Episodes of incident ICU-associated BSI were then identified and classified using previously developed algorithms (17). Incident BSIs were defined by the first positive (index) blood culture for a species per patient, with repeat isolations of the same species within 30 days deemed to represent the same episode. Polymicrobial BSI were those where two or more species were isolated within 48 hours of the index episode. For low-virulence organisms commonly associated with blood culture contamination, two separate positive blood cultures within 48 hours of the index draw were required for inclusion (17).

Definitions

ICU-associated BSI was defined by the index-positive culture obtained within 48 hours before ICU admission and within 48 hours of ICU discharge. Those represented all BSIs in ICU patients and were further subclassified as follows:

• Community-onset BSI, where the culture was obtained during the first 48 hours of hospitalization.

• Hospital-onset BSI, if the index culture draw was 48 hours or more after hospital admission and less than 48 hours after ICU admission.

• ICU-onset BSI, if the culture was drawn 48 hours or more after ICU admission.

Analysis

Analysis was performed using Stata 17 (StataCorp, College Station, TX, USA). The primary unit of analysis was ICU admission episodes. Where case-fatality was the outcome, only the first episode was included in analysis. Continuous variables were examined for underlying distribution using histograms. Skewed variables were reported as medians with interquartile ranges (IQRs) and were compared using Kruskal-Wallis tests. Categorical variables were reported as percentages and compared using χ² or Fisher exact tests. For multigroup comparisons by onset category, it was a priori established that further pairwise tests would be limited to hospital-onset and ICU-onset group comparisons, and only performed when the multigroup comparison was significant. Two-tailed p-values of <0.05 were deemed to represent statistical significance for all analyses.

Results

A total of 35,555 admissions among adults with an ICU length of stay of at least 48 hours were included. The median age was 60.7 (IQR 47.0–71.1) years, 21,864 (61.5%) were male, and the median APACHE 2 score was 19 (IQR 14–25). Admissions were for elective surgery in 5,913 (16.6%), emergency surgery in 6,367 (17.9%), and for other reasons in 23,280 (65.5%). Institutions were classified as tertiary, rural/regional, and metropolitan in 25,883 (72.8%), 5,641 (15.9%), and 4,031 (11.3%) instances, respectively.

Overall, a total of 3,540 episodes of ICU-associated BSI were identified among 3,368 ICU admissions, with 1,693 (47.8%) of these community-associated, 663 (18.7%) hospital-onset, and 1,184 (33.4%) ICU-onset. A total of 153 patients had multiple-incident ICU-associated BSIs during an admission episode, with 13 having three, 3 having four, 2 having five, and 1 individual having six episodes. Figure 1 maps the location of onset of ICU-associated BSIs and further episodes among the study cohort.

Figure 1: Study population and location of ICU-associated BSI episodes.

BSI = Bloodstream infection; ICU = Intensive care unit

Among those with a single episode of ICU-associated BSI (n = 3,215), the time from ICU admission to index blood culture draw was a median of −0.2 (IQR −0.4 to −0.1) and −0.1 (IQR −0.4 to −0.1) days for community-onset and hospital-onset BSIs, respectively, and was 6.9 (IQR 3.9–11.4) for ICU-onset BSIs.

Several significant differences were observed in relation to demographic, clinical, and microbiology variables according to onset type, as shown in Table 1. Notably, patients with ICU-onset BSIs were younger, had fewer comorbid illnesses, lower APACHE II scores, and were more likely male. ICU-onset BSIs were proportionally more likely to have a surgical admission diagnosis with cardiovascular and neurological systems most commonly observed.

Table 1:

Clinical characteristics of intensive care unit-associated bloodstream infections by onset classification

Factor	Community-onset (n = 1655)	Hospital-onset (n = 623)	ICU-onset (n = 937)	p-value
Median age (IQR)	62.4 (49.4–72.0)	64.6 (53.7–73.2)	58.4 (43.8–69.4)	<0.001*
Male sex	969 (58.6%)	396 (63.6%)	652 (69.6%)	<0.001*
Median APACHE2 (IQR)	22 (18–27)	23 (18–29)	21 (16–28)	<0.001*
Median Charlson (IQR)	2 (0–3)	2 (1–4)	2 (1–3)	<0.001*
Charlson categories				<0.001
None	414 (25.0%)	106 (17.1%)	228 (24.3%)
Mild (1–2)	689 (41.6%)	221 (35.5%)	371 (39.6%)
Moderate (3–4)	409 (24.7%)	191 (30.7%)	228 (24.3%)
High (5+)	143 (8.6%)	105 (16.9%)	110 (11.7%)
ICU admission				<0.001*
Diagnostic category	143 (8.6%)	97 (15.6%)	266 (28.4%)
Cardiovascular	227 (13.7%)	93 (14.9%)	115 (12.3%)
Respiratory	145 (8.8%)	55 (8.8%)	113 (12.1%)
Musculoskeletal/skin	234 (14.1%)	180 (28.9%)	145 (15.5%)
Gastrointestinal	16 (1.0%)	42 (7.7%)	13 (1.4%)
Blood disorders	89 (5.4%)	39 (6.3%)	192 (20.5%)
Neurological	61 (3.7%)	7 (1.1%)	33 (3.5%)
Endocrine/metabolic	325 (19.6%)	58 (9.3%)	24 (2.6%)
Genitourinary	6 (0.4%)	1 (0.2%)	1 (0.1%)
Pregnancy-related Other	409 (24.7%)	51 (8.2%)	35 (3.7%)
Surgical group
Non-surgical	1,378 (83.3%)	493 (79.1%)	647 (69.1%)
Elective surgery	20 (1.2%)	18 (2.9%)	100 (10.7%)
Non-elective surgery	257 (15.5%)	112 (18.0%)	190 (20.3%)	<0.001*
Hospital class				<0.001*
Rural/regional	300 (18.1%)	38 (6.1%)	77 (8.2%)
Metropolitan	391 (23.6%)	124 (19.9%)	88 (9.4%)
Tertiary	964 (58.3%)	461 (74.0%)	772 (82.4%)

P-values refer to three-group comparison. p-values were all <0.001 for two-group comparisons between hospital-onset versus ICU-onset two-group comparison, as indicated by *, with the exception of p = 0.015 for sex

ICU = Intensive care unit; IQR = Interquartile range

As shown in Figure 2, there was a decreasing proportion of Staphylococcus aureus, Escherichia coli, and -hemolytic streptococci, and an increasing proportion of other Enterobacterales, Pseudomonas spp, coagulase-negative staphylococci, and yeast/fungi observed with progression from community-, to hospital-, to ICU-onset BSI groups. Among the 569 BSIs caused by S. aureus, there was no significant difference (p = 0.9) observed in the proportions that were methicillin resistant among community-onset (51; 14.3%), hospital-onset (14; 13.9%), and ICU-onset (14; 12.6%). However, an increasing proportion of extended spectrum -lactamase-producing E. coli were observed, with 31/319 (9.7%), 14/100 (14.0%), and 14/63 (22.2%) among community-, hospital-, and ICU-onset BSIs, respectively (p = 0.021).

Figure 2: Frequency of pathogens for each onset location.

ICU = Intensive care unit

Overall, the all-cause 30-day case-fatality rate was 14.1% (5,004/35,555). Among those with first episodes of community-onset, hospital-onset, and ICU-onset BSI, the 30-day case-fatality rates were significantly different (p < 0.001) at 283/1,655 (17.1%), 135/623 (21.7%), and 220/937 (23.5%), respectively. There was no difference among those with hospital-onset and ICU-onset BSIs (p = 0.4).

Discussion

In this study, we examined the epidemiology of ICU-associated BSIs and found that episodes classified by onset location had different clinical, microbiological, and outcome characteristics. Patients with community-onset BSIs were younger, less frequently male, had fewer comorbidities, and were more frequently admitted for non-surgical reasons compared with those of hospital- and ICU-onset. Across community-, hospital-, and ICU-onset BSI groups, a decreasing proportion of S. pneumoniae, S. aureus, E. coli, and -hemolytic streptococci, and increasing proportion of Enterobacterales, Pseudomonas spp, Enterococcus spp, coagulase-negative staphylococci, and fungi were observed. The case-fatality rate progressively increased across the three categories, reaching 23.5% for ICU-onset patients.

It has been well established that community-onset and hospital-onset infections are different, such that our observations further support this and are not novel (9,11,18,19). Indeed, similar differences in pathogen distribution were shown in 2011 by Valles et al in a multicentre study investigating BSIs on admission to Spanish ICUs (11). They showed a predominance of S. pneumoniae, Streptococcus spp, and E. coli for community-onset BSIs and increasing proportions of Enterococcus spp, Klebsiella spp, Pseudomonas spp, and Candida spp for hospital-onset BSIs. Comorbidities, septic shock, and case-fatality rates were, as for our study, higher for patients with hospital-onset compared with community-onset. However, little attention has been given as to how the occurrence of critical illness may influence hospital-onset infections. Our study shows that ICU-onset BSIs are a distinct subgroup of hospital-onset infections and their identification has clinical and epidemiologic significance. Indeed, we showed lower age, lower comorbidity scores, and a higher proportion of surgical admissions, with more cardiovascular, musculoskeletal, and neurological admission diagnoses for those of ICU-onset when compared with the hospital-onset cohort. The trend in pathogen distribution between community-onset and hospital-onset affected the same pathogens and became more pronounced when we compared hospital-onset and ICU-onset BSIs.

It is an important distinction that our study classified BSIs by location of onset but did not attempt to attribute acquisition. Traditionally, infection-prevention and -control programs have aimed to classify infections as either community- or hospital-(nosocomial) acquired. This objective has, in part, sought to identify those infections that could arise as a complication of hospitalization to target hospital-based preventive efforts. However, an acquisition-based approach is challenging for several reasons. On the one hand, infections could be incubating at the time of admission such that while identified later during hospitalization, the infecting organism could have been community acquired. On the other hand, infections may be acquired in hospitals yet only manifest days or weeks later while in the community (20). Another consideration is that many or most infections that complicate a stay are endogenous in that they arise from a patient's own flora, but may only manifest as an infection due to a health care intervention such as an invasive procedure. While infections occurring in such a case represent an important opportunity for health care improvement, the exact location of acquisition of the organism itself may be of little relevance per se. Additionally, accurate determination of acquisition location requires extensive surveillance testing at and during admission, with subsequent molecular typing of infecting organisms. This is not routinely practical for clinical care purposes. Perhaps most importantly, the determination of acquisition often requires an assessment subject to significant inter-observer variability (21,22).

Classification of BSIs by onset category has been increasingly employed during the past 25 years. Morin and Hadler first proposed the classification of S. aureus BSIs by community-onset or not in the late 1990s (7). Subsequently, in a landmark paper by Friedman et al, the utility of further subcategorizing community-onset BSIs into community- and health care-associated infections was demonstrated (9). Their definitions have subsequently been validated in many populations, including critically ill cohorts. However, in the same way that not all community-onset BSIs are the same, hospital-onset BSIs may not be homogeneous. Our study provides empirical data to support that ICU-onset BSIs represent a distinct category of hospital-onset BSIs.

Although we and others have previously reported on ICU-onset/acquired infections, definitions have neither been widely standardized nor adopted. To explore this further, we conducted a PubMed search through to October 2023 using the terms ‘intensive care unit’ and ‘bloodstream infection’ and with limitation to articles with abstracts, humans, and English language. We then obtained a convenience sample of 20 reports by reviewing titles and abstracts in a reverse chronological fashion to identify articles examining BSIs among patients admitted to ICUs. Five studies used definitions similar to those used in our study (23, 24, 25, 26, 27). Eleven examined hospital-onset/acquired infections among patients admitted to ICU, but it was unclear as to whether onset was specifically in the ICU (2,28, 29, 30, 31, 32, 33, 34, 35, 36), albeit eight of these reported the average time to infection following ICU admission (2,30–36). Four did not classify infectious episodes (37, 38, 39, 40). While this survey only represents a brief look at the literature, it suggests that grouping hospital- and ICU-acquired BSIs together without differentiation is commonplace.

Our study has several strengths and weaknesses that merit discussion. We included a large, multicentred cohort of adult ICUs servicing a range of urban and non-urban settings. While our results should be generalizable to many other jurisdictions, we did not include pediatric cases, and accordingly, our results may not apply to children. Further studies are required to address that specific population. Our study benefited from including data from unit-based, state-wide, and national databases. However, this study was retrospective in design, and we therefore were limited to collecting data fields previously obtained through usual patient care and registry purposes. It is also notable that other than case-fatality assessments, which were limited to one observation per individual, our main study analysis was based on ICU admission episodes. As a result, patients could have been included in more than one group. Our study utilized definitions based on the location at the time of the index blood culture draw. While this may be related, it does not imply a source of acquisition of infection per se. Finally, while we identified significant differences in the microbial etiology of BSIs by onset group, we did not include detailed susceptibility data and may not make suggestions for different empiric therapy regimens.

Conclusion

In conclusion, this large, multicentred study shows that the location of onset is an important determinant of the epidemiology of ICU-associated BSIs and provides a justification for further classifying hospital-onset BSIs identified in critically ill patients as to whether they have onset in the ICU or in other locations.

Funding Statement

This study was funded in part by a Health Translation Queensland (formerly Brisbane Diamantina Health Partners [BDHP]) grant.

Contributors:

Conceptualization, A Tabah, M Ramanan, KC White, K Laupland; Data Curation, A Tabah, M Ramanan, KC White, F Edwards, K Shekar, P McIroy, A Attokaran, S Senthuran, J McCullough, A Kumar, S Luke, N Bhadange, P Garrett; Writing – Original Draft, A Tabah, K Laupland; Writing – Review & Editing, A Tabah, M Ramanan, KC White, K Laupland, F Edwards, K Shekar, P McIroy, A Attokaran, S Senthuran, J McCullough, A Kumar, S Luke, N Bhadange, P Garrett; Methodology, K Laupland; Formal Analysis, K Laupland; Investigation, K Laupland; Project Administration, K Laupland, F Edwards; Funding Acquisition, K Laupland.

Ethics Approval:

This study was approved by the Royal Brisbane and Women's Hospital, Brisbane, Queensland, Australia on October 14, 2020.

Informed Consent:

N/A

Registry and the Registration No. of the Study/Trial:

N/A

Funding:

This study was funded in part by a Health Translation Queensland (formerly Brisbane Diamantina Health Partners [BDHP]) grant.

Disclosures:

The authors have nothing to disclose.

Peer Review:

This manuscript has been peer reviewed.

Animal Studies:

N/A

Data Availability:

Data cannot be shared publicly due to institutional ethics, privacy, and confidentiality regulations. Data release for research under Section 280 of the Public Health Act 2005 requires application to the Director General (PHA@health.qld.gov.au).