Hospital-acquired infections caused by enterococci: a systematic review and meta-analysis, WHO European Region, 1 January 2010 to 4 February 2020

Background Hospital-acquired infections (HAI) caused by Enterococcus spp., especially vancomycin-resistant Enterococcus spp. (VRE), are of rising concern. Aim We summarised data on incidence, mortality and proportion of HAI caused by enterococci in the World Health Organization European Region. Methods We searched Medline and Embase for articles published between 1 January 2010 and 4 February 2020. Random-effects meta-analyses were performed to obtain pooled estimates. Results We included 75 studies. Enterococcus spp. and VRE accounted for 10.9% (95% confidence interval (CI): 8.7–13.4; range: 6.1–17.5) and 1.1% (95% CI: 0.21–2.7; range: 0.39–2.0) of all pathogens isolated from patients with HAI. Hospital wide, the pooled incidence of HAI caused by Enterococcus spp. ranged between 0.7 and 24.8 cases per 1,000 patients (pooled estimate: 6.9; 95% CI: 0.76–19.0). In intensive care units (ICU), pooled incidence of HAI caused by Enterococcus spp. and VRE was 9.6 (95% CI: 6.3–13.5; range: 0.39–36.0) and 2.6 (95% CI: 0.53–5.8; range: 0–9.7). Hospital wide, the pooled vancomycin resistance proportion among Enterococcus spp. HAI isolates was 7.3% (95% CI: 1.5–16.3; range: 2.6–11.5). In ICU, this proportion was 11.5% (95% CI: 4.7–20.1; range: 0–40.0). Among patients with hospital-acquired bloodstream infections with Enterococcus spp., pooled all-cause mortality was 21.9% (95% CI: 15.7–28.9; range: 14.3–32.3); whereas all-cause mortality attributable to VRE was 33.5% (95% CI: 13.0–57.3; range: 14.3–41.3). Conclusions Infections caused by Enterococcus spp. are frequently identified among hospital patients and associated with high mortality.


Introduction
Enterococcus spp. is a genus of Gram-positive, facultative anaerobic, catalase-negative bacteria that commonly inhabit the intestinal tracts of healthy humans and animals [1]. In addition to their role as commensals, enterococci are known for being associated with hospital-acquired infections. They can cause a wide range of infections, including infections of the urinary tract, bloodstream, and endocardium [2]. Enterococci, particularly E. faecalis and E. faecium, are among the most frequently isolated pathogens from patients with hospital-acquired infections (HAI) [1,3,4]. Hospital-acquired infections with enterococci are associated with considerable mortality [5][6][7], morbidity [8,9] and economic burden [10]. The clinical relevance of Enterococcus spp. is emphasised by their intrinsically low susceptibility to a wide range of antimicrobial drugs, including aminoglycosides, cephalosporins and sulphonamides and in the case of E. faecium, low-dose penicillin and ampicillin [11,12]. In view of the dwindling number of treatment options, vancomycin is commonly used to treat enterococcal infections, especially E. faecium. After the introduction of vancomycin in 1958 [13], a profound increase in prescriptions was recorded in the early 1980s [14]. Consequently, the first vancomycin resistance in clinical Enterococcus spp. isolates was observed in 1988 in London, United Kingdom [15]. Since then, vancomycinresistant Enterococcus spp. (VRE) has spread and been detected in healthcare facilities across the world [16]. A rise of vancomycin resistance has been observed in clinical Enterococcus spp. isolates (especially in E. faecium) in many European countries in the last decade

Study outcomes
The primary outcomes of this review are the prevalence, incidence and incidence density of hospital-acquired Enterococcus spp. / E. faecium and VRE / vancomycin-resistant E. faecium (VREF) infections among hospitalised patients and at the population level. Incidence density is defined as new cases per 1,000 patient hospitalisation days. The mortality of patients with HAI caused by Enterococcus spp. / E. faecium and VRE / VREF was additionally studied as a primary outcome. Secondary outcomes are (i) the proportion of vancomycin resistance among all Enterococcus spp. / E. faecium HAI isolates and; (ii) the proportion of HAI with Enterococcus spp. / E. faecium and VRE / VREF among all identified microorganisms from patients with HAI. In our review, cases of HAI caused by Enterococcus spp. / E. faecium include both vancomycin-resistant and sensitive strains.

Search strategy, study selection criteria and data extraction
We searched Medline and Embase for epidemiological and surveillance studies reporting data on HAI. The search was carried out for studies published between 1 January 2010 and 4 February 2020 without any language restrictions. This timeframe was chosen because we aimed to summarise recent data on the epidemiology of hospital-acquired Enterococcus spp., especially given the rise of vancomycin resistance in Europe in the last decade. The detailed search strategy, including search strings, is provided in Supplementary Material. Title, abstract and full-text screening were

Study selection criteria
Studies were included if they met all of the following criteria: • The study provided data for at least one of the predefined primary outcomes for Enterococcus spp. and/or E. faecium. Studies were only included if they provided microbiological results where either the pathogen was identified or the culture was negative for more than 90% of all HAI episodes.
• The study was conducted in the WHO European Region.
• Data collection was completed before 2008 and the study was published after 2009.
• A largely unselected patient cohort was studied, i.e. not only high-risk patients such as low birthweight neonates or elderly patients, etc. or those with a specific underlying disease.
• The study was published in English, French, German or Spanish.
• Only studies that reported data for total HAI and HA-BSI were included.
Studies were excluded if: • Data was provided for HAI outside of hospitals, such as nursing homes.
• Studies with any of the following study designs were excluded: literature reviews, intervention studies, case-control studies, outbreak studies and case series. independently performed by three authors (SB, OA, RM) using Covidence, a screening and data extraction tool recommended by the Cochrane Community [22]. All disagreements were discussed for consensus or resolved by a third reviewer.
The study selection criteria are presented in Box 1.
The data of all eligible studies were independently extracted by three authors (SB, OA and RM). All disagreements were resolved through discussion. The data extraction included the primary and secondary outcomes and the following study characteristics: authors, year of publication, study period, country, study design, setting (e.g. hospital, intensive care unit (ICU), etc.), age groups, patient inclusion criteria and the HAI infection type (i.e. total HAI and HA-BSI).
We contacted study authors via email where details regarding outcomes and reporting were needed.

Risk of bias assessment and statistical analysis
The risk of bias for individual studies was assessed by two authors (SB and RM) using the risk of bias tool developed by Hoy et al. [23]. For data analysis and presentation, studies were grouped into hospital-wide, ICU-based, neonatal ICU-based studies, and other hospital units/wards (e. g. internal medicine, surgical units, etc.) as well as by HAI types (i. e. total HAI and HA-BSI). All statistical analyses were performed using R version 3.6.1 and the R package meta version 4.9.7 (R Foundation, Vienna, Austria) [24]. Pooled estimates were calculated using random-effects models with a Tukey Double Arcsine transformation [25] of the raw proportions. The DerSimonian-Laird estimator was used to define τ 2 (between-study variance). The I 2 statistics quantified the statistical heterogeneity of the selected studies.

Results
In total, we identified 6,069 unique records. After title and abstract screening, 362 studies were assessed in full-text review and 75 [6,7, met all inclusion criteria ( Figure 1).
The results specifically for E. faecium and VREF are not presented in the main text of this study but are instead described in the Supplementary Material.

Risk of bias assessment
The risk of bias for the representativeness of the studied hospital population was assessed as high in the majority of studies (69/75) (Supplementary Material, Table S8. Risk of bias assessment of included studies). Since these studies were single centre studies and/or included data from patients treated in academic medical centres, the representativeness of the included patients for the general hospital population in a given region or country was therefore unclear or low in these studies. Six studies [33,36,37,40,46,61] included nationally representative hospital populations. The risk of bias for the applied case definitions (i.e. hospital-acquired infections) was judged as low for most studies, since the majority of the studies (55/75) used HAI definitions based on the United States (US) Centers for Disease Control and Prevention criteria and the National Healthcare Safety Network criteria [99,100]. These validated definitions are widely used in the surveillance of HAI. More than half (46/75) of the studies did not report the used pathogen identification and/or antimicrobial susceptibility testing method and/or interpretation guideline (e.g. The European Committee on Antimicrobial Susceptibility Testing, Clinical and Laboratory Standards Institute). For this reason, the risk of bias with regards to the validity and reliability of the methodology used in these studies to identify enterococci and vancomycin-resistant strains was considered high (item 7, Supplementary Material, Table S8). In epidemiological surveys, HAI are typically defined as infections that occur 48 h after admission. That means that only patients with a hospital stay longer than 48 h in these studies are at risk of developing HAI and hence represent the appropriate denominator population for the parameters of interest (i.e. prevalence, incidence and mortality). Consequently, only studies including patients with a hospital stay longer than 48 h are judged as low risk of bias for item 10 (33/75 studies).

Figure 2
Incidence of hospital-acquired infections caused by Enterococcus spp. and vancomycin-

Mortality
The all-cause mortality recorded among patients with HA-BSI caused by Enterococcus spp. ranged between 14.3% and 32.3% (pooled estimate: 21.9%; 95% CI: 15.7-28.9, five studies) (Figure 3; Table). Based on two studies [40,47]  Pooled estimates were calculated using random-effects models with a Tukey Double Arcsine transformation of the raw proportions. The DerSimonian-Laird estimator was used to define τ2 (between-study variance).

Discussion
In  [4]. Another explanation might be broader screening practices and the implementation of contact precaution measures within the US healthcare system, particularly to control meticillin-resistant Staphylococcus aureus (MRSA) and VRE [108].
Our study emphasises the importance of Enterococcus spp. as a nosocomial pathogen, since it accounts for 6.1% to 17.5% of all pathogens isolated from patients with HAI. Enterococcus spp. usually remains among the top five most frequent nosocomial pathogens in Europe, despite the variation in species distribution across hospitals and regions [101,[109][110][111]. In comparison, Enterococcus spp. is less frequently found in isolates from patients with HAI in the US [104] and China [105], 5% and 3.1% of all HAI pathogens, respectively. Our data show that VRE was found in 1.1% (range: 0.4-2.0) of all pathogens isolated from HAI patients, which is lower than the mean proportion of MRSA (ca 5%) observed in Europe [112]. However, in Germany [101] and Greece [109] VRE and MRSA are equally often found in HAI patients and in studies from Italy [110] and Ukraine [49], VRE is even more frequently isolated than MRSA, underlining the local heterogeneous distribution of nosocomial antibiotic-resistant pathogens. Interestingly, we found that Enterococcus spp. is less frequently isolated from HAI patients in ICU compared with patients treated hospital wide (10.9% vs 3.8%). However, the reasons for this observation are unclear.
Our study shows that the pooled vancomycin resistance proportions among HAI Enterococcus spp. were 7.3% hospital wide and 11.5% in isolates from patients in ICU, although individual study estimates varied somewhat. These pooled estimates are similar to the European Centre for Disease Prevention and Control data from the European Point Prevalence Survey [113,114]. In comparison to these European data, vancomycin resistance proportions are substantially lower in China [115,116] and Japan [117], where VRE proportions lower than 2% were observed. Interestingly, other countries in eastern Asia observed much higher VRE proportions, such as in South Korea (33.4%) [118] and Taiwan (40%) [119]. Compared with the European estimates, VRE proportions in the US are also generally higher (> 20%) [120,121], which might be explained by the widespread use of vancomycin in US hospitals, which increased by more than 30% between 2006 and 2012 [122].
For patients with HA-BSI caused by Enterococcus spp., all-cause mortality estimates ranged between 14.3% and 32.3% (pooled estimate: 21.1%). These are higher [35,38,40] or similar [7,48] to the all-cause mortality rates observed for S. aureus and generally higher than those reported for E. coli [7,35,38,40], which are other frequently encountered nosocomial pathogens. Substantial attention is paid to infection prevention and control (IPC) measures to address VRE, but our results show that enterococcal HAI as a whole are associated with a high incidence and mortality in Europe and should therefore receive more attention in IPC strategies.
An important observation of our study is that there is a large variation between individual study estimates of incidences/prevalences of HAI caused by Enterococcus spp. as well as for VRE proportions. This finding is similar to other systematic reviews around the world that also found large inter-study variations in the frequency of HAI [123][124][125]. Some of this heterogeneity might be explained by different methodological approaches, including different inclusion/exclusion criteria and microbiological sampling routines. In many published studies, data on the causative pathogen are not available for a substantial proportion of HAI episodes (> 40%) because of the lack of microbiological samples taken or incomplete data. This would ultimately lead to a substantial underestimation of the frequency of HAI caused by Enterococcus spp. To avoid this source of bias, we only included studies where pathogen identification results were reported for almost all HAI episodes. In addition to methodological differences, the large variation between individual study estimates also reflects true differences in the occurrence of nosocomial pathogens, including Enterococcus spp., between countries, regions and individual hospitals. For example, in a large multicentre study from Ukraine [49], Escherichia coli, Staphylococcus aureus and Enterococcus spp. were the predominant pathogens isolated from patients with HAI, while in a multicentre study from Greece, Klebsiella spp., Pseud omonas aeruginosa and Acinetobacter spp. were the most frequently identified nosocomial pathogens [126]. Furthermore, there is great variation in IPC policies and resources across Europe [127], which also explains the observed variations of HAI caused by Enterococcus spp. and VRE.
Since vancomycin resistance is predominantly found in E. faecium and less in E. faecalis and/or other enterococci species [7,18,75,118], vancomycin resistance proportions in Enterococcus spp. HAI isolates are also largely influenced by the proportion of E. faecium among all Enterococcus spp. isolates. Moreover, vancomycin resistance proportions in E. faecium differ across countries [18] and even within countries [17], which also explains the observed variation in VRE proportion described in our study. Moreover, nosocomial outbreaks and local spread of E. faecium genotypes associated with vancomycin resistance especially VanA and VanB in Europe [128] and increased virulence such as the esp and hyl genes can result in a higher VRE incidence in certain regions and hospitals. Another explanation for the observed inter-study variations in HAI caused by VRE are the profound differences in the consumption of glycopeptides/vancomycin, fluoroquinolones and third generation cephalosporins in Europe [129,130], whose usage is associated with VRE infections and colonisations in hospitals [131][132][133][134][135].
This systematic review is a comprehensive summary of recent data on the epidemiology of Enterococcus spp. and VRE in the WHO European Region, including 75 studies with data on over 8.5 million hospitalised patients with 154,000 HAI episodes. The majority of studies were based on routine HAI surveillance systems, including data from unselected patient cohorts. However, because of language restrictions in the literature selection, potentially relevant studies might have been excluded, for example from eastern European countries. Also, the majority of the included studies were conducted in academic medical centres and/ or tertiary care hospitals and the representativeness of hospitalised patients and external validity of the study results might therefore be limited. Despite unclear representativeness of most studies, the overall quality of the studies and thus the quality of evidence was moderate to high. Another limitation is that many studies did not report vancomycin resistance profiles of Enterococcus spp. and data on the epidemiology of VRE are therefore limited. Importantly, since enterococci frequently colonise healthy people and are often detected in mixed infections, they may not be the causative microorganism in all HAI reported by the included studies. Especially in intra-abdominal, pelvic and soft tissue infections, the clinical relevance of Enterococcus spp. is debated [136]. Although the included studies were conducted in 21 different countries in the WHO European Region, the studies were not evenly distributed across Europe, which might lead to a geographical bias. For example, ICU-based studies were predominantly from studies in eastern and southern Europe and none was conducted in Scandinavia. Notably, studies from Turkey were overrepresented within the study set reporting VRE data. However, Turkish data did not systematically differ to data from other European countries. More nationally representative studies with complete microbiological and antimicrobial resistance profiles, including populations-based data, are needed in order to fully understand the epidemiology of HAI caused by Enterococcus spp. and VRE. In most analyses, a large statistical heterogeneity was observed (I 2 > 80%) and the pooled estimates should be interpreted with caution. We therefore also provided the range of individual study estimates for all outcomes.

Conclusions
Our data show that HAI caused by Enterococcus spp. and VRE are frequently identified among hospital patients and associated with high mortality in the WHO European Region. Continuous monitoring and the improved implementation of infection prevention and control programs as well as antibiotic stewardship measures are essential to reduce the burden of HAI caused by enterococci.

*Note
This designation is without prejudice to positions on status, and is in line with UNSCR 1244/99 and the ICJ Opinion on the Kosovo declaration of independence.

Authors' contributions
SB, TE and RM designed the study. SB, OA and performed literature screening, study selection and data extraction. SB and RM assessed the risk of bias. RM conducted the statistical analyses. SB and RM led the writing of the manuscript. All authors revised the manuscript for important intellectual content.