Clinical and Molecular Epidemiology of Extended-Spectrum Beta-Lactamase-Producing Klebsiella spp.: A Systematic Review and Meta-Analyses

Healthcare-related infections caused by extended-spectrum beta-lactamase (ESBL)-producing Klebsiella spp. are of major concern. To control transmission, deep understanding of the transmission mechanisms is needed. This systematic review aimed to identify risk factors and sources, clonal relatedness using molecular techniques, and the most effective control strategies for ESBL-producing Klebsiella spp. A systematic search of PubMed, Embase, and Outbreak Database was performed. We identified 2771 articles from November 25th, 1960 until April 7th, 2014 of which 148 were included in the systematic review and 23 in a random-effects meta-analysis study. The random-effects meta-analyses showed that underlying disease or condition (odds ratio [OR] = 6.25; 95% confidence interval [CI] = 2.85 to 13.66) generated the highest pooled estimate. ESBL-producing Klebsiella spp. were spread through person-to-person contact and via sources in the environment; we identified both monoclonal and polyclonal presence. Multi-faceted interventions are needed to prevent transmission of ESBL-producing Klebsiella spp.


Introduction
Healthcare-related infections (HRIs) are a major clinical problem worldwide. In 2011, the World Health Organization (WHO) reported that in a mixed patient population the pooled HRI-prevalence was 10.1% in low-and middle-income countries and 7.6% in high-income countries [1]. Prolonged hospital stay, higher costs, increased antimicrobial resistance, and risk of potentially life-threatening conditions indicate the enormous burden of HRIs [2]. Further, we are facing HRIs caused by multidrug-resistant gram-negative bacteria (MDR-GNB) without a parallel progression of the novel antibiotic classes [3].
Klebsiella spp. have been recognized as the most frequent cause of MDR-GNB outbreaks, particularly after the emergence of the extended-spectrum beta-lactamase (ESBL) enzymes [4,5]. As a result, infections in hospitalized patients with this ESBL-producing Klebsiella spp. have

Data Analyses
We included articles related to ESBL-producing Klebsiella spp. that described the factors associated with the presence or acquisition of ESBL-producing Klebsiella spp. using a multivariate model. We took into account studies that have suggested and proven the sources of ESBL-producing Klebsiella spp. using molecular typing techniques. However, we also included studies that suggested the potential reservoirs. In addition, we included studies about the associated factors for mortality related to ESBL-producing Klebsiella spp.
In order to assess clonal relatedness and transmission patterns of ESBL-producing Klebsiella spp., studies that only performed phenotypic typing methods were excluded and studies that did use molecular typing were included. We merged studies that used polymerase chain reaction (PCR)-based techniques for typing. We assessed the result of molecular typing methods and calculated the total number of identified patterns. We defined a cluster as two similar patterns of ESBL-producing Klebsiella spp. isolates. Likewise, a unique isolate was defined as a single pattern. The term monoclonal presence referred to a single cluster and the term polyclonal presence referred to two clusters. We calculated the total number of patterns, the clusters including cluster sizes, and the single patterns. If the information was available, we performed the cluster analyses based on the number of patients, otherwise on the number of isolates. We also reviewed studies about infection control strategies and prevention programs. We assessed the standard interventions possibly combined with additional control strategies, and reported which were most successful strategies to stop transmission. We compiled data from two studies that were presented in four publications in the result section [15][16][17][18].

Statistical Analysis
We combined all associated factors that reported an odds ratio (OR) and a 95% confidence interval (95% CI) into ten different categories: 1, medical devices (e.g. mechanical ventilation, intra-vascular devices); 2, prior cephalosporin exposure; 3, prior quinolone exposure; 4, prior other antibiotic exposure; 5, prior antifungal exposure; 6, length of hospital stay; 7, patient characteristics (e.g. age); 8, underlying disease or condition (e.g. malignant disease); 9, medical procedures (e.g. surgical intervention); 10, other (e.g. exposure to the hands of HCWs). Studies reporting associated factors for mortality were excluded. Random-effects meta-analyses were performed for all categories except the "prior antifungal exposure", "medical procedures" which had less than three factors and "other" that comprised many various factors. Lytsy et al., used three different models of multivariable analyses to find reliable estimates for the most important variables. However, we chose to only include model 1 in our meta-analyses [18]. We applied the method of DerSimonian and Laird and meta-analyses were performed using Stats-Direct statistical software (StatsDirect, Version 2.8.0, Altrincham, StatsDirect Ltd, 2013) [19]. We considered P values <0.05 as statistically significant. Bias assessment plots were constructed to explore publication bias using the Egger and Begg-Mazumbdar (Kendall's tau) indicators [19,20].

Study Quality
The methodological quality of all studies in the random-effects meta-analyses was assessed using the strengthening the reporting of observational studies in epidemiology (STROBE) guidelines (Table A and Table B in S3 File) or the Newcastle-Ottawa quality assessment scale (Table A and Table C in S3 File), based on the study design [21,22]. Furthermore, the methodological quality of all molecular epidemiological studies in cluster analysis was assessed by the strengthening the reporting of molecular epidemiology for infectious disease (STROME-ID) guidelines (Table D in S3 File) [23]. However, the study quality was not considered as an exclusion criterion.

Results Description
We identified a total of 5,608 articles as potentially relevant when using our search strategy (Fig 1). Of these, 835 articles met the eligibility criteria and 25 articles were retrieved from citations of interesting reviews. We received 45 full-text articles from 80 authors. We got three further responses to the information requests that were sent to the authors of 17 articles. We ultimately included 148 articles in this systematic review (Fig 1). Five articles were written in Spanish, one article was written in French, one article was written in Turkish and 141 articles were written in English. The non-English articles were translated prior to data extraction. Most studies were conducted in Europe (39.9%; n = 59) particularly in France (n = 16), followed by Asia (18.9%; n = 28), North America (16.9%; n = 25), Africa (9.5%; n = 14), South America (10.8%; n = 16), multiple regions: Europe and Asia (3.4%; n = 5) and Australia (0.7%, n = 1) with 49 countries in total. All studies were published between 1991 and 2014 with study periods of less than a month up to seven years. Thirty-one studies (20.9%) indicated the suspected or identified index case. The study population was predominated by adult patients (45.3%, n = 67), followed by neonates (33.1%; n = 49), pediatric patients (5.4%; n = 8), and mix groups including the studies located in hospitals but which did not mention the study population (16.2%; n = 24). Seventy-six percent of all studies took place at the ICU (n = 113). These studies can be further divided in neonatal ICU (38.1%; n = 43), pediatric ICU (3.5%; n = 4), adult ICU (48.7%; n = 55) and mix ICU units (9.7%; n = 11). However, fifty-two studies were located both at ICU and non-ICU.

Associated Factors
We identified 26 studies reporting associated risk factors with a statistically significant OR above one (Table 1) and seven studies reporting associated protective factors with a statistically significant OR below one (Table 2) for the presence of ESBL-producing Klebsiella spp. In addition, four studies identified the associated factors for bloodstream infections caused by ESBLproducing K. pneumonia (Table 1). In general, prior antibiotic exposure was the most common associated factor in all studies. Four studies reported associated factors for mortality [24][25][26][27]. It was published in these four studies that the presentation with septic shock had the highest odds ratio (205.99) [24].

Eight random-effects meta-analyses
Twenty-three studies were included in the seven random-effects meta-analyses, reporting 54 associated risk factors with a statistically significant OR above one and five associated protective factors with a statistically significant OR below one (Fig 1 and Table 3). The category of underlying disease or condition (OR = 6.25; 95% CI = 2.85 to 13.66) and prior cephalosporin exposure (OR = 4.65; 95% CI = 2.83 to 7.65) generated the highest pooled estimates (Fig 2). The publication bias indicators showed no significant results (Table 3).

Sources and reservoirs
Fifty-seven studies identified environmental sources and/or surrounding reservoirs for ESBLproducing Klebsiella spp. (Table 4). Contaminated sinks were the most reported sources in the environment (13.8%; n = 4) whereas the patients were the main reservoirs (48.9%; n = 23), followed by the hands of HCWs (25.5%; n = 12). Interestingly, one study showed food as a transmission source for ESBL-producing K. pneumonia [28].

Cluster Analyses
One hundred-twenty studies described the molecular methods used to type ESBL-producing Klebsiella spp. and the analyses of genetic similarity (Table 5). In particular for ESBL-producing K. pneumoniae, ninety-two studies used PFGE and 16 studies used PCR-based techniques Prior antibiotic exposure a 1 54 OR 0.003 [56] Prior carbapenem exposure a 1 206 HR 0.22 [55] Age c 1 47 OR 0.95 [50] Prior cephalosporin exposure 1 204 OR 0.1 [44] Abbreviations: RE, risk estimate; HR, hazard ratio; OR, odds ratio. a This factor was classified in the category of prior antibiotic exposure for a random-effects meta-analysis study.
b This category was not included in a random-effects meta-analysis study. c This factor was classified in the category of patient characteristics for a random-effects meta-analysis study.

Effective Interventions
The identification of ESBL-producing Klebsiella spp. transmission in hospitals should be followed by infection control strategies. However, not all studies provided detailed information  regarding the interventions. We identified 84 studies that described the interventions during the study period. Twenty-eight studies reported the standard interventions that did not succeed. All but three studies described the additional and/or successful strategies to prevent the spread of ESBL-producing Klebsiella spp. in hospitals. Ultimately, we presented the top ten strategies to control ESBL-producing Klebsiella spp. in hospitals from 71 studies ( Table 6). Reinforcement of hand hygiene (46.5%) was the most successful intervention in all studies, followed by adequate compliance with the antibiotic control programs (33.8%). Removal of contaminated tools was also found in the list of top 10 strategies to control ESBL-producing Klebsiella spp. infections within the hospital.

Summary of evidence
This is the first systematic review and meta-analyses to identify the transmission pattern of ESBL-producing Klebsiella spp. among hospitalized patients worldwide. Our random-effects meta-analyses showed the underlying disease including malignancy or particular condition (OR = 6.25; 95% CI = 2.85 to 13.66) and prior cephalosporin exposure (OR = 4.65; 95% CI = 2.83 to 7.65) as the most significant associated factors for the presence of ESBL-producing Table 6. Top 10 strategies to control ESBL-producing Klebsiella spp. outbreaks in hospitals.
Our study confirmed the importance of interaction between hospitalized patients, environmental sources, and surrounding reservoirs, in particular the HCWs. Obviously, the colonized and/or infected patients were the consistent reservoirs (Table 1). Further, it was clear that surrounding reservoirs and sources must be taken into consideration in order to end the vicious cycle of the spread of ESBL-producing Klebsiella spp. In general, likewise in case of Methicillinresistant Staphylococcus aureus, HCWs who are colonized with ESBL-producing Klebsiella spp. play a pivotal role in the transmission through their hands [33]. Even more, using artificial nails and having onychomycosis can be the transmission source for the presence of ESBL-producing Klebsiella spp. [34,35]. Molecular typing. PFGE still is the gold standard for molecular typing, in particular to prove the transmission of ESBL-producing Klebsiella spp. among patients [9]. However, it is useful to perform another molecular technique in order to confirm the result of the main diagnostic technique or transmission [36,37]. Furthermore, our findings showed that both monoclonal and polyclonal presence occur in outbreak events. This indicated that clonal dissemination plays a role in outbreaks [38]. However, it is also important to consider the horizontal plasmid transfer among different bacterial species [36]. Studies on plasmid fingerprinting showed the importance of plasmid genotyping to predict the mode of plasmid transmission among patients [37,39].
Interventions. To stop the spread of ESBL-producing Klebsiella spp., reinforcement of hand hygiene and an antibiotic control program were the most successful interventions. Firstly, hand hygiene is a simple and low cost intervention to prevent the presence of ESBL-producing Klebsiella spp. The introduction of hand antiseptic will reduce the contamination of the hands with of ESBL-producing Klebsiella spp. [40]. Recent guidelines also highlight hand hygiene as the top priority to prevent the transmission of nosocomial infections [11]. Secondly, a metaanalysis study suggested cycling empirical antibiotic therapy to prevent antibiotic resistance [41]. This adjustable cycling model confirmed the antibiotic control program as an important strategy for ESBL-producing Klebsiella spp. transmission.
Study quality. Most studies included in the meta-analyses and cluster analysis had a low study quality using the recommended guidelines. However, in particular for the studies that were assessed with STROME-ID guidelines, they had wide range of key objectives that did not focus solely on molecular epidemiology of ESBL-producing Klebsiella spp. Nonetheless, reporting the cases of ESBL-producing Klebsiella spp. according to STROME-ID guidelines can give clear information about the evidence to detect the transmission dynamics with molecular typing and helps in making the health-policy decision. In general, the studies that were included answered our research questions; however, our study indicated the importance to report structured articles that follow appropriate guidelines for a high study quality.

Strengths and Limitations
The major strength of our study was the inclusion of a large number of hospitalized patients with ESBL-producing Klebsiella spp. In the meta-analyses, we identified several risk factors that were associated with the presence of ESBL-producing Klebsiella spp. We also summarized the most successful interventions to prevent the spread of ESBL-producing Klebsiella spp., which to our knowledge has never been done before. However, our study also has some limitations. Firstly, the heterogeneity of all included studies in combination with the diverse models used in all individual statistical analyses, here combined in the meta-analyses. Therefore, we performed a random-effects model for the meta-analyses. Secondly, due to the different definitions of colonization and infection in the studies, we termed them as presence of ESBL-producing Klebsiella spp. As a consequence, reporting them separately was not possible. Thirdly, our study focused on clonal spread; hence, we did not include studies about plasmid transfers. Fourthly, some studies only performed molecular typing investigation on selected samples due to cost effectiveness. Therefore, most probably we have bias on the number of clusters. Fifthly, publication bias might have occurred since we have a broad range of articles from different objectives to comprehensively answer our research questions.

Conclusion
The presence of ESBL-producing Klebsiella spp., which results in increased morbidity and mortality, may occur by either direct spread from patient to patient or indirect transmission via surrounding reservoirs and sources in the environment. Obviously, molecular typing techniques can identify transmission of ESBL-producing Klebsiella spp. within the hospital. Prior antibiotic exposure holds a key role for the presence of ESBL-producing Klebsiella spp. particularly cephalosporin use. Multi-faceted interventions, including reinforcement of hand hygiene and control of antibiotic use, are necessary to prevent the spread of ESBL-producing Klebsiella spp. Further studies on plasmid transfer are needed to learn more about transmission of ESBLproducing Klebsiella spp. within a hospital. In addition, it is important to report studies in a more structured way that systematically follows suitable guidelines. S3 File. Study quality. Quality assessment scores of the 23 articles that were included in the random-effects meta-analysis study (Table A). Quality assessment scores of the 23 articles that were included in the random-effects meta-analysis study, using the STROBE guidelines (Table B). Quality assessment scores of the 15 case-control studies that were included in the random-effects meta-analysis study, using the Newcastle Ottawa Scale (Table C1). Quality assessment scores of the three cohort studies that were included in the random-effects metaanalysis study, using the Newcastle Ottawa Scale (Table C2). Quality assessment scores of 120 articles that were included in the cluster analysis study (Table D)