Background
Escherichia coli and other
Enterobacteriaceae frequently cause nosocomial and community acquired infections [
1] and the burden due to infection with
Enterobacteriaceae resistant to extended-spectrum cephalosporins (ESC) is increasing worldwide [
2]. Infections are often preceded by colonisation of the gut with these bacteria. Among
Enterobacteriaceae, resistance to ESC is often caused by production of extended-spectrum beta-lactamases (ESBLs) or plasmid mediated AmpC (pAmpC) [
2]. In a hospital setting, carriage of
Enterobacteriaceae resistant to ESC (ESCr-E) is not only a concern for the individual patient, but also for the surrounding patients as these bacteria can spread via the faecal-oral route between patients within the hospital [
3]. In addition, plasmids conferring ESC resistance can be transferred between different bacterial strains and species. Outbreaks of ESBL-producing
E. coli and
Klebsiella spp. have been reported in healthcare institutions in Norway [
4‐
6]. Yet another concern is that ESC-resistance often occurs along with co-resistance to important non-beta-lactam antibiotic classes [
7], significantly limiting treatment options for these patients.
Enterobacteriaceae are part of the normal gut flora and pose a complex challenge as eradication is problematic [
8]. If introduced within a hospital, infection prevention and control rely on strict compliance to standard precautions [
3,
9]. The Norwegian Institute of Public Health (NIPH) recommends isolation of patients in general hospital wards if they carry carbapenemase producing
Enterobacteriaceae or vancomycin resistant enterococci. Patients who carry ESC resistant
Enterobacteriaceae should be isolated if admitted to departments with particularly vulnerable patients (neonatal, cancer or intensive care units) [
10]. Even though the Nordic countries have a low prevalence of antimicrobial resistance among
Enterobacteriaceae, resistance to ESC in
E. coli and
Klebsiella ssp in urine and blood specimens is increasing [
11]. In order to better target interventions to prevent spread between patients and to guide empirical antibiotic treatment we aimed to estimate the prevalence of faecal carriage of ESC resistant
E. coli and
K. pneumoniae (ESCr-EK) and vancomycin resistant enterococci (VRE) in patients upon hospital admission, and to identify factors associated with carriage. Additionally we aimed to explore the diversity of ESCr-EK and determine the genetic background for resistance to ESC.
Methods
Design, setting and study population
The cross-sectional study was carried out at Akershus University Hospital, a secondary care facility located in a mixed urban and rural area east of the Norwegian capital Oslo. The catchment area covers around 10% (500,000 persons) of the Norwegian population. Trained nurses recruited patients upon admission. In the adult’s emergency department, recruitment took place on selected weekdays between October 2014 and July 2015, all patients presenting between 8 am and 11 am were invited to participate. In the children’s emergency department, all patients/parents admitted during selected weeks in the period between January 2015and July 2016 were invited to participate. From October 2015 to December 2016 recruitment was expanded to three surgical units (thoracic, urology, and orthopaedic) where all patients presenting for pre-surgical examination prior to elective surgery were invited to participate.
Patients with suspected acute coronary disease, in need of immediate intensive care, and adult patients unable to give informed consent were not recruited.
Data collection
Consenting patients or parents of patients below 16 years of age completed a questionnaire with 26 questions, including demographics and possible risk factors for colonisation with ESCr-EK such as travel outside of the Nordic countries, antibiotic consumption, health care and medical devices within the 12 months prior to recruitment. Only information about risk factors prior to admission was collected. One rectal swab was collected at admission or maximum 24 h after admission.
Microbiology
Rectal swabs were collected and transported in liquid medium (eSwab, Copan, Italy). The samples were stored at 4 °C for maximum 14 h, with subsequent plating on blood agar for growth control, and on two selective lactose agar plates with the addition of either 1 mg/ml cefotaxime or 1 mg/ml ceftazidime for detection of ESCr-EK, produced in-house. For the detection of vancomycin resistant Enterococci, we used Brilliance VRE plates (Oxoid, ThermoFisher Scientific, MA, USA). Morphologically distinct colonies from each selective plate were subcultured, and identified by MALDI-TOF mass spectrometry (MALDI-TOF, Bruker Daltonics, Bremen, Germany). Isolates of
E. coli or
Klebsiella spp. were frozen at -80 °C for later analysis. We did susceptibility disk testing and interpretation of clinical breakpoints according to EUCAST methodology (
http://www.eucast.org/fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_tables/Breakpoint_table_v_4.0.xls, 2017), and divided the antibacterial agents into six classes: Penicillins (Ampicillin), cephalosporins and penicillins in combination with betalactam-inhibitors (cefoxitin, cefotaxime, ceftazidime and piperacillin-tazobactam), carbapenems (meropenem), (fluoro)quinolones (ciprofloxacine and nalidixin), aminoglycosides (gentamicin), and folate-pathway inhibitors (trimethoprim-sulphamethoxazole). We defined multidrug resistance as non-susceptibility (i.e. resistant or intermediate susceptible) to three or more of the above antibiotic classes.
From each patient positive for ESCr-EK, one isolate was selected for whole genome sequencing (WGS). If the patient carried both ESC resistant
E. coli and
Klebsiella spp., one isolate per species was included for WGS. Furthermore, if a patient carried several morphologically different strains of the same species, and if the strains had different antimicrobial susceptibility patterns (indicative of polyclonal carriage), one representative isolate for each pattern was included for WGS. WGS was done with Illumina technology (Illumina, San-Diego, USA), and data were analysed on the publicly available platform at the Center for Genomic Epidemiology [
12]. Default threshold settings were used for MLST and ResFinder [
13,
14].
Statistical analysis
We expressed the prevalence of ESCr-EK upon admission as the number of patients positive for ESCr-EK per 100 patients enrolled in the study. For the descriptive analysis, and to evaluate independent factors associated with colonisation we performed uni- and multivariable binominal regression analyses. We grouped countries into five regions (Europe, Asia, Americas, Africa and Australia). The category “Asia” included countries in the Middle-East as well as Central and East Asia (listed in Additional file
1). In the multivariable analyses variables with a
p-value of < 0.2 were included in the model one by one starting with the factor with the highest prevalence ratio. Model fit was evaluated using the likelihood-ratio test.
P-values < 0.05 were considered significant. Health care outside of the Nordic countries, included in the national screening guidelines, was included in the final model irrespective of level of significance.
Human subject protection
Informed consent was obtained from each patient or parents of patients below 16 years of age. The Regional Committees for Medical and Health Research Ethics (REC 2012/2234) and the hospital’s privacy protection officer approved the study protocol. The study was funded by the NIPH.
Discussion
In this study, including 747 patients from two emergency departments and three surgical units, we found no E.coli or Klebsiella spp. resistant to carbapenems, and no vancomycin resistant enterococci. The prevalence of colonisation of ESCr-EK upon admission was 6.0%. A major predictor for colonisation was travel to Asia. However, the majority of colonised patients had not been to Asia 12 months prior to admission.
The observed prevalence of ESCr-EK (6.0%) and ESBL-EK colonisation (5.6%) at admission was in the same range but lower than reported in similar studies from Germany (7.5–9.5%) [
15], the Netherlands (8.2%) [
16], and Israel (10.8%) [
17] and slightly higher than previously reported in healthy individuals in Sweden (4.7%) [
18] and Norway (4.9%) [
19]. Varying methods and definitions challenge direct comparison between studies and this may, at least in part, explain these differences between countries. Other explanations may be differences in age, ethnicity, co-morbidity, previous hospital admissions or a true lower prevalence upon admission in the Norwegian population compared to the prevalence in the Netherlands, Germany or Israel. To our knowledge, there has been no previous Norwegian studies of ESBL prevalence in patients at admittance to hospital.
Self-reported use of antibiotics has been found to be a risk factor for ESC-E colonisation at admission in Germany [
15]. We could not confirm this finding, nor could studies from the Netherlands [
16] and Sweden [
18]. Differing antibiotic-prescription policies may be one explanation for this. It was not possible to perform stratified analyses on antibiotics class as one third of participants who had used antibiotics 12 months prior to the admission did not indicate which class.
Consistent with other studies, visiting Asia within 12 months prior to faecal sampling, was associated with ESCr-EK colonisation (PR 6.8,
p < 0.001). Prevalence ranging from 13 to 45% has been found in Swedish [
18,
20] and Danish [
21] travellers to Asia. In accordance with our study, none of these studies found carbapenem-resistant
Enterobacteriaceae. However, the methods we used were not optimised to detect carbapenemases like OXA-48, which do not also confer resistance to cefotaxime or ceftazidime. The higher prevalence in patients who had travelled to Asia is also consistent with results from prevalence studies from Asia (Pakistan [
22], Thailand [
23‐
25], Turkey [
26]) on different populations and types of specimens, where a prevalence of up to 40–77% has been reported even in healthy adults and food products.
This study covers a convenience sample of a population utilizing the hospital. We recruited patients from two emergency departments and three elective surgery units, and a wide range of ages from infancy to elderly patients. It was not practically feasible to recruit critically ill patients admitted to intensive care units in the hospital. Hence, we do not describe the prevalence of ESCr-EK in units with patients at the higher risk for a severe outcome if colonised with resistant bacteria. The prevalence tended to be higher amongst patients who reported having received health care outside of the Nordic countries. However, there were not enough patients in this category to give sufficient power to quantify a difference. The data collected was not detailed to a level allowing for stratification on type of health care (hospital admission, dental care, primary health care, etc.).
The genetic mechanisms behind ESC resistance was mainly ESBL production due to
blaCTX-M. Only two ESCr-EK could not be explained by ESBL or pAmpC production. We found a large diversity of MLSTs in the 45 patients carrying ESCr-EK, making comparison with previous findings challenging. Sixteen different MLSTs were found in patients who had travelled to Asia, nine of these were found in travellers who had only been in Asia. However, many of these MLSTs represent well-known clones that have been found in various sources in several countries, including in Norway [
7]. The dominant ESBL allele was
blaCTX-M-15 which has emerged and disseminated worldwide [
27] and has previously been found in travellers returning from all over the world [
20], and in environmental samples from our area [
7]. The ESBL allele
blaCTX-M-55 has been found to be one of the most commonly encountered ESBL-encoding genes in
Enterobacteriaceae from food animals in China and has also been found in food animals in Europe together with other CTX-M types [
28,
29] suggesting the potential importance of animal-to-human spread via contaminated food [
20].
In our study, travel to Asia was primarily to Thailand and Pakistan. We expect that travel to other high endemic areas in Asia like India, or to Africa as found in other studies [
18,
20], might also be a risk factor for colonisation as also indicated in our study. All the ESCr-EK isolates from patients who had travelled to Asia were multidrug resistant.
Urinary tract infections and bacteraemia are often preceded by colonisation of the gut [
17,
30]. As more than one third of patients who had travelled to Asia within the last 12 months carried ESCr-EK, we would recommend that these patients receive empiric antimicrobial treatment effective against ESCr-EK if they present with signs of serious infections of possible Gram-negative origin. According to findings in several studies, and repeated here, ESCr-EK are often multidrug resistant, and carbapenems is probably the safest choice for empirical treatment [
31]. Treatment should always be adjusted and, if possible, de-escalated as soon as culture results are available [
32,
33].
Certain ESCr-EK phylogroups or subclones have been linked to increased potential to cause severe infections because of higher virulence [
18]. More knowledge of the more virulent phylogroups or subclones and risk factors for colonisation would allow targeted screening/treatment guidelines. As of today, the national guidelines for ESCr-EK infection control and prevention do not differentiate between MLSTs.