Import of multidrug-resistant bacteria from abroad through interhospital transfers, Finland, 2010–2019

Background While 20–80% of regular visitors to (sub)tropical regions become colonised by extended-spectrum β-lactamase-producing Enterobacteriaceae (ESBL-PE), those hospitalised abroad often also carry other multidrug-resistant (MDR) bacteria on return; the rates are presumed to be highest for interhospital transfers. Aim This observational study assessed MDR bacterial colonisation among patients transferred directly from hospitals abroad to Helsinki University Hospital. We investigated predisposing factors, clinical infections and associated fatalities. Methods Data were derived from screening and from diagnostic samples collected between 2010 and 2019. Risk factors of colonisation were identified by multivariable analysis. Microbiologically verified symptomatic infections and infection-related mortality were recorded during post-transfer hospitalisation. Results Colonisation rates proved highest for transfers from Asia (69/96; 71.9%) and lowest for those within Europe (99/524; 18.9%). Of all 698 patients, 208 (29.8%) were colonised; among those, 163 (78.4%) carried ESBL-PE, 28 (13.5%) MDR Acinetobacter species, 25 (12.0%) meticillin-resistant Staphylococcus aureus, 25 (12.0%) vancomycin-resistant Enterococcus, 14 (6.7%) carbapenemase-producing Enterobacteriaceae, and 12 (5.8%) MDR Pseudomonas aeruginosa; 46 strains tested carbapenemase gene-positive. In multivariable analysis, geographical region, intensive care unit (ICU) treatment and antibiotic use abroad proved to be risk factors for colonisation. Clinical MDR infections, two of them fatal (1.0%), were recorded for 22 of 208 (10.6%) MDR carriers. Conclusions Colonisation by MDR bacteria was common among patients transferred from foreign hospitals. Region of hospitalisation, ICU treatment and antibiotic use were identified as predisposing factors. Within 30 days after transfer, MDR colonisation manifested as clinical infection in more than 10% of the carriers.


Introduction
The spread of antimicrobial resistance (AMR) is strongly associated with international travel: 20-80% of visitors to high-risk regions become colonised and carry multidrug-resistant (MDR) bacteria back to their home country [1]. In high-income countries, rising background resistance and, particularly, import of MDR bacteria into hospitals from overseas is a concern.
Compared with infections by bacteria susceptible to antibiotics, infections by resistant bacteria are associated with greater mortality, longer hospitalisation and higher costs [2,3]. Colonisation by MDR bacteria, such as extended-spectrum β-lactamaseproducing Enterobacteriaceae (ESBL-PE), carbapenemase-producing Enterobacteriaceae (CPE), meticillin-resistant Staphylococcus aureus (MRSA), MDR Acinetobacter species (MDRACI), MDR Pseudomonas aeruginosa (MDRPA) and vancomycin-resistant Enterococcus (VRE), often remains asymptomatic but increases the risk of developing an infection. Colonised individuals may spread bacteria to contacts and the broader environment. Recently, we showed that approximately half of all ESBL-PE imported by travellers carry either intestinal or extraintestinal/ uropathogenic virulence genes [4].
Research into types and rates of travel-acquired MDR bacteria and the associated risk factors aids prioritisation of infection control resources and selection of empirical antibiotics. Interhospital patient transfers from abroad involve an increased risk of MDR bacterial carriage [13]. Such transfers may pose a substantial threat especially to hospitals in countries with a lower AMR prevalence, yet data on actual rates and risk have thus far been scarce. This study was undertaken to provide region-specific rates intended to provide basis for infection control management when devising guidelines and targeting resources.

Study design
The prevalence and risk factors of colonisation by MDR bacteria were studied among patients transferred from hospitals abroad to Helsinki University Hospital, Finland (HUH) between 1 January 2010 and 30 June 2019. We searched the HUH electronic infection control database for patients screened for both MRSA and MDR Gram-negative bacteria (MDRGNB). The latter screen comprises detection of CPE, ESBL-PE, MDRACI and MDRPA. For such patients, electronic medical records were explored and those with interhospital transfer data included. Patients with records from 2010 to 2013 were covered in a previous AMR investigation [13] that did not report separately on direct hospital transfers.

Definitions
Hospitalisation abroad was defined as a hospital stay of more than 24 h or an admission involving surgery or some other major invasive procedure. Direct transfer was defined as hospitalisation that continued immediately on return to Finland, with no overnight stay outside hospitals (excluding flights).

Inclusion criteria
The microbiological inclusion criteria comprised (i) a record of rectal swab or stool sample for MDRGNB screening within 3 days and (ii) MRSA screening from all three sites (nose, throat/trachea, groin/perineum)

Exclusion criteria
Medical records were searched for other documented stays and hospitalisations abroad. Stays outside Europe in regions other than those where hospitalised during the previous 12 months led to exclusion. Since travel within Europe is common and often not recorded in patient files, this was not taken as an exclusion criterion. Patients with multiple foreign hospitalisations were included only if all the hospitalisations took place in the same European country or, for other parts of the world, within the same geographical region.

Data collection
We collected data on underlying diseases, chronic alcohol abuse, travel-related factors and information linked to hospitalisation (antibiotics, interventions, intensive care unit (ICU) treatment, duration, diagnosis), and calculated the Charlson comorbidity index (CCI) [14]. In addition to documented antibiotic use, records of bacterial infections known to require antibiotics or surgery for which prophylaxis is advised were classified as 'antibiotic use abroad'. Ongoing antibiotic treatments at the time of screening were recorded but marked as negative if at least one set of samples had been taken without any antibiotics administered for 24 h. Because of the complexity of potential antimicrobial effects, we chose not to consider whether or not The map was created with mapchart.net.
an antibiotic acted against the MDR bacteria carried by the patient.
Symptomatic, microbiologically verified MDR infections and direct mortality caused by such infections were assessed by an infectious diseases specialist; they were recorded from transfer until discharge or for a maximum of 30 days. Because of incomplete data, potential MDR bacterial infections treated abroad were not evaluated.

Microbiological methods
The various MDR bacteria were identified by the standard methods used in the HUH laboratory HUSLAB, as follows.
VRE were detected using enrichment Enterococcosel broth (BBL, Cockeysville, MD) followed by culture on in-house selective media, as previously described [15], or CHROMagar VRE media. Positive findings were confirmed by in-house PCR [15].
ESBL-PE and CPE were analysed by plating directly on CHROMagar ESBL and CHROMagar Klebsiella pneumoniae Carbapenemase (KPC) or CHROMagar mSuper-CARBA, respectively. Identification of ESBL-PE species was confirmed by matrix-assisted laser desorption ionisation time-of-flight (MALDI-TOF; Vitek-MS, bioMérieux) and resistance was determined according to the guidelines from the Clinical and Laboratory Standards Institute (CLSI) and, from 2011, the European Committee on Antimicrobial Susceptibility testing (EUCAST) [15][16][17]. CPE were confirmed with in-house PCR targeting the carbapenemase gene [15].
MDRACI and MDRPA were screened on ESBL and KPC plates. Cultures were tested by C-390, VITEK-GN or MALDI-TOF for species identification. Acinetobacter isolates resistant to meropenem and Pseudomonas isolates resistant to both meropenem and ceftazidime were analysed by PCR for carbapenemase genes [15].

Statistical analyses
We used SPSS v. 25.0 (IBM Corp., Armonk, New York, United States) for all statistical analyses. For univariate analyses the chi-squared test, Fisher's exact test or binary logistic regression were used, as appropriate. For multivariable analysis, variables with a p value below 0.2 in univariate analysis, or those assessed as clinically relevant, were included. In cases of two strongly correlating explanatory variables, only one was chosen. The most parsimonious model was found by backward selection based on Akaike information criteria.

Ethical statement
The present study was approved by the research board of HUH Department of Internal Medicine. Since this investigation did not involve an intervention, an ethics committee review was not required (Finnish Medical Research Act).

Study population
A total of 698 patients undergoing direct hospital transfers between 1 January 2010 and 30 June 2019 met the inclusion criteria ( Figure 1

Antibiotic use abroad and during screening
The medical records included documentation of antibiotic use abroad for 383 patients (54.9%). After transfer, 458 (65.6%) patients had three-site MRSA and faecal MDRGNB screening at least once without ongoing antibiotic treatment, while 182 (26.1%) had ongoing antibiotic therapy at screenings. For 58 patients, data were missing or antibiotics were used during part of the screenings.

Findings of multidrug-resistant bacteria
A total of 208 patients (29.8%) were colonised by MDR bacteria, 41 by more than one class of MDR bacteria ( Figure 1). ESBL-PE were the most common findings with 163 ( Figure 2 shows the colonisation rates by geographical regions.

Risk factors for colonisation with multidrugresistant bacteria
In univariate analysis, several factors were associated with MDR carriage (Table 1) and in multivariable analysis, geographical region, ICU treatment and antibiotic use abroad were identified as independent risk factors. Separate analyses of individual MDR types are presented below.
Univariate analysis indicated that country of hospitalisation, antibiotic use abroad, duration of hospitalisation and surgery were risk factors associated with ESBL-PE carriage (Table 2), whereas multivariable analysis yielded only geographical region as risk factor.
Results of the univariate analyses conducted for MDRACI and MRSA are presented in Tables 3 and 4. Carriage of each of these two was associated with antibiotic use abroad, while only MDRACI carriage was associated also with antibiotics during screening. In addition, both were associated with ICU treatment abroad. Colonisation with MDRACI was more common among those hospitalised in Asia (14/96; 14.6%) than Europe (10/524; 1.9%). Carriage of MRSA was associated with chronic alcohol abuse, a finding independent of antibiotic use in a model with two explanatory variables. Significant associations were also recorded between CCI and MDRACI (Table 3) and between duration of hospitalisation and MRSA (Table 4).  Tables 3 and 4, no significant association was found (data not shown).
Risk factor analyses were not carried out for CPE or MDRPA due to the small number of colonised individuals (14 with CPE and 12 with MDRPA).

Clinical infections caused by multidrugresistant bacteria
During post-transfer hospitalisation, 22 of 698 patients (3.2% of the whole study population and 10.6% of those identified as MDR bacteria carriers) had a microbiologically verified clinical MDR bacterial infection, most commonly pneumonia which was found in nine patients (1.3% of all), surgical site infection (six patients, 0.9%), and urinary tract infection (six patients, 0.9%). One patient had MDR bacteraemia. For sites of infection and causative MDR bacteria, see Supplement (Supplementary Table S4). Infection by MDR bacteria was recorded as cause of death for two patients (0.3%).

Discussion
Of the 698 patients transferred to a Finnish hospital directly from hospitals abroad, 29.8% were colonised by MDR bacteria, the rates varying considerably by geographical region visited. From these 208 patients, 383 MDR bacterial strains were recorded. While these figures indicate the burden of MDR bacteria related to interhospital transfer, closer scrutiny reveals background data applicable to infection control practices and even choice of empiric antibiotics.
At first glance, the 29.8% overall MDR carriage rate appears low. As the rates typically decrease after patients return to low-prevalence countries [18][19][20][21][22][23], one could expect colonisation to be particularly common in these patients who were screened soon after return to Finland. By contrast, the prevalence proved to be similar to that of our previous data from 2010 to 2013 showing 29.7% carriage rates for patients screened within 12 months after hospitalisation abroad [13]. However, the closeness of the rates may be ascribed to at least two factors. Firstly, up to 23% of the patients in that previous investigation were, in fact, direct-transfer patients, and the time taken from return from abroad to sampling for the rest of the patients was short (median: 11 days). Secondly, the proportion of patients hospitalised in Europe where acquisition is less common [13,15] was higher in the current (75%) than in the previous dataset (64%).
Other European studies have reported MDR bacterial colonisation in similar patient groups but with differences in research design [5][6][7][8][9]11,12]. Among 1,167 patients directly transferred from hospitals abroad to the Netherlands between 1998 and 2001, Kaiser et al. show a colonisation rate of 18.2% [6], but their designation of resistant Gram-negative bacteria was solely based on gentamicin resistance [6] and there has been a substantial general increase in AMR rates since that study. In more recent research among patients screened on direct transfer or within 14 days of hospitalisation abroad, between 7.2 and 28.6% were colonised [7][8][9]11,12]. For patients hospitalised abroad and examined within 14 days in Switzerland between 2009 and 2011, Nemeth et al. showed the rate to be 17%, although VRE was not included in this study [7].  In another Swiss study where outpatient treatment abroad was also included but without screening for VRE, Kaspar et al. report a 16.3% carriage rate for direct transfer patients in 2012 to 2013 [11]. Josseaume et al. (2010Josseaume et al. ( -2011 and Birgand et al. (2012Birgand et al. ( -2013 show MDR bacterial colonisation for 7.2% and 28.6% of repatriated patients in France, respectively [8,9]. In 2012 to 2013, Mutters et al. detected a colonisation rate of 21.0% among patients hospitalised abroad at least 48 h and subsequently transferred to a German hospital [12]. The present study identified region visited, ICU treatment, and antibiotic use during travel as independent risk factors of MDR bacterial colonisation. All three factors have previously been associated with MDR acquisition [13,[24][25][26][27]. In our study population, association with geographical region proved particularly strong, for example, the colonisation rate was 18.9% among patients transferred from European countries and 71.9% among those returned from Asia (p < 0.001; OR = 10.5; 95% CI: 6.3-17.3). This difference was mainly ascribed to Gram-negative MDR bacteria rather than MRSA or VRE.
The overall colonisation rate by ESBL-PE was 23.4%, accounting for 68.1% of all MDR strains identified. This rate is substantially higher than among the general Finnish population: in 2009 to 2010, 1.2% of 430 Finnish travellers were colonised with faecal ESBL-PE before travel [27]; in 2015 to 2017, a study among Finnish elective surgery patients and medical students reported that 4.7% were colonised with ESBL-producing E. coli, and 1.1% with ESBL-producing K. pneumoniae [28].
Several prospective investigations have shown that visitors to (sub)tropical regions acquire ESBL-PE, with antibiotic use predisposing to colonisation [18,27,29]. In the present study looking at hospitalised travellers, multivariable analysis identified region of hospitalisation as the sole factor independently associated with ESBL-PE colonisation. ESBL-PE acquisition is so common among travellers visiting high-risk regions that any additional impact of nosocomial transmission may remain modest. While antibiotic use was found to predispose to MDR colonisation as a whole, its association with contracting ESBL-PE was significant in univariate but not in multivariable analysis, possibly because of an insufficient number of observations. The colonisation rates for other MDR bacteria were low, but not without relevance in a low-prevalence country like Finland [30]. In the stool specimens of 33 patients within this study, 46 carbapenemase-producing strains were recorded: 20 CPE and 26 MDRACI or MDRPA. In comparison, between 2010 and 2018, only 136 CPE strains were reported for the entire Helsinki and Uusimaa hospital district serving a population of 1.7 million, and the rates of MDRACI and MDRPA have been very low [30,31]. Thus, direct hospital transfers evidently contribute considerably to these cases. The MRSA colonisation rate of 3.6% is in line with that of 1.2-4.1% observed in earlier studies [5][6][7][8]12]. Although the rates for MRSA (3.6%) and VRE (3.8%) may appear low, they exceed those typical for Finland in general [30].
In univariate analyses conducted separately for MDRACI, MRSA and VRE, we found several associations. Antibiotic use abroad and ICU treatment were associated with each of the three. Indeed, both antibiotic use and intensive care predispose to MDR acquisition [24][25][26]. The association observed between MRSA carriage and chronic alcohol abuse confirms results of a previous study [32].
Clinical, microbiologically verified MDR bacterial infections were identified in 10.6% of the colonised patients in our study within 30 days after transfer, consistent with the rate of 11.4% observed in 2010 to 2013 as reported by Khawaja et al. [13] The rate of infection observed in the study by Mutters et al. was significantly higher at 29.9%, however, the demographics of patients in that study were different: over half of the patients were transferred to Germany from their country of origin in the Middle East [12]. As for nonhospitalised healthy travellers, we recently showed a maximum clinical infection rate of 17% for ESBL-PE carriers, with travellers' diarrhoea (TD) as the most common manifestation, whereas the estimated maximal rate of infections other than TD was 3% [4]. Indeed, the data suggested higher rates of clinical MDR infections (TD excluded) among travellers who were hospitalised compared with non-hospitalised travellers [4].

Limitations of the study
This investigation had limitations typical of a retrospective study design, such as incomplete data in some patient records and missing information on pretravel colonisation. Comparisons can nevertheless be made with background colonisation rates, as mentioned above. Due to lack of non-hospitalised controls, it is difficult to determine the respective proportions of nosocomial and community-acquired infection during travel. In numerous reports on ESBL-PE colonisation after travel without hospitalisation [1] the rates resemble our data, but community acquisition of other MDR bacteria appears limited [27,29,[33][34][35]. Furthermore, since the time frame for recording MDR bacteria was up to 1 month, nosocomial transmission after return to Finland cannot be ruled out. However, the risk was considered marginal on account of the low background prevalence: only 2% of S. aureus isolates are MRSA strains [30], and the background colonisation rate of ESBL-PE in Finland remains under 5% [28].
As culture-based assays lack sensitivity, a greater number of different MDR bacterial strains could be expected when employing modern genome-based methods, as shown in our recent study on travellers [36]. However, culture-based approaches are used in clinical practice and allow comparisons with earlier studies. For some patients, MDR bacteria from non-screened anatomic sites, such as skin lesions, could have been missed. However, we believe that the strict inclusion criteria for triple-site MRSA and faecal MDRGNB screenings together with clear hospital guidelines have resulted in a realistic yield.
With regards to antibiotic treatments, records from hospitals abroad were often not available, and antibiotics in use at the time of screening may have affected MDR bacterial findings. In general, depending on the setting, a concomitant antimicrobial effect may lead to unsuccessful bacterial culture or, contrarily, a selection of resistant strains. The complex effects of various antibiotics and their combinations could not be analysed. Finally, the rate of symptomatic MDR bacterial infections may be an underestimate, since infections without microbiological verification were not covered. As infections treated abroad and those diagnosed after discharge (or 30 days) were not recorded, a different design would be needed to evaluate the total burden of MDR bacterial infections among this patient population.

Conclusions
Colonisation by MDR bacteria is common among patients transferred from hospitals in high-prevalence countries. The most prevalent bacteria, ESBL-PE, are also frequently carried by non-hospitalised travellers. In addition, a substantial number of non-ESBL-PE strains, such as carbapenemase-producing bacteria, was detected. Among the variety of risk factors of MDR bacterial colonisation that were identified, geographical region of hospitalisation proved the strongest predictor of MDR findings. The study indicates that systematic screening of international transfer patients is warranted; our data serve as valuable background for devising infection control policies.