Antimicrobial resistance and antibiotic consumption in intensive care units, Switzerland, 2009 to 2018

Background Intensive care units (ICU) constitute a high-risk setting for antimicrobial resistance (AMR). Aim We aimed to describe secular AMR trends including meticillin-resistant Staphylococcus aureus (MRSA), glycopeptide-resistant enterococci (GRE), extended-spectrum cephalosporin-resistant Escherichia coli (ESCR-EC) and Klebsiella pneumoniae (ESCR-KP), carbapenem-resistant Enterobacterales (CRE) and Pseudomonas aeruginosa (CRPA) from Swiss ICU. We assessed time trends of antibiotic consumption and identified factors associated with CRE and CRPA. Methods We analysed patient isolate and antibiotic consumption data of Swiss ICU sent to the Swiss Centre for Antibiotic Resistance (2009–2018). Time trends were assessed using linear logistic regression; a mixed-effects logistic regression was used to identify factors associated with CRE and CRPA. Results Among 52 ICU, MRSA decreased from 14% to 6% (p = 0.005; n = 6,465); GRE increased from 1% to 3% (p = 0.011; n = 4,776). ESCR-EC and ESCR-KP increased from 7% to 15% (p < 0.001, n = 10,648) and 5% to 11% (p = 0.002; n = 4,052), respectively. CRE, mostly Enterobacter spp., increased from 1% to 5% (p = 0.008; n = 17,987); CRPA remained stable at 27% (p = 0.759; n = 4,185). Antibiotic consumption in 58 ICU increased from 2009 to 2013 (82.5 to 97.4 defined daily doses (DDD)/100 bed-days) and declined until 2018 (78.3 DDD/100 bed-days). Total institutional antibiotic consumption was associated with detection of CRE in multivariable analysis (odds ratio per DDD: 1.01; 95% confidence interval: 1.0–1.02; p = 0.004). Discussion In Swiss ICU, antibiotic-resistant Enterobacterales have been steadily increasing over the last decade. The emergence of CRE, associated with institutional antibiotic consumption, is of particular concern and calls for reinforced surveillance and antibiotic stewardship in this setting.


Introduction
Intensive care units (ICU) are considered high-risk settings for antimicrobial resistance (AMR) [1]. This can be explained by an increased risk for nosocomial infections in ICU patients who have underlying comorbidities and frequently use of medical devices, but also by potentially increased exposure to resistant organisms where there is high patient turnover and by the large number of transmission opportunities [2]. Another particularity of the ICU setting is the extensive use of broad-spectrum antibiotics, which is considered to be one of the main drivers for the development and dissemination of AMR, in particular for resistant Gramnegative pathogens [3]. Indeed, several studies have reported an association between carbapenem consumption and the emergence of carbapenem-resistant Enterobacterales (CRE) and Pseudomonas aeruginosa (CRPA) [4,5]. The CRE and CRPA are pathogens of particular interest, as both have been classified by the World Health Organisation (WHO) as critical pathogens regarding the need for new antibiotic substances [6].
Surveillance is considered a cornerstone in the fight against AMR [7]. Specific surveillance data from AMR in Swiss ICU have not been published so far, although outbreaks caused by resistant pathogens have been reported [8]. Data from the Swiss Centre for Antibiotic Resistance (ANRESIS) for non-ICU settings have shown a steady rise of AMR among Gram-negative pathogens between 2007 and 2017, while meticillin-resistant Staphylococcus aureus (MRSA) are declining [9,10]. Antibiotic consumption in Swiss ICU has been shown to be somewhat lower than in other European countries; however, these data were collected between 2004 and 2008 [11].
In this study, we describe secular trends of AMR and antibiotic consumption from Swiss ICU using surveillance data from ANRESIS. For CRE and CRPA, we sought to identify patient-level and institutional factors associated with resistance, including institutional antibiotic consumption. We also assessed the representativeness of ANRESIS for the ICU setting in Switzerland.

Data source
This laboratory-based surveillance study was conducted using antimicrobial resistance and antibiotic consumption data from ANRESIS (www.anresis.ch). ANRESIS, led by the Institute for Infectious Diseases at the University of Bern, is the national centre for surveillance of antibiotic resistance and antibiotic consumption and is supported by the Swiss Federal Office of Public Health. Antimicrobial susceptibility testing is performed at local laboratories according to guidelines from the Clinical and Laboratory Standards Institute (CLSI) or the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [12,13]. All laboratories are accredited by Swiss authorities and participate in at least one external quality programme of either the National External Quality Assessment Service (https://ukneqas.org.uk) or the Swiss quality control programme of the Institute for Medical Microbiology, University of Zurich (www.imm.uzh.ch/de/services/ qc.html).

Representativeness of ANRESIS
ANRESIS continuously collects routine antibiotic resistance data (along with basic patient information including age category, sex and sample location) from 33 laboratories distributed homogeneously all over Switzerland and representing 89% of all hospital samples in 2020. Antibiotic consumption data are collected once yearly from around 70 hospital pharmacies across Switzerland. For the analysis of AMR in ICU, we assessed the representativeness of ANRESIS for ICU registered by the Swiss Society for Intensive Care (www.sgi-ssmi.ch/de/zertifizierte-is.html).
Because several small-size ICU have been closed or have been incorporated into larger institutions over the last decade, representativeness was only assessed for the year 2018. This was done by calculating the percentage of ICU beds represented in ANRESIS among all registered ICU beds in Switzerland, stratified by administrative subdivisions (i.e. cantons) speaking mostly French or Italian (cantons of Fribourg, Geneva, Jura, Neuchâtel, Ticino, Valais and Vaud; henceforth referred to as south-west) or German (all other cantons; henceforth referred to as east). In order to detect differences in microbiological sampling frequencies between geographical regions, we calculated the sampling density, defined as the number of microbiological samples per represented ICU bed.

Inclusion criteria for antimicrobial resistance and antibiotic consumption data
For the analysis of AMR, we extracted microbiological isolates with the label 'intensive care' sent to ANRESIS between January 2009 and December 2018 from patients 15 years and older. Although data were anonymous, every patient possesses a unique identity number within every institution, which was used to exclude repeat isolates (i.e. same pathogen and resistance profile) from the same patient in a specific calendar year. If patients contributed both resistant and susceptible isolates (per species) in the same year, only the resistant isolate was included. Therefore, each patient contributed only one sample per species, institution and year. Bacterial isolates from non-specified swabs (presumably rectal swabs taken at time of admission) were excluded in order to minimise the impact of patients colonised with resistant pathogens before ICU admission.
For the analysis of antibiotic consumption, we extracted consumption data for the same time period (2009 to 2018) from those ICU where these data were available. Of note, antibiotic consumption data were only available on the institutional level (per year) and not on the patient level.
For the analysis of risk factors for CRE and CRPA, we included only data from years where institutions provided AMR data along with corresponding antibiotic consumption data.

Definitions for antimicrobial resistance
Resistance to a particular substance was defined as either an intermediate or a resistant phenotype in susceptibility testing according to the individual laboratories. Group resistance (e.g. glycopeptide, carbapenem, aminoglycoside or quinolone resistance) was defined as resistance or intermediate susceptibility to at least one antibiotic of the respective group; extended-spectrum cephalosporin-resistance was defined as resistance or intermediate susceptibility to at least one third or fourth generation cephalosporin.

Data analysis and statistics
We expressed AMR as a percentage of all isolates and consumption of systemic antibiotics as defined daily doses (DDD) per 100 bed-days (BD), compliant with the World Health Organization (WHO) definition from 2019 [14]. Categorical variables were reported as frequencies and proportions, continuous variables as median with interquartile range (IQR). For dichotomous variables, we used chi-squared or Fisher's exact test, as appropriate. For continuous variables, we used the Wilcoxon rank-sum test. For each type of resistance, the proportion of pathogens with the resistance trait (as percentage of the total number of isolates reported and tested for the respective key antibiotic) was analysed over time. Linear regression analysis was used to estimate temporal trends, using resistance or antibiotic consumption as dependent and year of surveillance as independent variable.
For the analysis of risk factors for CRE and CRPA, we performed a mixed model with ICU as random effect, considering the hierarchical data structure. Patientlevel variables included age group (< 60, 60-75 and > 75 years), sex and site of sample collection (blood, respiratory, urogenital or other). Institutional variables included geographical region (south-west vs east), hospital type (university vs non-university hospitals) and total annual antibiotic consumption. All co-variables were entered into multivariable analysis. Two-sided p values ≤ 0.05 were considered statistically significant. All statistical analyses were performed using R statistical software version 3.5.2 (R Foundation, Vienna, Austria).

Sensitivity analyses
Sensitivity analyses were performed for the assessment of temporal trends for AMR and antibiotic consumption, including only those ICU that provided data for at least 8 years between 2009 and 2018, in order to eliminate potential selection bias.

Ethical statement
Because ANRESIS contains routinely collected anonymised surveillance data, ethical consent was not required according to the Swiss law for research on human beings.

Representativeness of included intensive care units
Over the 10-year period, we included 52 and 58 ICU for the analysis of AMR and antibiotic consumption, respectively (for an overview see Supplementary Figure S1). By 2018, 11 of 52 ICU did no longer exist, resulting in 41 ICU represented in our database for the analysis of AMR. Of these 41 ICU (with a median bed number of nine), 17 were located in south-western, and 24 in eastern Switzerland; the median bed number was 39 (range: 34-64) for the five university and eight (range: 6-42) for the 36 non-university ICU (listed in Supplementary Table S1).
The 41 ICU covered 55% of the 74 registered ICU in Switzerland for the year 2018, or 70% (611/871) of all Swiss ICU beds. Coverage was higher for south-western (203/237; 86%) than for eastern Switzerland (408/634; 64%) (p < 0.001). For the year 2018, sampling density was 10.2 samples per represented ICU bed; sampling density was higher for ICU from the south-west (12.2 per represented bed) compared with those from the east (9.3 per represented bed) (p = 0.002) ( Figure 1).

Antibiotic consumption
Among 35 ICU from south-western and 23 from eastern Switzerland, penicillins, cephalosporins and carbapenems were the three antibiotic groups used the most. For 2018, they accounted for 68% (52.9 DDD/100 BD) of the total (78.3 DDD/100 BD) antibiotic consumption (for a detailed list see Supplementary Table S2). Temporal trends of these antibiotic groups, for quinolones and for the total antibiotic consumption are shown in Figure  3  (p = 0.044). Similar patterns were observed for the 'watch and reserve' group of antibiotics (WHO AWaRe classification) [15], as well as for carbapenems and quinolones. Consumption of piperacillin/tazobactam increased from 8.0 to 11.0 DDD per 100 BD (p = 0.003) from 2009 to 2018; ceftriaxone consumption increased from 6.3 to 7.9 DDD per 100 BD (p < 0.001). Notable geographical differences were a lower carbapenem (11.4 vs 15.6 DDD/100 BD; p = 0.002) and quinolone consumption (3.0 vs 4.0 DDD/100 BD; p = 0.014) in eastern compared with south-western Switzerland (Supplementary  Table S3).

Sensitivity analyses for temporal trends
Considering only those ICU that provided data for at least 8 years between 2009 and 2018, we observed the   same temporal trends regarding AMR (in total 28 ICU; 15 from the south-west, 13 from the east of Switzerland) and antibiotic consumption (in total 29 ICU; 14 from the south-west, 15 from the east of Switzerland) (data not shown).

Discussion
In this laboratory-based analysis of national surveillance data from Swiss ICU, we have shown that the occurrence of MRSA has been decreasing steadily, whereas resistant Gram-negative bacteria including ESCR-EC, ESCR-KP and -of particular concern -CRE have become more frequent between 2009 and 2018. The only modifiable factor associated with CRE was institutional antibiotic consumption. On a national level, antibiotic consumption peaked in 2013 and has since been declining. The strengths of this study were its large sample size, the representativeness of the database for the Swiss ICU setting and the robustness of temporal trends in sensitivity analyses.   Our main finding was the emergence of resistance among Enterobacterales in Swiss ICU over the last decade. These data are in line with trends described from other European countries, although the different observation periods overlap only partly with our analysis. In German ICU, ESCR-EC increased from 1.3% to 16.3% and imipenem-resistant K. pneumoniae from 0.4% to 1.6% between 2001 and 2015 [16]. In France, ESBL-producing Enterobacterales increased by 73% between 2009 and 2013, with higher rates in ICU [17]. In Italy, bloodstream infections caused by CPE, mainly originating in ICU, increased steadily between 2014 and 2017 [18]. Resistant Gram-negative pathogens are also increasing in Swiss non-ICU settings, with up to 22% of E. coli being ESCR-EC in long-term care institutions [19,20].
The emergence of CRE, associated with increased morbidity and mortality, is particularly worrying [21]. Enterobacter spp. were the predominant species among our CRE. Carbapenem-resistance in Enterobacter spp. can be conferred either by de-repressed AmpC in combination with porin loss/modification or by carbapenemase production [22,23]. De-repression of AmpC in Enterobacter spp. is usually observed after beta-lactam exposure, which is in accordance with our finding that institutional antibiotic consumption was associated -although not very strongly -with detection of CRE. However, we cannot exclude the possibility that the rise in carbapenem-resistant Enterobacter spp. may be due to dissemination of high-risk clones such as ST171 and ST78, which can also carry carbapenemase-encoding plasmids [23]. Another argument suggesting the emergence of CPE in Swiss ICU is the fact that antibiotic consumption, including carbapenem and quinolone consumption, is declining and that K. pneumoniae and E. coli are increasing among CRE. Increased importation of CPE from high-risk countries or even acquisition in local hospitals could account for their spread, as CPE are considered to spread primarily through horizontal transmission in the healthcare setting [8,24,25]. For instance, the hospital water environment (including drains and sinks) has increasingly been recognised as a reservoir for outbreaks of carbapenem-resistant organisms in the ICU setting [26]. Also, national surveillance data from Switzerland (inpatients and outpatients) show increasing numbers of CPE [27].
We found 27% of P. aeruginosa isolates to be carbapenem-resistant. This figure is comparable to data from 36 ICU in the United States which showed 35% CRPA, mostly from the respiratory tract [28]. The association between institutional antibiotic consumption and detection of CRPA was not significant in the multivariable analysis. Indeed, previous studies have shown that total antibiotic consumption is less predictive of CRPA than the consumption of broad-spectrum antibiotics such as carbapenems [5]. In a recent study from selected ICU in Serbia, conducted between 2014 and 2018, overall antibiotic consumption has been steadily decreasing, whereas the occurrence of CRAB and CRE increased [29]. On the other hand, in a before-after study from a French ICU, a reduction in quinolone and carbapenem use was associated with a decrease in CRPA, although total antibiotic use increased [30]. On the individual patient level, previous carbapenem exposure has also been identified as risk factor for CRPA [31].
A favourable trend was observed regarding MRSA in Swiss ICU. The fact that the decline in MRSA was exclusively attributable to haMRSA (and not to caMRSA) has been observed in other countries and has partly been attributed to increased awareness and infection control interventions, including the Dutch search-anddestroy approach [32]. Molecular typing is not routinely performed in Switzerland; this would allow to better discriminate haMRSA from caMRSA. In this context it is important to note that the concept of distinguishing haMRSA and caMRSA has been challenged in recent years, with traditional caMRSA such as USA300 increasingly detected in the healthcare setting and even showing a multi-resistant phenotype [33]. In contrast to MRSA, GRE have become more frequent, particularly in ICU from eastern Switzerland. This signal most probably reflects an interregional VRE outbreak caused by ST796 detected in 2017 [34].
Geographical differences have to be taken into account when explaining AMR trends in Switzerland. Historically, the south-western part of the country has been more affected by AMR than the east [10]. However, our data suggest that these differences are more and more levelling out. Not only were MRSA rates in 2018 very similar between the two regions, but CRE and also GRE were more common in the east than in the south-west. In addition, ICU in the east provided fewer bacterial isolates per represented bed than in the south-west (i.e. lower sampling density), which presumably led to further underestimation of resistance rates in the east. Antibiotic consumption was also somewhat lower in the south-west compared with the east for the year 2018. The reasons for this paradigm shift remain elusive. One (speculative) argumentation is that these favourable trends are a result of enhanced infection control measures implemented during the MRSA epidemic in south-western Switzerland.
Antibiotic consumption in Switzerland has traditionally been lower compared with other European countries [11]. Our data show that antibiotic consumption increased in Swiss ICU from 2009 to 2013. Similarly, data collected from German ICU between 2001 and 2015 have shown a steady increase in antibiotic consumption along with the above-mentioned increase in resistant Gram-negative bacteria [16]. However, consumption in Swiss ICU, including use of antibiotic substances on the 'watch' group of the WHO's AWaRe classification, declined between 2013 and 2018. Likewise, in countries of the European Union and the European Economic Area, overall use of cephalosporins, quinolones and carbapenems has been stabilising or even declining in the same time period [35].
Despite the decreased antibiotic use in intensive care, many antimicrobial stewardship activities are still lacking in a majority of Swiss hospitals [36]. Given the importance of rational antibiotic consumption in the control of AMR, the Swiss National Center for Infection Control launched an initiative to promote antimicrobial stewardship programmes (ASP) in Swiss hospitals in 2016 [37]. The aim was, among others things, to formulate requirements for institutional ASP and facilitate surveillance of antibiotic consumption and antibiotic resistance data using the ANRESIS database.
In addition to ASP, particularly in light of the increasing numbers of resistant Gram-negative pathogens, we advocate for a multimodal infection control strategy to monitor and control the spread of AMR in ICU patients, including intensified admission screening programmes along with systematic molecular sequencing of pathogens, contact isolation of colonised or infected patients, strict adherence to standard precautions such as correct hand hygiene and use of personal protective equipment, and alertness to search and destroy environmental sources such as contaminated sink drainage systems, if necessary. We would like to emphasise that ASP (and also infection control measures) should not be reduced to human healthcare; a One Health approach should be pursued considering the complex interplay between human, animal and environmental health. Examples include the use of avoparcin as animal growth promoter in the 1990s, which may have contributed to the spread of glycopeptide-resistant genetic elements in humans [38], or the recent detection of carbapenemase-producing E. coli among employees of a Swiss veterinary clinic, which was probably due to cross-transmission from hospitalised companion animals [39].
An important limitation of this study was the lack of information on day of patient admission to the ICU. Although we excluded screening samples from the analysis, part of our data reflect antibiotic susceptibilities of samples collected at the time of ICU admission. This means that some resistant pathogens in our dataset were in fact community-acquired and were not a consequence of high institutional antibiotic consumption. We also suppose that restriction to bacterial samples collected 48 h after ICU admission would have resulted in even higher resistance rates. Furthermore, we could not distinguish between pathogens representing true infection and those only colonising. Also, because of the institution-specific patient identifier, we cannot exclude the possibility of duplicate isolates for patients being admitted to more than one ICU in the same year. Another limitation was the lack of information on individual antibiotic treatment before admission. This prevented us from including antibiotic consumption as a patient level variable in our hierarchical model.

Conclusion
Resistance among Enterobacterales in Swiss ICU is increasing. The rise in carbapenem resistance calls for a multimodal approach to improve surveillance (including systematic screening of ICU patients and molecular typing of resistant isolates) and adherence to standard infection control measures. Although antibiotic consumption is decreasing, the observed association between total antibiotic consumption and the occurrence of CRE underscores the need for antimicrobial stewardship in the Swiss intensive care setting.