National point prevalence survey on healthcare-associated infections in acute care hospitals, Switzerland, 2017

Background As a part of the national strategy on the prevention of healthcare-associated infections (HAI), a point prevalence survey (PPS) was conducted in acute care hospitals in Switzerland. Aim Our objective was to assess the burden of HAI in Swiss acute care hospitals. Methods All acute care hospitals were invited to participate in this cross-sectional survey during the second quarter of 2017. The protocol by the European Centre for Disease Prevention and Control was applied. Patients of all ages, hospitalised on the day of survey were included, except when admitted to outpatient clinics, emergency and psychiatry. Results Ninety-six acute care hospitals (79% of all hospitals ≥ 100 beds) provided data on 12,931 patients. Pooled and randomised HAI prevalences were 5.9% (95% confidence interval (CI): 5.5–6.3) and 5.4% (95% CI: 4.8–6.0), respectively. The HAI incidence was estimated at 4.5 (95% CI: 4.0–5.0). The most common type of HAI was surgical site infection (29.0%), followed by lower respiratory tract (18.2%), urinary tract (14.9%) and bloodstream (12.8%) infections. The highest prevalence was identified in intensive care (20.6%), in large hospitals > 650 beds (7.8%), among elderly patients (7.4%), male patients (7.2%) and patients with an ultimately (9.3%) or rapidly (10.6%) fatal McCabe score. Discussion This is the first national PPS of Switzerland allowing direct comparison with other European countries. The HAI prevalence was at European Union average (5.9% in 2016 and 2017), but higher than in some countries neighbouring Switzerland. Based on the limited information from previous surveys, HAI appear not to decrease.


Introduction
The occurrence of healthcare-associated infections (HAI) is a preventable patient safety concern [1,2]. An estimated 2.7 million new HAI cases occur every year in the European Union (EU) and the European Economic Area (EEA) together, contributing to a cumulative burden of 501 disability-adjusted life years per 100,000 general population [3]. Surveillance of HAI incidence with timely feedback has become standard in HAI prevention [4]. However, because of the high work effort, prospective HAI surveillance is mostly oriented towards high prevalence areas such as intensive care units or is confined to specific infection types such as surgical site infection (SSI) [5,6] or bloodstream infection (BSI) [7].
As an alternative, point-prevalence surveys (PPS) have been performed for many years to estimate the burden of HAI [8,9]. The "Study on the Efficacy of Nosocomial Infection Control" (SENIC) by the United States (US) Centers for Disease Control and Prevention (CDC), applied repeated PPS to assess the effectiveness of infection prevention and control (IPC) programmes in acute care hospitals in the US [10]. Although challenging by its cross-sectional design [11], PPS offer a valid perspective on the hospital-wide burden of HAI. In 2011 and 2012, the European Centre for Disease Prevention and Control (ECDC) conducted its first PPS in the EU Member States, Iceland, Norway and Croatia [12] and 5 years later, it conducted its second PPS in the EU and EEA countries as well as in some EU candidate countries [13,14]. Switzerland, however, did not participate in these surveys.
Switzerland had performed national period prevalence surveys in the past, for the last time in 2003 and 2004 [15,16]. In January 2013, the Swiss Federal Council passed its 'Health 2020' agenda [17], which sets priorities in healthcare management in Switzerland. The prevention of HAI was identified as a first order measure and therefore, the Federal Office of Public Health (FOPH), together with stakeholders from various areas of health delivery in Switzerland, defined a national strategy on HAI prevention [18]. The strategy, called 'Strategie NOSO', aims at reducing HAI and containing the emergence and spread of antimicrobial resistance in the various healthcare settings in Switzerland.
FOPH mandated Swissnoso to perform a national PPS (CH-PPS) on HAI in 2017 to estimate the burden of HAI in Swiss acute care hospitals. Swissnoso is a publicly funded association of experts in the field of IPC and infectious diseases in Switzerland ( www.swissnoso. ch ). The goal of this survey was to estimate the burden of HAI using the ECDC protocol and to compare the results with other European countries.

Setting and study population
In 2016, a pilot survey was performed in three large acute care hospitals in Switzerland with the objective to test the ECDC PPS protocol [19]. The CH-PPS protocol ( www.swissnoso.ch/prevalence ) was translated into

Data collection
The CH-PPS coordination centre organised seven interactive training workshops for hospital investigators and data collectors in three languages (German, French, and Italian) in March and April 2017. The courses followed a structured methodology, encouraging a participative, problem-solving approach by discussing clinical cases and using the database interactively. Patient data were collected on individual case report forms with demographic information and data on medical devices, antimicrobial use, HAI and microorganisms. Healthcare-associated infections were defined as outlined in the ECDC protocol version 5.3 [13]. Outcomes were stratified by hospital size (small: < 200 beds; medium-size: 200-650 beds; large: > 650 beds), hospital type (primary care, secondary care, tertiary care, specialised care), hospital ownership (public, private-not-for-profit, private-for-profit) and university-affiliation. The detailed methodology is described elsewhere [19].
Bars and whiskers indicate the HAI prevalences with 95% confidence intervals.
the electronic CH-PPS database or by direct data entry into the database. The format of the CH-PPS database featured plausibility algorithms to rule out missing data; it was provided by the Institute of Hygiene and Environmental Medicine, Charité University Medicine Berlin, Berlin, Germany [20]. Hospitals had the option to download their own data in different formats (HTML, CSV, pdf).

Data validation
To assess the sensitivity and specificity of the PPS, a validation survey was performed in six hospitals, applying the ECDC protocol on data validation [13]: one university-affiliated hospital, two public medium-size hospitals, two public small hospitals and one private hospital. Three investigators from the CH-PPS coordination centre performed validation in 50 patients of each hospital, prioritising high-prevalence areas.

Data analysis
Descriptive data are reported as medians with interquartile range (IQR) or means with 95% confidence intervals (CI), where appropriate. Statistical analysis of patient characteristics relative to hospital size was performed using the non-parametric Kruskal-Wallis test.
A fatal McCabe score [21], age groups (0-17, 18-40, 41-60, 61-80, > 80 years), admission in an intensive care unit (ICU) on the survey day, exposure to a medical device (peripheral venous catheter, central venous catheter, urinary catheter, endotracheal tube) on the survey day, having undergone National Healthcare Safety Network (NHSN) surgery since hospital admission [22] and private-for-profit ownership were tested in a univariable logistic regression analysis as risk factors for HAI. Variables with a p value ≤ 0.2 were tested in a multivariable model. Observations were clustered on the hospital level and a two-sided p value of 0.05 was considered significant. Data analysis was performed using STATA version 13 (STATA Corporation). The HAIincidence was estimated using the ECDC methodology Ger: geriatrics; gyob: gynaecology and obstetrics; icu: intensive care unit; med: medicine; mix: mixed ward (< 80% of one specialty); neonat: neonatology; other: other ward type; ped: paediatrics; rhb: rehabilitation; surg: surgery.
Bars and whiskers indicate the HAI prevalences with 95% confidence intervals.
based on the Rhame-Sudderth formula [12,23]. Median time from admission to survey data was used to estimate length of stay; median difference between length of stay and time from admission to first HAI was used to calculate the denominator [12]. Only the first HAI was taken into account.

Benchmarking
A representative sample was obtained from the pooled data to allow benchmarking with the ECDC PPS data. Applying the ECDC methodology (design effect of 4.7, calculated from the data of the invited hospitals; precision of +/− 1%; HAI-prevalence of 5.9; the measured pooled prevalence; 95% CI), the randomised sample needed data from 56 hospitals, representing the distribution of hospital size in Switzerland. Six randomised sequences were generated, stratified by hospital size, using research randomiser ( www.randomizer.org ). The final sequence was randomly selected from the six previous sequences. To benchmark antimicrobial resistance (AMR), a composite index was calculated using the definitions of the second ECDC PPS [14]: number of first level AMR isolates, divided by the sum of the isolates for which results from antimicrobial susceptibility testing were reported. These first level markers were defined as Staphylococcus aureus resistant to meticillin (MRSA), Enterococcus faecium and Enterococcus faecalisresistant to vancomycin, Enterobacteriaceae resistant to third generation cephalosporins, and Pseudomonas aeruginosa and Acinetobacter baumannii resistant to carbapenems.

Ethical statement
No institutional review board approval was deemed necessary, similar to the ECDC PPS, given the quality improvement character of the survey. Only anonymous patient and ward data were collected and analysed.        Table 3 summarises uni-and multivariable testing of risk factors for HAI. The HAI prevalence was significantly higher in large hospitals, tertiary care hospitals,  intensive care, patients with a fatal McCabe score, male patients and patients at older age. The lower HAI prevalence in private-for-profit hospitals (2.9%; 95% CI: 1.7-4.0) was not statistically significant compared with public hospitals and private-not-for-profit hospitals in the multivariable analysis.

Benchmarking to the second ECDC point prevalence survey
Given that Switzerland used the ECDC protocol and conducted the CH-PPS at the same time as the second ECDC PPS, data could be benchmarked to the EU and EEA countries. Figure 5 shows the position of Switzerland compared with the pooled PPS results of the second ECDC PPS [14].
The past HAI prevalence surveys in Switzerland used the period methodology and thus, comparison with historical data is difficult. However, it has been estimated that, compared with the point methodology, the period methodology inflates the HAI prevalence by 32% [8]. Thus, the reported period prevalence of 7.2% in the most recent national survey from 2004 would translate to a point prevalence around 5.5%. We may therefore hypothesise that the HAI prevalence did not change in Swiss acute care hospitals in the past years. The stable course of the yearly period prevalence surveys at the University of Geneva Hospitals between 2006 and 2013 further supports this hypothesis [8].
Switzerland established a national strategy aiming at HAI prevention only in 2016 [18]. At the same time, Swissnoso initiated the first HAI strategy aiming at SSI prevention [24]. This ongoing programme is the first national HAI initiative in Switzerland since the hand hygiene campaign in 2006 [25], which was part of the World Health Organization "Clean Care is Safer Care" campaign launched in 2005 [26,27].
With a proportion of 29%, SSI was the most common HAI. This is substantially higher compared with the ECDC PPS, where the proportion of SSI among HAI was 19.6% in 2011 and 2012 and 18.4% in 2016 and 2017 [12,14]. This difference can be explained in part by the fact that the proportion of both surgical patients and patients undergoing NHSN surgery was higher in the CH-PPS than in Europe in 2011 and 2012 (36.1% vs 24.8 and 30.6% vs 20.2%, respectively) [12]. The HAI prevalence among patients undergoing NHSN surgery was even lower in the Swiss PPS (8.8%) than in Europe in 2011 and 2012 (10.1%) [12]. The lower HAI prevalence in private-for-profit hospitals was due to more favourable risk factors such as lower age, more favourable McCabe scores and shorter length of stay before survey. Gram-positive cocci were the most common pathogens in the CH-PPS (40.2%), followed by enterobacteria   [14], ranking in the lower range similar to Austria (12.4%), the Netherlands (14.5%), and Luxemburg (14.9%). The proportion of carbapenem-resistant enterobacteria was much lower than in the second ECDC PPS (6.2%) [14]. While C. difficile contributed 12.1% in the most recent US CDC PPS, its role in Switzerland was much smaller (4.7%) [28], similar to the second ECDC PPS in 2016 and 2017 (4.9%) [14]. In the absence of a national surveillance system for the incidence of C. difficile infection in Switzerland, this finding must be interpreted with caution.
Our survey has limitations. Firstly, the prevalence is a point estimate, and deduction of the burden of HAI is not suitable, particularly because of the length bias caused by over-representation of patients with increased length of stay [29]. However, even if a point estimate is limited for an individual hospital, on a national basis, the results can be considered valid and the participation of 80% of Swiss acute care hospitals with 100 or more beds makes this study representative of the burden of HAI in Switzerland. Secondly, variation in the application of HAI definitions was likely despite multiple centralised workshops for hospital data collectors. This is evidenced by the results of the validation study, which resulted in an unexpected high adjusted HAI prevalence. Thirdly, significant differences in healthcare practices and lengths of stay in Europe limit comparison of HAI between Switzerland and the EU/EEA countries. Finally, compared with Switzerland as a whole, medium-size and large hospitals were over-represented in our database, and thus, the measured HAI prevalence may have been overestimated. This was addressed by analysing a randomised subsample, taking into account the size distribution of acute care hospitals in Switzerland.

Figure 5
Prevalence of patients with healthcare-associated infections in the Swiss and the ECDC point prevalence surveys combined Bars and whiskers: HAI prevalences and 95% CI by country. Vertical lines: mean prevalence and 95% CI across the countries participating in the 2016 and 2017 ECDC point prevalence survey. Source: [14].
From an organisational perspective, the survey was successful in achieving the desired objectives: engaging the majority of Swiss acute care hospitals in a national project, achieving good data representativeness, updating national data on HAI given that the last period-prevalence in Switzerland was performed in 2004 and, for the first time, generating data that are comparable with other countries in Europe.

Conclusions
This first national point-prevalence survey in Switzerland identified an HAI prevalence at European average. Compared with previous national surveys, which used a different methodology, it still can be estimated that the HAI prevalence in Switzerland was stable in the past decade. Given the unchanged situation in Switzerland with a higher prevalence compared with its geographical neighbours, efforts in HAI prevention should be established.