Ethics Statement

Ethical approval for the study was obtained by the medical ethics committee of the St. Elisabeth Hospital in Tilburg (NL 19489.008.07, protocol 0749, March 9^th, 2009). All participating cases, controls and their household members completed and signed an informed consent form. For children aged below 18 years the informed consent form was completed by both parents or a legal representative.

Study design and setting

A case-control study was conducted in patients recruited from sixteen hospitals throughout the Netherlands, i.e. three university, six teaching and seven non-teaching hospitals. Patients were recruited from July 2009 until July 2011.

Study population

An eligible case was defined as a hospitalised patient or a patient visiting the outpatients' clinic of one of the participating hospitals, who had an MRSA positive clinical culture from any site, was aged above one year, had no previous history of MRSA colonisation or infection, were not known to have acquired their MRSA through nosocomial transmission, and had no known healthcare- or livestock-associated risk factor(s) for MRSA carriage as described in the Dutch MRSA guideline [17]. Screening for potential eligibility was performed by the infection control practitioners of the participating hospitals. All potentially eligible cases were subsequently confirmed and examined for definite eligibility by the coordinating investigator. Patients from abroad, patients living in a health care facility, and patients who had died before the MRSA positive culture result became available were classified as not eligible.

For every potentially eligible case, the last patient without a history of MRSA colonisation or infection, who was admitted to the same ward or visited the same outpatients' clinic as the case, prior to the MRSA positive culture result of the case, was selected as a potential hospital-based control patient. Potential control patients were approached and asked to participate in the study. When a control patient refused, the next to last patient was asked to participate, and so on. To confirm the MRSA negative status of control patients, nose and throat samples were taken during the home visit and cultured for the presence of MRSA. In the Netherlands, the prevalence of MRSA carriage is 0.11%, so the probability that the selected control patients were MRSA positive was negligible. The remaining criteria for eligibility were identical to those for cases. A hospital-based control group was preferred over a population-based control group. First, the potential risk factors for CA-MRSA carriage that were examined in this study might well be associated with health status and the need for hospital admission. Second, the detection of CA-MRSA in cases was unexpected per definition, i.e. CA-MRSA was detected in cultures taken for other reasons. Therefore, it was important to select control subjects from a population where the potential presence of CA-MRSA would have been detected in the same way as in cases. As cases could be either MRSA infected patients or asymptomatic MRSA carriers, control subjects were selected independently of the presence or absence of infections.

Data collection during home visit

Cases and controls were visited at home by one of three trained interviewers. A standardised questionnaire was taken to identify risk factors for CA-MRSA carriage. Based on what was known from the available literature the following potential risk factors for CA-MRSA carriage were included: being of foreign origin, travel to a foreign country, day care centre attendance, professional contact with children, practicing sports, scuba diving, sharing diving equipment, and visiting a sauna [5], [10]–[13], [18]–[21]. Furthermore, as MRSA has been found in meat, the consumption of meat (beef, pork, poultry) and fish, and the preparation of dinner were included as potential risk factors [16], [22]. Finally, the questionnaire included questions on demographic characteristics. A separate questionnaire, including the same items, was completed for all household members who were willing to participate. A nasal swab and a throat swab were taken and cultured to confirm the MRSA status of cases and control subjects.

Genotyping of MRSA isolates

All MRSA isolates were genotyped by multiple-locus variable number of tandem repeat analysis (MLVA) [23]. MLVA is known for its higher discriminatory power as compared to either multilocus sequence typing (MLST) or pulsed field gel electrophoresis (PFGE ) [23]. The MLVA profiles were clustered using a categorical clustering coefficient (unweighted-pair group method using arithmetic averages, UPGMA).

Geospatial analysis

Data on municipality level variables of livestock and human population density, and residential locational were collected by geocoding cases and controls to their 6-digit postal code using geographic information system software (ArcGIS version 9·3), or when this was not sufficient, Google Earth. Municipality level data of the population, land area, numbers of chickens, pigs, and cattle were downloaded from the website of the Central Institute for Statistics (CBS) [24]. Population and livestock densities were calculated as the number of animals or persons per hectare of land in a municipality. For regression analysis, the natural logarithms of population and livestock densities were used.

Statistical analysis

All analyses were performed using SPSS version 19.0 (SPSS Inc. Chicago, IL, USA). For continuous variables differences between groups were tested with Student's t-test or Mann-Whitney U test when applicable. For categorical variables differences between groups were tested with the Pearson Chi-square test. Univariate and multivariate odds ratios (OR) with 95% confidence intervals (CI) were estimated using logistic regression analysis. All variables that were found to be associated with CA-MRSA carriage in the univariate analysis at a p-value≤0.10 were included in the multivariate analyses. As the potential risk factors for MRSA MC398 and MRSA non-MC398 might be different, multivariate analyses were stratified based on the MLVA-complex. P-values≤0.05 were considered statistically significant. Adjusting for multiple testing was not performed as this study was designed as an exploratory study [25].

Interpretation and implications

The finding that the regular consumption of poultry is associated with CA-MRSA carriage may be explained by the abundant presence of MRSA in meat [14], [15]. A recent study demonstrated that a substantial part of the meat pieces obtained from retail stores in the Netherlands were colonised with MRSA, including both MC398 and non-MC398 strain types. Colonisation rates varied from 10⋅6% for beef, 10⋅7% for pork, up to 16·0% for chicken and 35·3% for turkey [16]. To the best of our knowledge this study is the first to identify an association between the consumption of poultry and MRSA carriage in humans, which suggests that MRSA in the food chain may be a source for MRSA carriage in humans. The observed association might be explained by the incomplete cooking of the meat, but also by the consumption of simultaneously prepared food products, such as salads, using contaminated kitchen equipment.

In this study the sharing of scuba diving equipment was found to be independently associated with CA-MRSA carriage. This finding corresponds with previous publications of outbreaks of CA-MRSA infections in divers, and the role of sharing of scuba diving equipment as a possible mode of transmission of CA-MRSA [13], [20]. Whether the presence of skin abrasions in divers, the lack of decontamination of diving suites and other diving materials after use, or the favourable high salt content of sea water play a role in the observed association remains to be elucidated. The number of cases and control subjects that had dived the year preceding the detection of CA-MRSA is notably low, indicating that the majority of patients who had shared diving equipment acquired their CA-MRSA more than one year before the (unexpected) detection during hospital admission.

Based on the first studies that identified an association between the presence of MRSA in livestock and the occurrence of MRSA in humans, it was assumed that the risk for acquisition of MRSA MC398 was confined to persons with direct contact with livestock [2]. However, several outbreaks with MRSA MC398 in healthcare facilities have been reported since then, indicating that human-to-human transmission occurs, although it has been estimated that MRSA MC398 is 5.9 times less transmissible than MRSA non-MC398 [26]–[27]. In addition, a recently published case-control study identified pig-, cattle- and veal calf densities per municipality as independent risk factors for carriage of MRSA MC398 as compared to MRSA non-MC398 MRSA [4]. These latter finding is supported by the results of our study showing that persons without direct contact with livestock but living in areas with a high cattle density are at higher risk for MC398 MRSA carriage. Although it cannot be excluded that human-to-human transmission occurs in areas with a high MRSA MC398 pressure, environmental contamination with MRSA MC398 may play a role as well. MRSA has been shown to be present in air and soil samples collected downwind of pig and swine barns [28].

Several studies have found the presence of CA-MRSA to be associated with foreign origin [10]–[12]. The question is why being of foreign origin was not found to be associated with CA-MRSA carriage in our study. One explanation might be the selection of the study population. Where other studies have included CA-MRSA infections only, we included asymptomatic carriers as well. It might be that foreign origin is associated with the acquisition of strains that cause infection, e.g. PVL-positive strains, rather than asymptomatic carriage. In addition, previous reports may be biased by the inclusion of multiple patients from one cluster of CA-MRSA infections [10], [29]. In our study patients were derived from sixteen hospitals throughout the Netherlands. A total number of 58 different MLVA types were found, with no clustering of MLVA types, except for the livestock-associated MRSA MC398 (n = 29).

This study has several limitations. First, the study was aimed to identify risk factors for CA-MRSA, where the study population was hospital-based. This may have limited the extent of potential reservoirs to be detected as only patients were involved. Second, the participation rate was slightly higher for cases then for controls. However, as responders and non-responders were comparable with respect to age, gender, and social economic score, the potential for non-response bias is considered limited. In addition, the case definition used in this study assumes that the cases are primary cases, and does not take into account the possibility that a case might have been a secondary case, who had no exposure to the unknown source. This may have resulted in a decreased power to identify unknown risk factors. Finally, although the interviewers were intended to be blinded for the MRSA carriage status of the participants, some patients inadvertently mentioned their MRSA positivity during the interview. As a standardised questionnaire was used, the impact of this partial unblinding is expected to be limited.

The advantages of this study are the strict case definition, the acquisition of subjects from several study sites, and the performance of a home visit, which enabled thorough explanation of the questions to be completed in the questionnaire. Finally, this study was performed in a low MRSA prevalence setting (<0·1%), which allows for the identification of emerging sources [30].

Implications for practice and further research

This case-control study identified regular consumption of poultry, living in an area with high cattle density and sharing of scuba diving materials to be independently associated with CA-MRSA carriage. However, due to the case-control study design no indication can be provided on the absolute risk of CA-MRSA carriage when either one of these exposures is present. Further studies are warranted to confirm our findings and determine the absolute risks of MRSA acquisition.

Contributors

Members of the CAM Study Group:

Amphia Hospital, Breda: Yvonne Hendriks, Daniëlle op den Kamp, Jan Kluytmans, Marjolein Kluytmans – van den Bergh, Renée Ladestein, Rudolf Punselie, Miranda van Rijen, Erwin Verkade

Catharina Hospital, Eindhoven: Marieke Ernest, Mireille Wulf

Elkerliek Hospital, Helmond: Mariëlla Brekelmans, Patricia Willemse, Mireille Wulf

Franciscus Hospital, Roosendaal: Nelleke ten Feld, Rob Wintermans

Hospital Gelderse Vallei, Ede: Bent Postma, Marja Terwee

Groene Hart Hospital, Gouda: Truus de Ruiter, Eric van der Vorm, Antoinette Wijker

Johns Hopkins University, Baltimore: Beth Feingold, Ellen Silbergeld

Lievensberg Hospital, Bergen op Zoom: Henk Hamers, Rob Wintermans

Maxima Medical Centre, Eindhoven/Veldhoven: Philo Das, Ellen Sanders, Kees Verduin

Orbis Medical Centre, Sittard: Dick van Dam, Jan Diederen

Regional Health Authority, Leiden: Peter ten Ham

Reinier de Graaf Hospital, Delft: Loes Nolles, Eric van der Vorm

St. Anna Hospital, Geldrop: Anouk Smeulders, Kees Verduin

St. Elisabeth Hospital, Tilburg: Anton Buiting, Helma Hörmann, Ellen Nieuwkoop

University Medical Centre St Radboud, Nijmegen: Diana Haverkate, Paul Verweij

University Medical Centre Utrecht, Utrecht: Marc Bonten, Marischka van der Jagt-Zwetsloot, Annet Troelstra

VieCuri Medical Centre, Venlo: Hanneke Berkhout, Liduine van den Hout

VU Medical Centre, Amsterdam: Annie Kaiser, Christina Vandenbroucke-Grauls