A prospective study to examine the epidemiology of methicillin-resistant Staphylococcus aureus and Clostridium difficile contamination in the general environment of three community hospitals in southern Ontario, Canada

Three community hospitals, referred to as A, B, and C, located in southern Ontario, Canada, participated in this study. Information pertaining to each healthcare facility is presented in Table 1. Environmental sampling was conducted in February and March, 2010 in hospitals A and B and June 2010 for hospital C. During the study period, none of the hospitals identified MRSA or C. difficile outbreaks. This study was approved by the research ethics boards of the University of Guelph and the participating hospitals.

Structured sampling of a pre-determined set of environmental surfaces distributed over the general medicine and surgical wards at each participating hospital was conducted one day a week for four consecutive weeks. The selection of surfaces to be investigated for contamination was based on data presented in previous studies [11, 13, 20] and through discussion with hospital staff regarding cleaning and disinfection practices. Environmental surfaces were sampled four times as this would provide good quality data regarding contamination in the general ward environment. Surfaces sampled were broadly replicable between wards and hospitals, in addition to sites that were relevant and specific to each ward. Surfaces were sampled prior to that day’s cleaning by housekeeping staff. All hospitals used a one step cleaner and disinfectant protocol and wards were cleaned once per day. No changes in the cleaning protocol were instituted during the study period.

Dry sterile electrostatic cloths (Swiffer™, Proctor and Gamble, USA) were used for environmental sampling. Using a gloved hand, the investigator wiped the electrostatic cloth over half the environmental surface to be sampled, up to a maximum of 20 cm x 20 cm. The cloth was then placed in a sterile collection bag (Whirl-Pak®, Nasco, California, USA). A second electrostatic cloth was used to sample the other half of the environmental surface. Gloves were changed between each sample. During each hospital visit, two electrostatic cloths were not used for sampling but were handled and processed in parallel, to act as negative controls for quality assurance. Information collected with each surface sampled included: date, hospital, hospital ward, type of surface (e.g., sofa), surface material (e.g., fabric), and if the surface was accessible by the public or only by hospital staff. For MRSA, impact air sampling was also performed using an air IDEAL ®3P air sampler (Biomérieux, Saint Laurent, Quebec, Canada) in elevator areas, nurses’ stations, and waiting rooms. A total of 500 L of air was collected onto selective MRSA culture agar (BBL CHROMagar MRSA, Becton, Dickinson and Company, Sparks, Maryland, USA) over a five minute period. Air sampling was conducted at a height of 1 metre and a minimum of 1 metre from the nearest surface.

From all participating facilities, MRSA isolates from patients hospitalized in the medical and surgical wards during the study period were obtained from the microbiology laboratory following MRSA confirmation. Isolates were collected from culture plates using a culture swab with Stuart’s media. Stool samples from patients diagnosed with a C. difficile infection by a fecal toxin test (Tox A/B Quik Chek, TechLab, Blacksburg, Virginia, USA) were obtained from Hospital C only. All isolates were collected at the discretion of medical personnel. For MRSA and C. difficile, only one isolate per patient was collected.

For MRSA, cloths were immersed in 30 ml of enrichment broth containing tryptone (10g/L), sodium chloride (75 g/L), mannitol (10g/L), and yeast extract (2.5g/L) and incubated at 35°C, aerobically, for 24 hours. Following incubation, approximately 5 μL of broth was inoculated onto MRSA Chromogenic agar (BBL CHROMagar MRSA, Becton, Dickinson and Company, Sparks, Maryland, USA) and incubated, aerobically, at 35°C for 24-48 hours. Colonies with an appearance consistent with MRSA were sub-cultured onto blood agar (Oxoid, Nepean, Ontario, Canada) and identified as S. aureus by Gram stain, catalase test, tube coagulase test, and S. aureus latex agglutination assay (Pastorex Staph-plus, Bio-Rad Laboratories Ltd, Mississauga, Ontario, Canada). The presence of methicillin-resistance was confirmed by testing for penicillin-binding protein 2a (MRSA latex agglutination test, Oxoid Ltd., Hants, UK). For air samples, agar plates were incubated and processed as described above. For patient isolates, culture swabs were streaked onto blood agar (Oxoid, Nepean, Ontario, Canada) and processed as described above. Molecular typing of MRSA was conducted using sequence analysis of the X region of the staphylococcal protein A gene (spa typing) [23]. Sequences were then analyzed using two different methods; eGenomics software [24] and the Ridom system [25]. Based on eGenomics, spa types are reported using a numerical system (e.g., spa type 2) whereas Ridom spa types are reported using a numerical system preceded by a ‘t’ (e.g., t002). The spa types obtained were compared to epidemic MRSA clones that are frequently found in North America [26]. All MRSA isolates were investigated for the lukF-PV gene encoding the Panton-Valentine leukocidin toxin by real-time PCR [27].

For C. difficile, cloths were immersed in 30 ml of brain-heart infusion broth supplemented with 0.1% sodium taurocholate and incubated anaerobically at 37°C for 5 days. A 2 ml aliquot of broth was alcohol shocked by addition of an equal volume of anhydrous alcohol and incubated at room temperature for one hour followed by centrifugation at 4,000 rpm for 10 min. The resulting pellet was then inoculated onto C. difficile moxalactam-norfloxacin agar (Oxoid, Nepean, Ontario, Canada) and incubated anaerobically for 24-96 hours at 37°C. Presumptive colonies were sub-cultured onto blood agar (Oxoid, Nepean, Ontario, Canada) and identified as C. difficile based on characteristic colony morphology, odour, and production of L-proline-aminopeptidase (Prodisk, Remel, Lenexa, Kansas, USA). For patient isolates, approximately 1 g of feces was inoculated into 9 ml of brain-heart infusion broth and processed as described above. All isolates identified as C. difficile were investigated for the presence of genes for toxin A (tcdA) [28], toxin B (tcdB) [29], and binary toxin (cdtA) [30] using PCR. Ribotyping was also performed [31]. When a ribotype pattern was known to be an international ribotype based on comparison to reference strains, the appropriate numerical designation (e.g., 027) was assigned. Otherwise, an internal laboratory designation was assigned. Toxinotyping [32] was performed on a representative of each toxigenic ribotype.

The Spearman’s rank correlation test was used to identify correlations between independent variables and the correlation between MRSA and C. difficile contamination at the sample level. Independent variables with a correlation >0.8 were investigated and only the variable that was more biologically plausible was included in the model to avoid issues associated with collinearity [33]. Additionally, for common surfaces, a paired exact logistic regression was conducted to determine if there were any significant differences in the probability of a surface being contaminated with MRSA or C. difficile.

Univariable logistic regression models were constructed to screen the independent variables with each dependent variable using a significance level of α ≤0.25. Multivariable models were constructed by a manual backwards step-wise procedure starting with all significant variables based on the liberal P-value. Confounding was evaluated by examining the effect of the removed variables on the coefficients of the remaining variables. A variable was deemed to be a confounder if it was not an intervening variable and the log odds of a statistically significant independent variable changed by at least 20% [34]. Interaction terms were examined for all independent variables in the final main effects model. Using the final multivariable model, the investigators examined contrasts for independent variables with >2 categories. These contrasts allowed the researchers to investigate significant differences between any two categories.

Random intercepts were removed from the model if, based on a likelihood ratio (LR) test, they were not statistically significant and they did not confound the observed associations. Similarly, Akaike’s Information Criteria (AIC) were examined among models with and without the random intercepts to assess which model provided the best fit. If random effects were included in the final model, standardized Pearson residuals were assessed to identify outliers. In addition, normality and homogeneity of variance for the best linear unbiased predictors were examined to assess model fit. If random effects were not included in the final model, a Pearson χ² test was used to assess model fit and standardized Pearson residuals were evaluated to identify outliers.

All descriptive statistics, model building, and analyses were performed using Stata 11.0 (StataCorp LP, College Station, Texas, USA). All tests were two-sided and statistical significance was based on an α ≤ 0.05.

Due to the low number of surfaces positive for C. difficile, a statistical model could not be constructed for C. difficile. Consequently, only descriptive statistics for C. difficile are reported.

From the three participating hospitals, 208 different surfaces, for a total of 612 samples, were tested for MRSA, while 191 different surfaces, for a total of 552 samples, were tested for C. difficile. Overall, 11.8% (72/612; 95% CI 9.30-14.6%) and 2.4% (13/552; 95% CI 1.3-3.9%) of surfaces sampled were positive for MRSA and C. difficile, respectively. For common surfaces, there was no significant difference in the prevalence between MRSA and C. difficile (OR=1.13; 95% CI 0.15–6.92; P=0.999). The Spearman’s correlation coefficient indicated that MRSA and C. difficile contamination was not correlated at the sample level (ρ=0.05; P=0.233). The proportion of surfaces that tested positive at least once for MRSA or C. difficile is presented in Table 2, with Hospital C identified as having the highest prevalence of both MRSA and C. difficile in the general environment. Data pertaining to the prevalence of MRSA and C. difficile based on visit, ward, surface material, surface location, and surface sampled are presented in Table 3. Over the study period, the prevalence of MRSA and C. difficile fluctuated in all three hospitals. None of the negative control cloths tested positive for MRSA or C. difficile.

Among the 72 MRSA isolates collected from the hospital environment, seven different spa types were identified (Table 4). A total of 60 air samples were taken during the study with 6.7% (n=4) positive for MRSA. All MRSA air isolates were spa type 2/t002. Overall, 46 MRSA isolates from patients were obtained during the study. Eight different spa types were identified (Table 4).

For C. difficile, eight different ribotypes were identified among the 13 isolates from the environment including internationally recognized ribotypes 027 (15.4%, n=2) and 078 (7.7%, n=1). Five C. difficile isolates, each representing a different ribotype, were collected from patients from Hospital C only. Data pertaining to the characterization of C. difficile isolates are presented in Table 5.

In Hospital A, a drug cart was contaminated with three different spa types, 140/t954 (visit 1), 24/t242 (visit 3), and 2/t002 (visit 4), all consistent with the Canadian epidemic MRSA (CMRSA) 2 clone. In Hospital B, a handrail, an isolation cart, and an air sample taken in the elevator area were contaminated with MRSA on two visits each. For the isolation cart and elevator air sample, all MRSA were identified as spa type 2/t002. However, for the handrail, on visit two, spa type 2/t002 was identified and on visit 4 spa type 539/t034 was identified. In Hospital B, a sofa was identified as being contaminated with spa type 2/t002 on three different visits. In Hospital C, the back of a chair located in a nursing station was identified as being contaminated with MRSA on more than one visit; spa type 7/t064 on visit two and spa type 1/t008 on visit 4. No surfaces were identified as being contaminated with C. difficile on more than one visit.

Three surfaces were identified as being contaminated with both MRSA and C. difficile on the same visit. These surfaces included a glove box holder and a visiting room sofa in Hospital A and an isolation cart in Hospital B. Four surfaces were identified with being contaminated with MRSA and C. difficile, on different visits. These surfaces included a drug cart in Hospital A, and a nursing station counter top, an isolation cart, and a patient chart in Hospital C.

For the initial univariable analysis, the variables surface material, type of surface, hospital, and the presence of C. difficile were significant at the 25% level for the presence of MRSA on a surface (Table 6). There was no statistically significant association in the univariable models between type of ward and MRSA contamination, or surface location and MRSA contamination.

For the final multivariable model, three variables were included: hospital, surface location, and type of surface (Table 7). In constructing the multilevel models for MRSA contamination, the size of the variance components for ward and surface were extremely small (e.g., <10^-6). Furthermore, the LR test was not statistically significant (P>0.99) comparing models with one or both random intercepts compared to a regular logistic regression model. Similarly, the AIC was smaller when the random effects were not included. Therefore, a regular logistic regression model was used. The variable hospital was forced into the final model as a fixed effect to control for clustering at the hospital level and the potential confounding effects of management differences between hospitals. Surface location was included in the final model as it was a confounder for the variable type of surface. The only statistically significant independent variable in the final model was type of surface. Specifically, results from the multivariable logistic regression model indicated that the odds of contamination for chair backs, hand rails, isolation carts, and sofas were significantly higher than computer keyboards (Table 7). Statistically significant contrasts between the different types of surface categories are presented in Table 8. Similarly, the mean predicted probabilities for surfaces contaminated with MRSA, while fixing hospital and surface location at a referent category, are presented in Figure 1. As demonstrated in both model-based contrasts and the mean predicted probabilities, sofas, hand rails, chair backs, and isolation carts had a higher probability of being contaminated with MRSA compared to other surfaces commonly found in the ward environment. Interactions between variables could not be assessed due to the large number of categories for type of surface and the resulting small number of observations per interaction term.

Standardized Pearson residuals were assessed and only two outliers were identified (i.e., residual values of 2.55 and 2.71). For both outliers, data were assessed, found to be recorded correctly, and the observations were kept in the final model. The Pearson χ² test was not significant (P=0.25) indicating that the model fits the data.