Two surgical wards with endemic MRSA were matched for bed numbers, bed occupancy rates, floor area, staffing levels, case mix, antibiotic consumption, cleaning schedules and state and maintenance of internal fabric. Each ward contains 21 beds, with two side-rooms at the ward entrance and the remaining beds in groups of three to five in open cubicles on either side of the ward. Ward A is an acute male surgical ward situated on the first floor of a Victorian block, and Ward B, directly above, is for females, although each occasionally hosts patients of the opposite sex. Most patients were admitted for general surgical procedures, with occasional vascular and orthopaedic cases. Routine antibiotic prophylaxis for general surgical cases is amoxicillin, gentamicin and metronidazole, with cephalosporins used for patients allergic to penicillin. The wards share clinical staff and cleaning is performed by the same domestic team to a set specification 14. Bed occupancy rates are routinely monitored by Greater Glasgow & Clyde NHS Trust. In 2005, each ward experienced 20 new cases of MRSA, most thought to be ward-acquired.

We screened 10 hand-touch sites on both wards every week for 1 year in order to ascertain overall aerobic colony counts, presence of MRSA and meticillin-susceptible Staphylococcus aureus (MSSA) 78. Once-weekly screening took place on different days after the ward had received its routine daily clean. Sites chosen were hand-touch sites at bedside areas (patient lockers, overbed tables and bed frames), clinical equipment (patient hoist, infusion pump and blood pressure (BP) stand), sites at the nurses' work station (computer keyboard, desk and patient notes) and a side-room door handle. Patient curtains were not included because it was not possible to replace these outside of routine procedures. Hand-touch sites were sampled from one ward hoist, BP stand and computer keyboard but from different sites representing the remaining items throughout the ward, excluding those in the isolation rooms. Screening was continued for 7 weeks after completion of the study in order to monitor any residual effects.

Both wards adhered to identical infection control, screening and antimicrobial policies during the year. Any change in routine screening policy would have confounded the effect of cleaning as a single intervention, since it has already been shown to be an effective control mechanism in its own right 15. Patients awaiting surgery were routinely screened pre-admission or on the ward if they had been transferred from another unit, hospital or nursing home, or had a history of recent hospital admission, prior MRSA or recent contact with MRSA. All patients with newly diagnosed infections were also screened for MRSA, along with other clinically relevant specimens, in order to detect ward-acquired MRSA infection. Such infections were diagnosed by senior surgical staff using national criteria for surgical site, catheter-associated and respiratory infections 16. Those patients with no prior history of MRSA, who had been on the ward for >48 hours, had evidence of infection, and from whom the laboratory isolated MRSA were regarded as new cases of MRSA infection. These, along with known MRSA cases, were then screened every week until discharge from the ward.

Patients with MRSA were placed in isolation and managed according to local infection control policies. They were screened at carrier sites, started on a 5-day topical clearance regimen (chlorhexidine washes, mupirocin nasal cream and Corsodyl™ throat spray), and treated with appropriate antibiotics. Patients situated nearby were also screened for MRSA after any new case was found on the ward. If there were no isolation rooms available, patients were cohorted or moved to another ward; this included the other study ward. MSSA acquisition and infection were not audited during the study.

Enhanced cleaning began on Ward A in July 2006. The study domestic cleaned from 0730 h until 1530 h each weekday, with holiday and sickness cover provided by a colleague. Both cleaners were recruited from the existing domestic pool and trained to clean hand-touch sites. The study specification included all near-patient hand-touch sites two to three times per day, hand-touch sites at the nurses' station one to two times per day and clinical equipment one to two times per day regardless of nurses' cleaning responsibilities. The study cleaner also cleaned all door handles on the ward two to three times per day as well as items such as plastic wall-mounted racks for leaflets, visitors' chairs and hand-touch sites in the ward office and kitchen.

Day-to-day management responsibilities fell to senior nurses but the cleaners remained professionally accountable to the domestic supervisor. Routine ward cleaning continued as usual, including regular quality control audits on environmental cleaning 14. Routine cleaning comprised 2 to 3 hours each morning, concentrating on floors and bathroom facilities. There was a 'spot-check' performed later on during the day, in case toilets, in particular, required further cleaning attention. Equipment and cleaning consumables were standard NHS issue, stored separately and maintained according to current protocol. Consumables were chiefly CINCH™ detergent (AGMA, Haltwhistle, UK) and water and Tuffie™ wipes (Health Care Services, Nottinghamshire, UK). Disinfectants are not routinely used on these wards. The extra cleaner only operated on one ward at a time for a period of 6 months before being moved to the other ward. Each ward therefore acted as a control for the other. There was no additional cleaning performed out-of-hours, including weekends.

Total weekly MRSA patient-days were estimated by calculating all MRSA-positive patient-days each week for each ward 13. This was done by prospective auditing of all microbiological specimens submitted from both wards and knowledge of patient admission and discharge dates throughout the study. In addition, every patient with MRSA infection and/or persistent colonisation was screened weekly until discharge. Given that we did not screen every new admission to the ward, we probably underestimated the number of MRSA patient-days during the study, but this would have been consistent for both wards and study periods given the cross-over design. All cases of MRSA infection were identified promptly, whether acquired on the ward or before admission to the ward. Patients with MRSA infection are known to shed the organism into the environment more readily than those who are merely colonised 5.

Clinical samples were processed in the clinical laboratory according to standard operating procedures. Dipslides (Biotrace^®, Bridgend, UK) were used for environmental screening; these were coated with Baird-Parker (for staphylococci) and nutrient agars (for aerobic colony counts) 781417. Colonial growth was subjected to both quantitative (cfu/cm²) and qualitative assessment 2. One dipslide was used to screen adjacent sites, with each side pressed against the site for 5 seconds at a pressure of approximately 25 g/cm² 18.

After sampling, dipslides were incubated at 30°C in air for 48 hours. Microbial growth on nutrient agar was quantified as <2.5 cfu/cm², 2.5 to 12 cfu/cm², 12 to 40 cfu/cm², and 40 to 100 cfu/cm², according to the manufacturer's recommendations. Black colonies on Baird Parker were identified to genus level by colonial morphology and Gram film. Gram-positive cocci were differentiated by catalase test and staphylococci designated S. aureus or coagulase-negative staphylococci (CNS) by Staph-plus (Pastorex^®, Stockport, UK). Coagulase-positive isolates were subcultured onto MRSA chromogenic agar (Oxoid, UK) and incubated overnight at 37°C in air. Antimicrobial susceptibility testing was performed using the Vitek^® system (bioMerieux, Basingstoke, UK) standardised in accordance with the Clinical and Laboratory Standards Institute guidelines.

Clinical and environmental MRSA were sent to the Scottish MRSA Reference Laboratory, which confirmed identification by latex agglutination and DNase test. All isolates were subjected to PCR detection of S. aureus species-specific nuc gene, followed by PCR for the mecA gene 1920. Pulsed-field gel electrophoresis (PFGE) typing of SmaI- (Invitrogen, UK) digested DNA was performed by a modification of a previously described method 21. Briefly, S. aureus colonies from overnight cultures were incorporated into agarose plugs. After bacterial lysis, genomic DNA was digested using SmaI. PFGE was performed by clamped homogeneous electric field (CHEF) electrophoresis with a CHEF-mapper system (Bio-Rad Laboratories, California, USA). The fragments were separated with a linear ramped pulse time of 6.8 to 63.8 seconds over a period of 23 hours at 14°C. Gels were analysed using DNA analysis software GelCompar II version 5.1 (Applied Maths, Belgium) using the Dice correlation co-efficient. A dendrogram was generated using UPGMA with a tolerance of 1.5%.

The standard set for finding a potential pathogen is <1 cfu/cm² 2. For this study, we chose both MRSA and MSSA as indicator organisms 22. A second standard states that total aerobic colony counts (ACCs) from a hand-touch site should not exceed 2.5 to 5 cfu/cm² 223. Exceeding these levels suggests insufficient cleaning, masks the presence of a pathogen or implies an increased chance of finding a pathogen with similar epidemiological properties, for example, CNS and S. aureus 2. We therefore defined a hygiene failure as a site with ACC greater than 2.5 cfu/cm².

Environmental screening data and cases of hospital-acquired MRSA infection were modelled against routine and extra cleaning periods, MRSA patient-days and bed occupancy rates. Any potential transmission between patients and environment were investigated using molecular typing results, dates of isolation and location. Typing also allowed us to ascertain longevity of unique strains in the environment.

Since the project began in July, this allowed each ward to receive 6 months of enhanced cleaning to encompass both winter and summer months. The project was presented to domestic, nursing and medical staff and to the Trust Infection Control Committee for approval and comments before it was launched. It was approved by the R & D department and Ethics committee at the study hospital.

We calculated the overall costs of new MRSA infections potentially saved by the efforts from an extra cleaner. The study cleaner earned £12,320 pa and one new case of MRSA infection was assumed to cost £9,000 2425. Cost of cleaning consumables was included in the calculation.

From previous data we expected a background rate of two to four hygiene failures per week 7. The study was powered to detect a 50% reduction in hygiene failures over a 6-month period, with powers of 97% assuming a background rate of four per week in the normally cleaned ward, 93% with a background rate of three, and 81% with a background rate of two. We expected to find about 20 new MRSA cases on the control ward over the year, assumed to have been acquired on the ward. This only has a power of about 44% to detect a 50% reduction in the numbers of MRSA cases.

The main test of the hypothesis that extra cleaning has an effect on ACC is based upon a linear regression of the log total ACC with terms of cleaning, ward, period and quarter used to investigate the effect of enhanced cleaning on total ACC. A log transformation was required to satisfy the normality assumption in the linear regression and residual plots were used to verify that this assumption was appropriate. This analysis is more powerful than an analysis of hygiene failures, since ACC is a quantitative measurement and hygiene failure is a categorisation of that. The study was powered in terms of hygiene failures, and the powers for ACC analysis will be higher as it is a quantitative measurement rather than a categorical variable. Binomial logistic regression models with factors for ward, cleaning and site were used to estimate the effect of enhanced cleaning on environmental levels of MRSA/MSSA and hygiene failures. Poisson log-linear regression models were used to estimate the effect of enhanced cleaning on new MRSA infections in a ward adjusting for MRSA patient-days and ward, and interaction tests were used to assess the differences in effect over the two wards. Significance levels of 5% were used for the pre-specified hypotheses and 95% confidence intervals for the estimated effects, which, for the overall effect of enhanced relative to routine cleaning, were adjusted for ward differences.

Figure 1 shows the total quantity of microbial growth recovered from both wards over the year and for 2 months after the study finished. New patient MRSA infections are also indicated. ACCs were 32.5% less with enhanced cleaning compared with normal cleaning (P < 0.0001; 95% CI 20.2%, 42.9%) (Table 1). Ward B had an overall 19.0% lower ACC than Ward A (P = 0.015; 95% CI 4.3%, 31.4%). There was no residual evidence of any temporal effects of period (P = 0.58) or quarter (P = 0.18). In Ward B, the extra cleaning was associated with a 45.4% reduction in ACC (P < 0.0001; 95% CI 33.8%, 55.1%), whereas it was only associated with a 16.5% reduction in ACC for Ward A (P = 0.21; 95% CI 36.7% reduction,-10.1% increase).

The rate of hygiene failures with routine cleaning averaged 3.2 per week, as anticipated. Over both periods the enhanced cleaning was associated with a 50.3% reduction in the odds of a hygiene failure (95% CI 33.8%, 62.7% reduction), though there was evidence of a different effect between the two wards, P < 0.0001, and the two periods (P = 0.0002). In period 1 there were 86 hygiene failures in Ward B (normal cleaning) and 73 in Ward A (extra cleaning), corresponding to a 21.0% reduction (95% CI 45.6% reduction, 14.8% increase) in the odds of having a hygiene failure at a site with enhanced cleaning. In period 2 there were 82 hygiene failures in Ward A (normal cleaning) and 27 in Ward B (extra cleaning), corresponding to a 74.7% reduction (95% CI 84.3%, 59.3%) in the odds of having a hygiene failure with enhanced cleaning. In Ward A, where the extra cleaner was used in the first period, the odds of hygiene failure was 15.3% lower (95% CI 41.8% reduction, 23.5% increase) when the extra cleaner was used. In Ward B, the odds of hygiene failure was 76.5% lower (95% CI 85.4%, 62.1%) when the extra cleaner was used.

MRSA was recovered from 12 sites when the wards received extra cleaning compared with 16 sites when the wards were being routinely cleaned (P = 0.45) (Table 1). This was in contrast to the number of MSSA isolates recovered during routine and enhanced cleaning periods, since MSSA was recovered from 37 sites on wards receiving enhanced cleaning and from 30 sites on wards during routine cleaning periods (P = 0.37).

There was some evidence for different amounts of MRSA/MSSA across the sites (P = 0.065) (data not shown). Sites more frequently contaminated with MSSA/MRSA were the bedside locker (17 isolates), overbed table (13 isolates) and bed frame (12 isolates). These are the three sites closest to the patient. There was less contamination on the infusion pump, BP stand and computer keyboard. Whether or not the ward had enhanced cleaning did not affect the site contamination rankings (P = 0.52), nor did the ward on which sites were located (P = 0.15). The least-contaminated site throughout the study was a touch pad on the infusion pump.

There was little difference in bed occupancy rates between the two wards throughout the study, although both had higher rates during their respective extra cleaning periods (Table 1). Ward A had bed occupancy rates of 89.7% during routine cleaning and 90.5% during enhanced cleaning; Ward B had rates of 85.7% during routine cleaning months and 91.8% when receiving extra cleaning.

MRSA-positive patient-days ranged from 0 to 25 per week (Table 1). The total number of MRSA patient-days on Ward B during routine cleaning was 95, giving an average of 3.65 MRSA patient-days each week for the 6-month period. During enhanced cleaning, the total number of MRSA patient-days was 204, giving an average of 8 MRSA patient-days per week for this period. Thus, Ward B's clean period experienced more than double the number of MRSA patient-days than the period without extra cleaning. For Ward A, total MRSA patient-days during routine cleaning were 232, giving an average of 9 MRSA patient-days per week. During enhanced cleaning, the total was 271, giving an average of 10.42 MRSA patient-days per week for this period. As for Ward B, total MRSA patient-days were greater for the period without extra cleaning, but not by such a large margin.

The total number of positive MRSA screens on each ward for both periods were examined after removal of duplicates: there were 17 MRSA-positive patients identified on Ward A during the first period (enhanced clean) and 14 during the second period (routine cleaning); on Ward B, there were five MRSA-positive patients during the first period (routine clean) and nine during the second period (enhanced clean). These numbers include the patients who acquired their MRSA on the wards.

During normal cleaning there were a total of nine new cases of MRSA infection on both wards with a total of 327 MRSA patient-days. During enhanced cleaning periods, there were 475 MRSA patient-days overall, leading to an expected 13.1 (= (9/327) × 475) new cases of MRSA occurring during the months the wards were receiving extra cleaning. This assumes that there is a relationship between new MRSA cases and MRSA patient-days, and that enhanced cleaning has the same effect as routine. In fact only four new MRSA infections were observed during enhanced cleaning periods. The rate of new MRSA infections during enhanced cleaning was 26.6% (95% CI 7.7%, 92.3%) of the rate during normal cleaning (P = 0.032). It is important to emphasise that the power to detect a reduction in new cases of MRSA was low to start with, and the overall low numbers occurring during the study must be interpreted with caution.

There were no changes in antibiotic policies during the study, nor differences in antibiotic consumption, save for a small increase in use of linezolid on Ward A for complicated MRSA infections. There were no additional hand-hygiene interventions other than routine educational visits from the infection control team including poster reminders. We did not formally audit hand-hygiene compliance.

A simple cost analysis was performed. The study cleaner earned £12,320 per annum and consumables were £1,100; this was offset against the number of patients potentially spared MRSA infection. The average cost of one hospital-acquired surgical site infection (SSI) caused by MRSA is estimated as £9,000 2425. If it is correct to assume that five to nine patients were potentially spared MRSA SSI, the hospital saved in the region of £45,000 to £81,000 minus the costs of the cleaner and consumables for one year. This gave overall cost savings of £31,600 to £67,600.

Tables 2 and 3 show the PFGE profiles of patient and environmental MRSA strains and their isolation dates. Not all isolates were typed, since some did not survive storage. On Ward A, there were six different PFGE profiles from both patients and the environment (Table 2). Of these, one strain (15b) was recovered from patient notes on 25 October 2006 (E4A), two new patient infections on 14 February 2007 and 18 February 2007 (P14A, P15A), overbed table on 19 February 2007 (E11A) and a BP stand on 21 July 2007 (E16A). Strain 15z/15–71 was isolated from a desk on 4 July 2007 (E15A), followed by a new patient on 12 July 2007 (P18A). The dendrogram suggests a close genetic relationship between these two sets of strains (Figure 2). In addition, there were similar strains isolated from a desk on 13 February 2007, bed on 19 February 2007 and bed again on 22 June 2007 (E9A, E10A and E13A) (Figure 2).

On Ward B, there were four different PFGE profiles from both patients and the environment (Table 3). Of these, two types were isolated from the environment before they appeared as new infections in patients. Type 15c was found from an infusion pump on 8 March 2007 (E9B) before it was isolated from two new patients on 16 July 2007 and 24 July 2007 (P8B, P9B); type 16b was found on a computer keyboard on 25 September 2006 (E5B) and overbed table on 6 October 2006 (E6B) before it was recovered from a patient on 11 April 07 (P7B). Again, the dendrogram suggests a close genetic relationship between these strains (Figure 2). Two further clusters are shown, one involving two new patients (P1B, P2B) on 11 August 2006 and 27 November 2006, a door handle on 8 May 2007 (P10B) and a bedside locker on 22 June 2007 (P11B). The second shows four similar strains from a locker, bedside table and BP stand during the first month of the study (E1-4B).

There were thus indistinguishable strains from patients, which were later found in the environment, and vice versa (Figure 2). Indistinguishable strains were recovered from different sites around the ward, sometimes weeks apart 3567. Not surprisingly, some strains appeared to move between wards. Staff were not screened during this study, however, which means that a carrier role by staff cannot be excluded in any transmission hypotheses 67.