Baldwin 2010.

Methods	Randomised controlled trial. A (cRCT) design, in which the nursing home was a unit of randomisation, was used. Each group of residents in a participating home constituted one cluster. The intervention was tested in the intervention nursing homes for a period of 12 months while usual practice continued in the control homes. The study received ethical approval from the Office for Research Ethics Committees Northern Ireland. Baseline data were collected from residents and staff in all nursing homes at 3, 6 and 12 months. Resident data collected were age and gender, presence of wounds or indwelling devices, history of hospitalisation and antibiotic use in the previous 3 months and history of chronic illness. Consenting staff provided details on occupation, age and gender. Nursing home data such as number of beds (occupancy and capacity), staffing levels and ownership type were also collected. The participating nursing homes were then matched and paired using baseline data on number of beds per home, staffing levels, infection control audit scores and MRSA prevalence. NQuery version 6 produced the randomisation sequence in batches of two (for the pairs) with one home in each pair randomly allocated to the intervention or the control arm of the study.
Participants	Only facilities registered as general nursing homes with ≥20 residents, in one of four geographically defined health administration area in NI, and not participating in other studies were eligible to participate. All  nursing home residents aged ≥ 65 years were eligible ( excluding the terminally ill or those attending on a day‐care basis only) 32 nursing homes, 792 residents and 333 staff randomised
Interventions	Homes randomly allocated to the intervention arm received detailed information on their baseline infection control scores via a written report and verbal feedback from an infection control nurse which detailed infection control practice and how it could be improved. In‐depth infection control training, consisting of a 2 hour training session delivered via Powerpoint and DVD presentations, was provided to all intervention home staff by this same nurse. Practical demonstrations on hand hygiene and decontamination of equipment and the environment were also provided during the session. Selected staff from each intervention home were designated as infection control link workers, their role being to reinforce all aspects of good infection control throughout the study. These staff received additional training (5 hour during a period of one day). Infection control audits were also carried out in each home at baseline, 3, 6 and 12 months using an audit tool adapted from one previously developed for community practice. These measured compliance with infection control standards by collecting information using a standardised data collection form. Practice was observed and recorded for the following ten standards: cleanliness of the environment, cleanliness of the kitchen environment, decontamination of equipment, linen management waste and sharps management, hand decontamination, use of personal protective equipment, urinary catheter management, management of enteral feeding and management of wounds. On completion of each audit, an overall percentage score was calculated to determine compliance with good infection control practice with a score <75% indicating poor compliance, a score 0f 76 to 84% partial compliance and a score of >85% indicating compliance as recommended. Usual practice continued in the control sites with no training or feedback being delivered to staff, and no staff designated as infection control link workers.
Outcomes	A swab of the anterior nares was obtained from each consenting resident and staff member at each sampling time point in order to detect MRSA. A specimen of urine was collected from residents with indwelling urinary catheters, and those with wounds and any other indwelling devices provided swabs of these sites as relevant. The primary outcome was MRSA prevalence in residents and the secondary outcome was a change in infection control audit scores. MRSA prevalence was determined by calculating the proportion of residents and staff with MRSA at each home. Binary outcomes such as the proportion of residents positive for MRSA colonisation were compared between intervention and control homes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated NQuery version 6 produced a randomisation sequence in batches of two (for the pairs) with one home in each pair randomly allocated to the intervention or the control arm of the study.
Allocation concealment (selection bias)	High risk	The research team and nursing home staff were aware of home allocation.
Blinding (performance bias and detection bias) All outcomes	High risk	Infection control audits were undertaken at the time points by the researcher NB. Another infection control nurse blinded to allocation of homes performed audits in two randomly selected nursing homes at each time point, independent of the researcher to try to minimise measurement bias. Results from each set of audits were similar at each time point.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing outcome data balanced in numbers across the intervention and control groups, with similar reasons for missing data across groups. The total number of residents lost to follow‐up was 158 in the intervention group and 157 in the control group due to deaths in the resident population.
Selective reporting (reporting bias)	Low risk	MRSA prevalence rates in both arms remained similar. The infection control audit scores were comparable.
Other bias	Low risk	Characteristics of participating nursing homes and residents were described at baseline Contamination was unlikely as this study used a cRCT design in which the nursing home was the unit of randomisation. The intervention was tested in the intervention nursing homes for a period of 12 months while usual practice continued in the control homes.  A power calculation was given " based on a 10% reduction in MRSA prevalence at 5% significance and 80% power and an ICC of 0.01 indicated that 12 intervention and 12 control homes were required".