Gordon 2017.

Study characteristics
Methods	RCT ‐ cluster. Cluster‐randomised controlled trial was conducted in a UK teaching hospital (Blackpool Victoria Hospital), including all medical prescribers in four randomised inpatient ward areas. Four inpatient wards were purposefully selected for inclusion in the study: a children’s ward, an orthopaedic ward, an endocrine ward and a cardiology ward. These were selected as they represent a range of clinical specialisms, as well as all having almost mutually exclusive clinical teams, with the aim of preventing contamination. All medical staff who prescribe in each of these clinical areas were contacted by email, at departmental meetings and through on‐ward recruitment over an 8‐week period during March–April 2016. The number of clusters was fixed at four ward areas, two in each group, which all use paper‐based prescribing. Unit of allocation: wards Unit of analysis: prescriptions
Participants	All medical prescribers in four randomised inpatient ward areas of a UK teaching hospital. Consent was obtained from 55 prescribing doctors out of a possible 123 in those areas (44.7%). No one withdrew consent during the study (N not available). IP adults (medical and surgical wards)
Interventions	Intervention Structural/organizational, Organizational changes. Intervention Technology. The commercial e‐prescribing system was implemented in inpatient wards in November 2009, and includes barcode scanning technology for patient identification. Barcode wristbands are now given to blood transfusion and chemotherapy patients, while either visual or verbal identication is used to identify patients for other types of treatment. After an assessment of prescribing on each ward, a ward‐specific feedback document was prepared, giving general and anonymous feedback, and forwarded to all consenting participants in the intervention areas. Intervention wards: prospective ongoing prescribing error feedback Control wards: no feedback. No e‐prescribing system
Outcomes	The primary outcome was total prescribing order error rates (calculated as the number of medication orders with any error as a percentage of the total medication orders audited); secondary outcome measures included clinical order error rates, technical order error rates and cost per error prevented. Prescriptions were eligible for assessment, if they were active on the day of data collection: "once only" drugs, regular medication orders, "when required" drugs and continuous infusions. Errors were not recorded if they had been corrected by the prescriber immediately, but were recorded if they had been corrected by other staff.
Notes	No registration was obtained. Funding: Blackpool Victoria Hospital (host organisation) provided a pump priming grant of £6900 to support this work internally. The department had no involvement in the carrying out or writeup of the study, but did peer review the protocol before funding and as part of internal and ethics approval.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was completed used a computer‐generated random number list.
Allocation concealment (selection bias)	Low risk	Allocation was concealed using sealed opaque envelopes, with assignment to the next sealed envelope as per the random number list.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Involvement in the trial would not impact on the routine screening, quality assurance and intervention processes conducted by the pharmacists. Participants were not aware of which group their area would be randomised to on enrolment.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Ward pharmacists in each of the study groups began collecting data using a bespoke electronic pro‐forma, with several changes made to the interface and content based on feedback. A senior pharmacist acting as principal investigator performed reliability checks during this period to confirm the appropriate and consistent recording of data. The error data was aligned with the previously published EQUIP trial, in which this hospital participated for data collection. All interventions on the ward by the pharmacist were maintained as normal during this process. There is no mention regarding blinded assessment of outcomes but the process is very transparent and supervised. The outcome measurement is not likely to be influenced by lack of blinding.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No one withdrew consent during the study. There were no missing data.
Selective reporting (reporting bias)	Low risk	The study protocol is not available but it is clear that the publications include all the expected results, including those that were prespecified.
Conflict of interest	Low risk	Authors have completed the International Committee of Medical Journal Editors (ICMJE) disclosure form. With the exception of the declared funding, there has been no other financial support for this work, no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years, no other relationships or activities that could appear to have influenced the submitted work.
Other bias	High risk	The outcome was measured for each prescription, prescriptions seem to be clustered within audits, which were clustered within wards; wards were randomised. The analysis does not account for the possible cluster effects of ward or audit. It may be possible to use the published P value under the assumption that there is no effect of audit.