Ding 2012.

Study characteristics
Methods	RCT ‐ cluster. Cluster‐randomised control‐experimental design in a general surgery patient ward in a tertiary hospital in Beijing. Unit of allocation: nurses Unit of analysis: prescriptions
Participants	Medication nurses and pharmacists in the chosen patient wards (N not available) IP adults (surgical wards)
Interventions	Intervention Technology Dispensing systems (for "processing" of the order). Automated dispensing Intervention: the Unit Dose Dispensing System, which was installed in the experimental group. The Unit Dose Dispensing System was installed only on TPN (total parenteral nutrition) doses. The data analysis was limited to TPN doses. Control: hand‐written patient charts were the primary method of prescription.
Outcomes	Total no. errors (including discrepancies) The ultimate outcome measure of the medication use system from the patient’s perspective is the rate of errors which reach the patient at the point of administration.
Notes	No financial support stated. No trial number
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The clusters of doses for two units with 29 beds in one unit and 24 beds in the other unit on the general surgery ward were randomly assigned to the control group or experimental group by flipping a coin.
Allocation concealment (selection bias)	High risk	The clusters of doses for two units with 29 beds in one unit and 24 beds in the other unit on the general surgery ward were randomly assigned to the control group or experimental group by flipping a coin.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participating nurses were informed that their normal medication preparation and administration processes would be observed.
Blinding of outcome assessment (detection bias) All outcomes	High risk	The direct observation method was used to detect and measure medication errors.
Incomplete outcome data (attrition bias) All outcomes	Low risk	The Principal Investigator excluded 7 doses (pre‐test) and 11 doses (post‐test) from the TOEs in the control group because they did not meet a priori operational definitions. A final total of 517 ordered doses plus 4 unordered doses were analysed for the statistical analysis in the control group, 41.7% of the total prescribed TPN doses. The Principal Investigator excluded 14 doses (post‐test) from the TOEs because they did not meet a priori operational definitions.
Selective reporting (reporting bias)	Unclear risk	The protocol of the study was not available.
Conflict of interest	Unclear risk	The authors did not address this issue.
Other bias	High risk	This is a cluster‐RCT with two clusters, where each cluster is randomly assigned. The authors argued that there cannot be any cluster effects because nurses worked across the two clusters. However, this argument does not consider any other factors that might have cluster‐level effects (e.g. unobsevable variables that might differ between the clusters), so it is unconvincing. It is also not immediately clear whether within‐patient clustering is possible.