Twenty seven studies were included. The sample size ranged from 57 to 172,224 patients and from 67 to 204,681 orders (where stated). There were two randomised controlled trials, six time-series analyses, 14 pre/post studies, four cohort studies, and one pre/post study with untreated controls. Study quality was generally weak. Several studies were retrospective. Few had comparison groups recruited over the same period. Half conducted reliable outcomes measurement, but in most cases it was unblinded. Most studies used adequate hypothesis tests and reported measures of statistical variability, but few adjusted for confounding or clustering. Reporting quality was poor overall.
Electronic prescribing versus paper-based or less sophisticated electronic prescribing:
Medication errors (25 studies): Twenty three studies reported a statistically significant risk reduction in the intervention group, with relative risk reduction ranging from 13 per cent to 99 per cent. One study was inconclusive. One study reported a significant 26 per cent increase in errors in the intervention group, which was attributed to confounding and potential selection bias.
Potential adverse drug events (nine studies): Six studies reported a statistically significant risk reduction in the intervention group, with a relative risk reduction of 35 per cent to 98 per cent. Two studies were inconclusive. One reported a statistically significant 29 per cent increase in errors in the intervention group.
Adverse drug events: Four studies reported a significant risk reduction in the intervention group, with a relative risk reduction of 30 per cent to 84 per cent. Three studies (one with no events) found no statistically significant difference between the groups.
Subgroup analyses (seven studies): A greater reduction in medication errors was noted in studies using home-grown or more advanced electronic systems, and in those comparing electronic with handwritten prescribing. Effects differed depending on the method used for detecting errors.