Table 1

Quantitative studies of actual error rates

Type of studyInterventionSettingsStudy populationDrugsMain findingsOutcomeReferences
Cross over studyThree wards included WARD A: control (with two nurses) for first 23 weeks, and trial for second 23 weeks (with one nurse)
WARD B: trial with one nurse for first 23 week, and control (with two nurses) for second 23 weeks
WARD C: control for all study period with two nurses administering
Three wards of a geriatric assessment and rehabilitation unit, AustraliaRegistered nursesNon-restricted medicationsTotal errors=319; error rate/1000 medicines administered; one nurse 2.98; 95% CI 2.45 to 3.51; two nurses 2.12; 95% CI 1.69 to 2.55The use of two nurses to administer medication, statistically significantly reduced the medication error rate, but the clinical advantages were uncertain13
Retrospective studyRetrospective review of medication error reports completed from April 1994 to August 1999 (65 months)This study reviewed data routinely collected in Royal Hospital for Sick Children, Glasgow, UKNurses and pharmacy staffAll medicinesTotal errors=195; dispensing errors=39; Without double checking; 18 dispensing errors reported in 22 month period, that is, 9.8 per year; With double checking; 21 dispensing errors reported in 43 months period, that is, 6 per yearThe introduction of a policy of double checking for all drugs dispensed by pharmacy staff led to a reduction in dispensing errors from 9.8 to 6 per year14
Simulation studyTwo checklists for an ambulatory infusion pump were compared, one old, and one new. The new checklist had a specific item to check patient identity. Study was focused on the ability of the second nurse to detect errors by using the checklists. 14 pumps were checked by each nurseSimulated setting Toronto University Hospital, Canada10 Registered nursesChemotherapyOverall, the new checklist helped nurses to detect 76/130 (59%) of errors compared with 66/130 (51%) with old checklist; (p<0.01)No significant difference in detection of pump programming errors, but detection of errors in patient identification with new checklist (80%) was significantly higher than with the old checklist (15%)15