The study of medication errors at a teaching hospital using failure mode and effects analysis.

The prevalence of medication errors in a major teaching hospital was investigated using several methodologies. The existing ward stock drug distribution system was assessed and a new system designed based on a novel use of failure mode and effects analysis. The existing system was compared to the new unit supply individual patient dispensing system on two wards in terms of medication errors, nursing time, pharmacy time, drug costs, drug security and nurses' opinion. A review of a one year sample of reports submitted under the existing incident reporting scheme was also undertaken. Errors were categorised according to drug group, error type, reason cited for the error, and probability ranking (probability of occurrence, detection and harm). In addition, a "no-blame" medication error reporting scheme was implemented and assessed.Results of the study showed that in the newly designed individual patient dispensing system there was a reduction in nursing time associated with medication activities of approximately 29%, an increase in pharmacy staff time of 64%, a reduction in drug costs and an increase in drug security. Using the disguised observer methodology a reduction in medication errors by 23.5% (including timing errors) and 7.3% (excluding timing errors) was seen on Ward A. Similarly a reduction of 21.1% (including timing errors) and 9.8% (excluding timing errors) was observed on Ward B. Significant support for the individual patient dispensing system was given by nursing staff. Of the errors self-reported under the existing incident/accident reporting scheme the most common type of error was omissions (32.2%), the most common drug group was cardiovascular drugs (19.8%), and the most common cause of the error cited was a faulty check (42.3%). The probability ranking showed that 75% of errors reported scored between 12 and 17 points (from a possible 30 points). In the no-blame error reporting system, an error rate of 2.1% was detected in the existing system and 1.7% in the failure mode analysis designed phase