Background
A recent report from the Department of Health recommends that serious errors in the use of prescribed drugs should be reduced by 40% by year 2004^1. However, little is known about the current incidence of prescribing errors nor how this could be measured. We have previously attempted to determine the prescribing error rate in a UK hospital². This paper now describes some of the methodological issues encountered and our attempts to address them.

Methods
Following a series of training sessions, pharmacists in the study hospital recorded details of all prescribing errors identified during two one-month periods, about six months apart. Errors were assessed by a senior pharmacist and a senior clinical pharmacologist to identify those that were potentially serious. The number of medication orders written during each period was estimated using a 1 in 5 sample of all inpatients present in the hospital. During the first study period, medical staff were unaware that the study was taking place. However, during the second period, doctors making potentially serious prescribing errors were interviewed to explore the reasons why errors occurred; prescribers were therefore aware that a study was taking place. Data from the two periods were compared to explore the existence of any 'Hawthorne' effect, and to explore any differences amongst pharmacists in terms of their identification and documentation of prescribing errors.

Results
No evidence of a 'Hawthorne' effect was identified. However, there were considerable differences between wards and between pharmacists in terms of the numbers of errors identified. During each period, the investigator, who was acting as a ward pharmacist, identified the greatest number. Since many pharmacists changed wards between the two data collection periods, it could be concluded that most of the variation arose from the pharmacists themselves. However, there was much less variation in the numbers of potentially serious errors recorded, suggesting that pharmacists record serious errors most consistently.

Discussion
This analysis has highlighted potential sources of variation in the quantitative study of prescribing errors. These will be discussed, together with their implications for similar data collection methods and potential strategies for their minimisation.

References

1. Department of Health (2000). An organisation with a memory. Stationery office, London.
2. Dean, Schachter, Vincent and Barber (2000). Prescribing errors in hospital inpatients � how often do they occur? Paper submitted for publication.