Characteristics of clinical decision support alert overrides in an electronic prescribing system at a tertiary care paediatric hospital.

CONTEXT  Electronic prescribing (EP) systems are advocated as a solution to minimise medication errors. Benefits in patient safety are often as a result of some clinical decision support (CDS) within the system. OBJECTIVE  To study the characteristics of the CDS alerts generated within a commercially available EP system in use at a tertiary care paediatric hospital in the UK. METHODS  Retrospective review and characterisation of CDS alerts recorded in the EP system over 1 year. RESULTS  A total of 16 182 conflict alerts were recorded when ordering 26 836 items, of which 3507 (13 alerts per 100 prescription orders (95% confidence interval, 12.8 to 13.6)) were visible to the user. Eighty nine percent (3119/3507) of all visible alerts were overridden by the user at point of prescribing. Drug-allergy conflict alerts were the most accepted, and exact drug duplication alerts the least. CONCLUSION  We found a high incidence of alert override, which is undesirable but consistent with that reported in the literature. The results suggest that the underlying algorithms for alert generation in many EP systems are not specific and need to be reviewed.