Background and importance Drug–drug interactions (DDIs) are common and can result in preventable harm. Clinical decision support systems (CDSS) embedded within electronic prescribing software (eg DDI alerting tools) may improve clinical decision making. Studies have shown that prescribers override up to 96% of CDSS alerts and have questioned the usefulness of alerting systems.

Aim and objectives The study’s objective was to evaluate the characteristics and override rates of DDI alerts following a recent implementation of a hospital wide electronic prescribing systems incorporating a DDI CDSS which was set to flag ‘major contraindicated’ drug combinations only.

Material and methods A retrospective analysis of DDI alerts generated by Cerner electronic prescribing software over a 6 week period in a haematology–oncology inpatient cohort was completed. A parallel review of DDIs highlighted by clinical pharmacists in the same patient cohort was undertaken and the results were compared.

Results There were 310 electronic DDI alerts generated. Of these, 58 alerts were redundant as they referred to duplicates within the same prescribing episode (n=22) or were not triggered by current medications (n=36). The remaining 252 alerts involved 38 individual medicines and 44 medication pairs. Antiemetic medications accounted for over 50% of alerts and QTc interval prolongation was the most frequently alerted drug interaction adverse outcome. In 44 instances (17%) either the original prescription or the interacting medicine was changed by the prescriber following the DDI alert, reflecting an override rate of 83% (n=208).

A total of 37 DDI alerts were flagged by clinical pharmacists in the study. There were 42 individual medicines and 37 medication pairs involved. In 5 instances (14%), a change was made following the DDI alert. Thirty-one (84%) of the alerts flagged by clinical pharmacists did not trigger an electronic DDI alert.

Conclusion and relevance The volume and pattern of flagged DDIs varied between the electronic and pharmacist alerts. Override rates were high but consistent with the reported literature. Findings suggest changes which could be made to reduce the volume of redundant or irrelevant alerting.

References and/or acknowledgements No conflict of interest.