Grading the severity of drug-drug interactions in the intensive care unit: a comparison between clinician assessment and proprietary database severity rankings.

Background: Computerized provider order entry with decision support software offers an opportunity to identify and prevent medication-related errors, including drug-drug interactions (DDIs), through alerting mechanisms. However, the number of alerts generated can overwhelm and lead to "alert fatigue." A DDI alert system based on severity rankings has been shown to reduce alert fatigue; however, the best method to populate this type of database is unclear.

Objective: To compare the severity ranking of proprietary databases to clinician assessment for DDIs occurring in critically ill patients.

Methods: This observational, prospective study was conducted over 8 weeks in the cardiac and cardiothoracic intensive care unit. Medication profiles of patients were screened for the presence of DDIs and a severity evaluation was conducted using rankings of proprietary databases and clinician opinion using a DDI severity assessment tool. The primary outcome measure was the number of DDIs considered severe by both evaluation methods.

Results: A total of 1150 DDIs were identified after 400 patient medication profiles were evaluated. Of these, 458 were unique drug pairs. Overall, 7.4% (34/458) were considered a severe interaction based upon proprietary database ratings. The assessment by clinicians ranked 6.6% (30/458) of the unique DDIs as severe. Only 3 interactions, atazanavir-simvastatin, atazanavir-tenofovir, and aspirin-warfarin, were considered severe by both evaluation methods.

Conclusions: Since proprietary databases and clinician assessment of severe DDIs do not agree, developing a knowledge base for a DDI alert system likely requires proprietary database information in conjunction with clinical opinion.