A new knowledge structure for drug-drug interactions.

We developed a program to automatically screen patients' medication profiles for pairs of interacting drugs. Since some drug-drug interactions are indicated by changes in physiological parameters (e.g., ciprofloxacin and theophylline leading to an elevation of theophylline levels), the program considered the patients' relevant laboratory parameters prior to generating the alerts. We developed an editor to facilitate maintenance of the knowledge base. We evaluated the program for 3 weeks in two satellite pharmacies. The program reported 160 alerts of which 5 resulted in a change in the patients' therapies (one per 500 patient-days of care). These five interactions were potentially very serious. An additional 3 alerts led to changes in medication administration times. Subjectively, the program is well received and continues to be in routine clinical use.