Evaluation of the use of a medication database as a tool for detecting drug interactions

At the department of geriatrics, medication is prescribed using a computerised system. The system includes a database on which the medication history of the separate patients is saved. The database enables the system to check newly prescribed medication for interactions, overdoses or double medication. The system immediately warns the prescribing physician for possible unforeseen complications so that he/she can adapt the prescription if considered necessary. The aim of this survey is to register all medication control signals given by the automated database, to register the actions taken because of the signals and to evaluate the clinical relevance of the different signals. Finally the survey should result in recommendations for further use of automated medication control. A questionnaire was used for the registration and evaluation of all given signals. Each prescribing physician was asked to complete the questionnaire for each signal. Clinical relevance was discussed with the responsible physician and possible signs as a result of the (denied) signal were looked for in medical status. Signals given by the database were categorised as: ‐useful signals, ‐unnecessary signals although clinically relevant, ‐not clinically relevant, ‐useless or incorrect. In a survey period of 2 months, 60 signals were registered. Of these 78% concerned a warning for possible interactions. The number of signals with furosemide (21/60) and digoxin (14/60) were most frequent. Only 20 signals were considered to be useful. The other 40 were unnecessary, not clinically relevant or even incorrect. Signals involving coumarin anticoagulants were considered to be useful and relevant. Besides extra control of the prothrombin time, these signals resulted in a dose adjustment in four cases of the seven involving acenocoumarol. Possible pharmacokinetic interactions with digoxin (except for the majority of signals involving the combination with furosemide) and lithium were also considered to be useful and plasma concentrations were verified more often. Other relevant and useful signals were several interactions due to metabolic enzyme induction or inhibition and those involving magnesium and iron affecting the bioavailability of several compounds. Misplaced or even incorrect messages distract the attention of the prescribing physician from the useful signals. It should technically be possible to avoid incorrect messages (a warning for an overdose with digoxin while it was actually a normal loading dose) or misplaced signals (there is no risk of hypotension when furosemide is added to an ACE‐inhibitor, only the other way around). Whether a clinically relevant signal is useful or unnecessary is arbitrary and depends on the knowledge of the prescribing physicians. It is not necessary to warn for a clinically relevant interaction that occurs daily, since that will only result in a lack of interest in the whole signalling system. Knowledge about potential interactions of the physicians will differ per department and the usefulness should be listed per department. Shifting the possible important signals out will increase the value of a medication database as a tool for detecting drug interactions. An inventory of useful signals needs to be completed per department.