ABSTRACT
PURPOSE: The comparative evaluation of clinical decision support software (CDSS) programs regarding their sensitivity and positive predictive value for the identification of clinically relevant drug interactions. METHODS: In this research, we used a cross-sectional study that identified potential drug interactions using the CDSS MediQ and the ID PHARMA CHECK in 484 neurological inpatients. Interactions were reclassified according to the Zurich Interaction System, a multidimensional classification that incorporates the Operational Classification of Drug Interactions. RESULTS: In 484 patients with 2812 prescriptions, MediQ and ID PHARMA CHECK generated a total of 1759 and 1082 alerts, respectively. MediQ identified 658 unique potentially interacting combinations, 8 classified as "high danger," 164 as "average danger," and 486 as "low danger." ID PHARMA CHECK detected 336 combinations assigned to one or several of 12 risk and management categories. Altogether, both CDSS issued alerts relating to 808 unique potentially interacting combinations. According to the Zurich Interaction System, 6 of these were contraindicated, 25 were provisionally contraindicated, 190 carried a conditional risk, and 587 had a minimal risk of adverse events. The positive predictive value for alerts having at least a conditional risk was 0.24 for MediQ and 0.48 for ID PHARMA CHECK. CONCLUSIONS: CDSS showed major differences in the identification and grading of interactions, and many interactions were only identified by one of the two CDSS. For both programs, only a small proportion of all identified interactions appeared clinically relevant, and the selected display of alerts that imply management changes is a key issue in the further development and local setup of such programs.
Subject(s)
Central Nervous System Agents/adverse effects , Decision Support Systems, Clinical/organization & administration , Inpatients , Mass Screening/methods , Cross-Sectional Studies , Drug Interactions , Humans , Risk AssessmentABSTRACT
PURPOSE: The current study aimed at identifying and quantifying critical drug interactions in neurological inpatients using clinical decision support software (CDSS). Reclassification of interactions with a focus on clinical management aimed to support the development of CDSS with higher efficacy to reduce overalerting and improve medication safety in clinical practice. METHODS: We conducted a cross-sectional study in consecutive patients admitted to the neurology ward of a tertiary care hospital. We developed a customized interface for mass analysis with the CDSS MediQ, which we used for automated retrospective identification of drug interactions during the first day of hospitalization. Interactions were reclassified according to the Zurich Interaction System (ZHIAS), which incorporates the Operational Classification of Drug Interactions (ORCA). Dose adjustments for renal impairment were also evaluated. RESULTS: In 484 patients with 2812 prescriptions, MediQ generated 8 "high danger," 518 "average danger," and 1233 "low danger" interaction alerts. According to ZHIAS, 6 alerts involved contraindicated and 33 alerts involved provisionally contraindicated combinations, and 327 alerts involved a conditional and 1393 alerts involved a minimal risk of adverse outcomes. Thirty-five patients (6.2%) had at least one combination that was at least provisionally contraindicated. ZHIAS also provides categorical information on expected adverse outcomes and management recommendations, which are presented in detail. We identified 13 prescriptions without recommended dose adjustment for impaired renal function. CONCLUSIONS: MediQ detected a large number of drug interactions with variable clinical relevance in neurological inpatients. ZHIAS supports the selection of those interactions that require active management, and the effects of its implementation into CDSS on medication safety should be evaluated in future prospective studies.
