Clinical-pharmaceutical assessment of medication CDSS alerts: content appropriateness and patient relevance in clinical practice.

Background: Clinical pharmacy services and clinical decision support systems (CDSSs) are increasingly implemented to optimize medication safety. However, risks as overalerting can limit these benefits. Therefore, the Meona medication CDSS was interprofessionally evaluated and locally configured prior to implementation at Erlangen University Hospital. Aim: We aimed to analyze the displayed CDSS alerts and to evaluate the content appropriateness and patient relevance of CDSS alerts in a hospital with established clinical ward pharmacists. Furthermore, we characterized pharmaceutical interventions triggered by CDSS and CDSS-independent interventions. Methods: Pseudonymized clinical data of 160 patients from four clinical departments were prospectively included once between days 1 and 3 after hospital admission to analyze the frequency, type, and severity of the displayed CDSS alerts. All severe and "duplicate prescription" CDSS alerts were evaluated regarding their content appropriateness and patient relevance by clinical pharmacists using the four-eyes principle. For patient-relevant CDSS alerts, clinical ward pharmacists intervened during weekly ward rounds. All pharmaceutical interventions, including CDSS-independent interventions, were documented in ADKA-DokuPIK by recording reason, acceptance rate, and severity. Results: In total, 1,799 CDSS alerts (median 9.0/patient) were displayed. Of those, 33.9% (609/1,799) were classified as severe by Meona. Clinical pharmacists validated 647 CDSS alerts (609 severe and 38 "duplicate prescriptions"). Only 82.7% (535/647) were rated as content appropriate, of which 19.6% (105/535) were classified as patient relevant. The clinical ward pharmacists recorded 244 interventions in 150 patients discussed during rounds (1.6/patient). CDSS-independent interventions by clinical ward pharmacists (158/244, 64.8%) were significantly more frequent compared to pharmaceutical interventions triggered by the CDSS (86/244, 35.2%). (p = 0.0002). The acceptance rate of interventions was 92.2% (225/244). The most common severity category was C (error occurred, no harm). Conclusion: Despite the locally customized medication CDSS, a high number of CDSS alerts were displayed. Interestingly, we still observed content-inappropriate CDSS alerts defined by pharmaceutical validation. The majority of CDSS alerts with appropriate content were rated not patient relevant in clinical practice and could be considered as overalerting. Our results highlight that a CDSS can support healthcare professionals but underline (1) the continuing need for clinical pharmacists to improve medication safety by interpreting CDSS alerts and performing comprehensive medication reviews and (2) the further need for CDSS improvements.