Evaluation of the efficacy of conventional medication surveillance and clinical rules in daily hospital practice

Abstract

Background: Alert fatigue is probably one of the biggest downsides of clinical decision support systems based on medication surveillance used in current daily (hospital) practice. Different studies suggest that only 1.4-7.5% of conventional medication surveillance signals require intervention by a pharmacist, while for clinical rules intervention rates of 80-94% are possible. The purpose of this study is to determine the current extent of alert fatigue by calculating the positive predictive value (PPV) for intervention for both types of medication surveillance in the hospital pharmacy of Maasstad Hospital.

Materials and methods: Data of January 2022, which contained all generated medication surveillance signals during the study period, were extracted from the electronic health records and analyzed retrospectively. The amount of interventions arising from these generated signals was determined. This allowed the PPV to be determined, as well as the amount of advices which were followed by the physician. Based on the data, suggestions for improvement have been made.

Results and conclusion: In January 2022, 15.016 conventional medication surveillance signals and 1.544 clinical rules have been generated. 0.9% of conventional medication surveillance signals resulted in consultation of a physician by a pharmacist, while 18.9% of clinical rules led to an intervention. For conventional medication surveillance signals 83 of 132 advices (62.9%) were followed, while for clinical rules 165/260 advices (63.5%) were followed by the physician. For conventional medication surveillance, 789 signals were identified which, in future could be disabled and 1.473 signals have been identified which could (partially) be managed by technicians in the future. Optimization of current clinical rules could prevent 279 irrelevant rules.

Conclusion: The PPV of conventional medication surveillance is very low and introduces the risk of alert fatigue as only 9 out of 1.000 signals lead to an intervention. Although the PPV of clinical rules is higher than the PPV of conventional medication surveillance in the current situation, it is still lower than desired. There is room for improvement of both types of medication surveillance.