gms | German Medical Science

Introduction: In the intensive care unit (ICU), alarms notify clinical staff of a patient’s deterioration, indicating that a medical intervention may be required. Clinicians take decisions to counteract alarms. Due to their proximity to patients, nurses are the first point of contact for ICU alarms. Analyzing alarm data can be a starting point for a more informed alarm management and patient care [1]. Specifically, alarm metrics and their visualizations can help identify measures to reduce unnecessary alarms - one of the main burdens of ICU staff today [2]. However, there is no comprehensive overview of the types of metrics used in ICU alarm research. Furthermore, it is unclear how to best visualize these metrics to provide meaningful insight for ICU staff and facilitate data-driven clinical decision-making.

State of the art: Today, alarms are represented singularly without further context. Thus, the reasons for alarm analysis can be basic analyses of the alarm situation or reducing the number of alarms. Nursing staff could use alarm analysis for more effective shift handovers. ICU teams could conduct quality of care reviews [1], e.g., for alarm reaction times. For technically savvy ICU staff, do-it-yourself instructions exist to analyze alarm log data themselves [3], e.g., to visualize the number of technical alarms versus vital sign alarms. A focus group study for pediatric alarm management identified that alarm metrics should be categorized, contextualized, and discussed with an interprofessional team [4].

Concept: We want to define a framework for alarm metrics and visualizations that best support end users’ (nurses, physicians, data scientists) needs. We will measure how different end-users obtain insights from ICU alarm metric visualizations and how this affects decision-making, creating a benchmark for the future of ICU alarm management. More intuitive visualizations of alarm metrics need to be created and evaluated – for the purpose of enabling medical staff to make more informed decisions for alarm management.

Implementation: Currently, we systematically review how scientific literature presents alarm analyses, what kind of metrics and visualizations are used, and what factors may have influenced the type of analyses. We will cluster the findings by field of application and research question. Next, via an insight-based methodology [5], we will evaluate their gain in knowledge and its effect on decision-making. End-user insights will be quantified, e.g., observations from visualizations, how long it takes to reach an insight, perceived importance of an insight, or if hypotheses for alarm reduction can be derived.

Lessons learned: Summary metrics may be used to increase the active participation of medical staff in the improvement of alarm management [4]. A data-driven understanding of the alarm situation could lead to evidence-based alarm reduction strategies. It is likely that there is no single way of depicting multidimensional clinical data [5]. For ICU alarm management, this means that different types of visualizations need to be evaluated on how informative they are for different end-users. A framework for ICU alarm research will create a common basis to compare results across institutions – and finally, to improve care delivery for patients and medical staff. It can be a starting point for the development of alarm dashboards.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.