gms | German Medical Science

Introduction: Alarm management in intensive care units (ICUs) is influenced by a variety of factors. One major inhibitor for efficient utilization of ICU alarms is alarm fatigue: clinicians are bombarded with numerous alarms generated by patient monitors, ventilators, and other medical devices, almost all of them without clinical actionability [1], [2]. This can lead to desensitization and decreased responsiveness to critical alarms [3]. However, how to improve the specificity of alarms with technological and staff-based alarm management systems is still up for debate. By examining alarm data from various patient groups for the use case of ventilation therapy - a major source of ICU alarms [4] - we sought to determine the potential of data science to inform, guide, and enhance alarm management efforts.

Methods: Following IRB approval (Ethics vote no. EA1/127/18) we retrospectively extracted alarm data from ICU patient monitoring devices; ventilation therapy information from the electronic health record, and clinical interventions after an alarm for 15 ICUs of a tertiary care center between 07/2019 and 06/2021. Based on a set of annotation guidelines [5] we grouped the ventilation therapy by invasiveness: unassisted spontaneous breathing, non-invasive and invasive; and the non-technical vital sign alarms as clinically actionable or non-actionable based on whether clinical interventions were performed after the alarm.

Results: We extracted 9,248,889 vital sign alarms from 32,912 patients and 1,321,654 documented ventilation therapy changes. Patients receiving invasive ventilation therapy produce less peripheral oxygen saturation (SpO₂), heart rate, and technical alarms but more blood pressure alarms when compared to patients with unassisted spontaneous breathing (Figure 1 [Fig. 1]). Blood pressure alarms were also the only alarm type where patients with non-invasive ventilation therapy produced the most alarms. For all vital sign alarms, the invasive ventilation therapy group had the highest percentage of clinically actionable alarms, whereas the ventilation therapy group that generated the most overall alarms had the lowest actionable percentage. The absolute number of clinically actionable heart rate alarms was comparable across all three ventilation groups. For SpO₂ alarms, that number only decreased in the invasive ventilation group.

Discussion: These results suggest that the invasiveness of ventilation therapy is an important factor in determining the number of alarms in ICU patients. Clinicians should be aware of the increases and decreases associated with ventilation therapy, especially because the percentage of clinically actionable alarms does not always match the alarm rate. However, there are confounding factors that we didn't analyze but should be considered in future studies, like disease severity or patient characteristics.?????

Conclusion: Our study highlights the relationship between ventilation therapy and alarm rates in ICU patients and emphasizes the need for a more data-driven approach to alarm management in the ICU setting. By analyzing the complex relationships between patient characteristics, therapy types, and the number of alarms and their actionability, clinicians could adequately adjust ventilator alarm setting, and develop more effective, targeted alarm management strategies. This study provides a foundation for future research and highlights the potential for data science to improve patient outcomes and staff well-being in the ICU.

Grant: BMBF 16SV8559

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.