Background: Physiologic monitoring is a mainstay of neonatal intensive care. Patient monitoring devices rapidly detect fluctuations in vital signs, triggering alarms to alert providers to changes in patient status. While integral to clinical monitoring, these systems also generate an abundance of technical alarms and non-critical physiologic alarms. The resultant excess alarm burden risks delayed response to critical alarms, a serious patient safety hazard known as alarm fatigue.
Objective: Our study set out to quantify the audible burden of patient monitoring alarms in a Level-IIIB Neonatal Intensive Care Unit (NICU) at Beth Israel Deaconess Medical Center in Boston, MA. We also set out to characterize alarm silencing patterns among NICU providers and identify sources of disproportionate contribution to alarm burden.
Design: Our population consisted of all patients undergoing physiologic monitoring in the BIDMC Level-IIIB NICU between January 2018 and December 2019. Physiologic trend data and alarm data were captured by Philips IntelliVue patient monitors and archived using Data Warehouse Connect (DWC) technology. We leveraged alarm descriptors in the DWC database to classify each alarm according to its priority, type, and source. To estimate the duration of each alarm for which audible indicators were emitted, we extracted time intervals in which alarms were silenced or paused. We also used timestamp data to analyze providers' response times in silencing alarms, as a function of alarm type and physical location within the unit.
Results: We found that 5,750,033 alarms occurred over the study period, an average daily burden of 7,877 alarms. Among these alarms, 2,958,818 (51.5%) were technical alarms. Eight alarm types contributed most significantly to audible alarm burden, among them both advisory- and technical-level heart rate and oxygen saturation alarms. In response, providers silenced 1,713,282 (29.8%) alarms and utilized the pausing feature on bedside monitors 493,196 times. Providers' response times in silencing alarms varied according to alarm priority, with critical physiologic alarms silenced most rapidly and technical alarms most slowly. Silencing occurred most rapidly in patient rooms near the center of the unit, where critical and non-critical physiologic alarms were most concentrated in accordance with patients' clinical acuity.
Conclusion: Our work supports previous findings that NICU alarm burden is driven by an abundance of technical alarms and non-critical physiologic alarms, which may distract providers from rarer, more life-threatening physiologic alarms. Having identified several sources of excess audible alarm burden, we plan to develop and implement quality improvement interventions to reduce non-actionable alarms and measure their impact on alarm burden.
