Alarm fatigue: when the safety feature becomes the hazard

An alarm is a safety feature, bought to extend a nurse's surveillance past the bedside. The evidence in this unit is that the feature has become a hazard in its own right. One intensive care study recorded 187 audible alarms per bed per day, most of them non-actionable (Drew et al., 2014).

The claim is narrow. It is not that clinicians are careless with alarms: device design and settings produce the load, the desensitisation is measurable against exposure, and the interventions tried change the count more reliably than the outcome. This fifth of ten units turns from the limits of the operator to a designed system that harms the people it protects.

A hazard that reports the patient's noise as well as the patient's danger

Cvach's integrative review of 72 articles published between 2000 and 2011 found alarm fatigue rated the number one medical device technology hazard of that year, and traced it to a high false alarm rate and to the number of alarming devices (Cvach, 2012). Sendelbach and Funk report that 72 to 99 per cent of clinical alarms are false, define alarm fatigue as sensory overload from excessive alarms leading to desensitisation and missed alarms, and record deaths attributed to it (Sendelbach and Funk, 2013).

Funk and colleagues surveyed hospital personnel in 2005 and 2006 and again in 2011, and in both rounds respondents ranked frequent false alarms the most important of nine alarm issues (Funk et al., 2014). The survey is self-report from staff who chose to answer, so it measures what they believe about alarms rather than what the alarms did.

The load is a property of the configuration

Drew and colleagues measured it. They stored all monitor data for 461 adults over 31 days, capturing 2,558,760 unique alarms, of which 381,560 were audible, a burden of 187 audible alarms per bed per day (Drew et al., 2014). Annotators judged 88.8 per cent of 12,671 arrhythmia alarms false positives, and 93 per cent of the 168 true ventricular tachycardia alarms too brief to warrant treatment. The excess is attributed to settings, patient conditions and algorithm deficiencies, all of them configurable.

Desensitisation is real, and it is measured as response time

Bonafide and colleagues video-recorded 36 nurses for 210 hours in a children's hospital, capturing 5,070 alarms, of which 87.1 per cent in intensive care and 99.0 per cent on the ward were non-actionable (Bonafide et al., 2015). Response time to alarms raised while no clinician was in the room lengthened with the number of non-actionable alarms in the preceding 120 minutes, at log-rank P under 0.001 in intensive care and 0.009 on the ward.

Dursun Ergezen and Kol observed 13 nurses for 328 hours in an adult intensive care unit, recording 1,781 alarms (Dursun Ergezen and Kol, 2020). Nurses answered about 46.9 per cent of those requiring a response, and the response rate fell as the false alarm rate rose; Gazarian's observation of nine nurses in non-critical care found a response to 46.8 per cent of alarms and no dysrhythmia adverse event (Gazarian, 2014).

The interventions move the count, not the outcome

Sowan and colleagues changed 17 default monitor parameters and gave in-service education in a 20-bed transplant and cardiac intensive care unit with 39 nurses (Sowan et al., 2016). The alarm rate fell from 87.86 to 59.18 per patient day. Nurses' attitudes did not change measurably, and half of the 24 nurses completing both surveys still wanted more training. The education also assumes a reader whose first language is the ward's.

Nyarko and colleagues reviewed 13 studies and found that most reduced total and false alarms and improved nurses' knowledge, while the reduction in alarm fatigue itself remained uncertain (Nyarko et al., 2022). Gul and colleagues screened 11,632 studies and pooled 7, and concluded that the pooled effect size was not significant, so the interventions may not reduce alarm fatigue (Gul et al., 2024).

Winters and colleagues, for a critical care task force, found 62 articles with usable data (Winters et al., 2018). Most of the studies were successful, most were inseparable bundles, and none measured alarm fatigue directly. That gap has since narrowed: Wunderlich and colleagues tested the nine-item Charite Alarm Fatigue Questionnaire on two German intensive care samples, 265 and then 1,212, and confirmatory factor analysis supported its alarm stress and coping scales (Wunderlich et al., 2024).

The signal was mis-specified upstream of the ward

The audible alarms associated with IEC 60601-1-8, the global standard for medical device alarm sounds, are difficult to learn and retain. Edworthy and colleagues designed four alternative sets of eight alarms matched to the functions the standard specifies, and tested them for recognisability and localisability; every set but one outperformed the alarms currently specified (Edworthy et al., 2017). The standard requires sound, so a clinician with hearing loss depends on the ward's other equipment, and the authors present their work as the benchmarking phase of replacing that standard.

The contested question: count the alarms, price their information, or admit that nothing was measured?

The informativeness position, argued by Rayo and Moffatt-Bruce, holds that the alarm count is not what drives response (Rayo and Moffatt-Bruce, 2015). The contributor most frequently associated with response is informativeness, the proportion of alarms conveying a specific event. It is low across healthcare, consistently between 1 and 20 per cent.

Bonafide and colleagues found response time lengthening with non-actionable alarms in the preceding two hours, and Dursun Ergezen and Kol found response rates falling as the false alarm rate rose (Bonafide et al., 2015; Dursun Ergezen and Kol, 2020). The measurement position is put most explicitly by Winters and colleagues, who deny that either quantity has been tied to the construct: no included study defined alarm fatigue or measured it (Winters et al., 2018), though an instrument now exists (Wunderlich et al., 2024). Sowan's before and after data supports that position, since cutting the alarm rate by about a third left attitudes where they were (Sowan et al., 2016).

What would settle it is a three-armed trial, carrying an informativeness intervention, a volume intervention and a direct metric for alarm fatigue, judged against response time to actionable alarms and the number missed. No such study is in this unit's literature.

What to do while the dispute is open

The evidence on the two sides is unequal. That informativeness drives response rests on an argument from the guidance literature. That exposure drives response rests on two observational studies with a dose-response shape and one before-and-after project. The strongest objection to this unit's thesis is that the interventions measured no alarm-fatigue outcome, so the claim that the feature is the hazard rests on proxies. While the dispute is open, a designer should measure informativeness and the response to actionable alarms, and treat a fall in the count as a result to be justified, not one achieved.

Where this is heading

Winters and colleagues named a validated alarm-fatigue metric as the investment priority, and Wunderlich and colleagues have published one (Winters et al., 2018; Wunderlich et al., 2024). Rayo and Moffatt-Bruce offer informativeness as the signal to measure (Rayo and Moffatt-Bruce, 2015), and Edworthy and colleagues have completed the benchmarking phase of replacing the sounds a global standard specifies (Edworthy et al., 2017).

Sources

• Cvach, M. (2012). Monitor Alarm Fatigue: An Integrative Review. Biomedical Instrumentation and Technology. doi:10.2345/0899-8205-46.4.268 Synthesises 72 articles published between 2000 and 2011 and records alarm fatigue as the number one medical device technology hazard in 2012. The causes it lists, a high false alarm rate, poor positive predictive value, missing standardisation and the number of alarming devices, are properties of the configuration.
• Sendelbach, S. and Funk, M. (2013). Alarm fatigue: a patient safety concern. AACN Advanced Critical Care. doi:10.4037/nci.0b013e3182a903f9 Reports that 72 to 99 per cent of clinical alarms are false, defines alarm fatigue as sensory overload from excessive alarms leading to desensitisation and missed alarms, and records deaths attributed to it. Its closing recommendation carries the thesis: strategies that reduce false alarms still need trials showing that they cut alarm burden without compromising safety.
• Drew, B. J., Harris, P., Zegre-Hemsey, J. K., Mammone, T., Schindler, D., Salas-Boni, R., Bai, Y., Tinoco, A., Ding, Q. and Hu, X. (2014). Insights into the Problem of Alarm Fatigue with Physiologic Monitor Devices. PLoS ONE. doi:10.1371/journal.pone.0110274 Stored all monitor data for 461 adults over 31 days: 2,558,760 unique alarms, 381,560 audible, 187 audible alarms per bed per day, and 88.8 per cent of 12,671 annotated arrhythmia alarms false positives. The excess is attributed to settings, patient conditions and algorithm deficiencies; the study was funded by a monitor manufacturer.
• Bonafide, C. P., Lin, R., Zander, M., Graham, C. S., Paine, C. W., Rock, W., Rich, A., Roberts, K. E., Fortino, M., Nadkarni, V. M., Localio, A. R. and Keren, R. (2015). Association between exposure to nonactionable physiologic monitor alarms and response time in a children's hospital. Journal of Hospital Medicine. doi:10.1002/jhm.2331 Video-recorded 36 nurses for 210 hours, observing 5,070 alarms, of which 87.1 per cent in intensive care and 99.0 per cent on the ward were non-actionable. Response time to alarms raised while no clinician was in the room rose with non-actionable alarms in the preceding 120 minutes, at log-rank P under 0.001 and 0.009. It measures exposure against response, not fatigue itself.
• Dursun Ergezen, F. and Kol, E. (2020). Nurses' responses to monitor alarms in an intensive care unit: An observational study. Intensive and Critical Care Nursing. doi:10.1016/j.iccn.2020.102845 Observed 13 registered nurses for 328 hours in an adult intensive care unit, recording 1,781 alarms: blood pressure, respiration and oxygen saturation, and arrhythmia accounted for 37.6, 35.3 and 27.1 per cent. Nurses answered about 46.9 per cent of the alarms requiring a response, and the response rate fell as the false alarm rate rose.
• Gazarian, P. K. (2014). Nurses' response to frequency and types of electrocardiography alarms in a non-critical care setting. International Journal of Nursing Studies. doi:10.1016/j.ijnurstu.2013.05.014 Observed nine registered nurses caring for patients on continuous electrocardiography, each for two three-hour periods. Nurses responded to 46.8 per cent of alarms and no dysrhythmia adverse event occurred, so a low response rate is ordinary on the ward rather than evidence of harm.
• Rayo, M. F. and Moffatt-Bruce, S. D. (2015). Alarm system management: evidence-based guidance encouraging direct measurement of informativeness to improve alarm response. BMJ Quality and Safety. doi:10.1136/bmjqs-2014-003373 Argues that many supposed drivers of alarm response, including total alarm rate, are unsupported, while informativeness, the proportion of alarms conveying a specific event, is most frequently associated with response. Informativeness is low across healthcare, consistently between 1 and 20 per cent, and should be measured first.
• Sowan, A. K., Gomez, T. M., Tarriela, A. F., Reed, C. C. and Paper, B. M. (2016). Changes in Default Alarm Settings and Standard In-Service are Insufficient to Improve Alarm Fatigue in an Intensive Care Unit: A Pilot Project. JMIR Human Factors. doi:10.2196/humanfactors.5098 Changed 17 default cardiac monitor parameters and gave in-service education in a 20-bed transplant and cardiac intensive care unit with 39 nurses, comparing ten weeks before and after. The alarm rate fell from 87.86 to 59.18 per patient day, attitudes did not change measurably, and half of the 24 nurses completing both surveys still asked for training.
• Edworthy, J., Reid, S., McDougall, S., Edworthy, J., Hall, S., Bennett, D., Khan, J. and Pye, E. (2017). The Recognizability and Localizability of Auditory Alarms: Setting Global Medical Device Standards. Human Factors. doi:10.1177/0018720817712004 Designed four sets of eight audible alarms matching the functions specified in IEC 60601-1-8 and tested them for recognisability and localisability against the alarms the standard specifies. Every set but one outperformed the current alarms, so the mis-specification sits in a global standard, upstream of the ward.
• Winters, B. D., Cvach, M. M., Bonafide, C. P., Hu, X., Konkani, A., O'Connor, M. F., Rothschild, J. M., Selby, N. M., Pelter, M. M., McLean, B. and Kane-Gill, S. L. (2018). Technological Distractions (Part 2): A Summary of Approaches to Manage Clinical Alarms With Intent to Reduce Alarm Fatigue. Critical Care Medicine. doi:10.1097/ccm.0000000000002803 Systematic review for a critical care task force that found 62 articles with usable data. Most studies reduced total or false alarms, most were inseparable bundles, and none measured alarm fatigue directly or defined it, so the field has measures but no measure.
• Wunderlich, M. M., Krampe, H., Fuest, K., et al. (2024). Evaluating the Construct Validity of the Charite Alarm Fatigue Questionnaire using Confirmatory Factor Analysis. JMIR Human Factors. doi:10.2196/57658 Tested a nine-item alarm-fatigue questionnaire on two German intensive care samples, 265 and then 1,212. Confirmatory factor analysis supported its alarm stress and coping scales and the scores correlated moderately with self-reported fatigue, so the construct can now be measured, though the coping scale was weaker and no outcome was recorded.
• Nyarko, B. A., Nie, H., Yin, Z., Chai, X. and Yue, L. (2022). The effect of educational interventions in managing nurses' alarm fatigue: An integrative review. Journal of Clinical Nursing. doi:10.1111/jocn.16479 Reviews 13 studies of educational interventions from five databases covering 2016 to 2021. Most found that education reduced total and false alarms and improved nurses' knowledge, while the reduction in alarm fatigue itself remained uncertain. The remedy improves every outcome except the one in its title.
• Gul, G., Seren Intepeler, S. and Bektas, M. (2024). Effectiveness of interventions developed to reduce alarm fatigue in intensive care units: systematic review and meta-analysis study. Fatigue: Biomedicine, Health and Behavior. doi:10.1080/21641846.2024.2332071 Screened 11,632 studies, included 19 in a systematic review and 7 in a meta-analysis of interventions to reduce alarm fatigue in intensive care units. The pooled effect size was not significant, so the authors conclude that the interventions may not reduce alarm fatigue. Their abstract describes a p value above 0.05 as significant in one sentence and non-significant in the next, so the conclusion is the more reliable half.
• Funk, M., Clark, J. T., Bauld, T. J., Ott, J. C. and Coss, P. (2014). Attitudes and Practices Related to Clinical Alarms. American Journal of Critical Care. doi:10.4037/ajcc2014315 Surveyed hospital personnel in 2005 to 2006 and again in 2011. Respondents ranked frequent false alarms the most important alarm issue in both years, and 18 per cent in 2011 reported patient adverse events related to alarms at their institution.