Electronic medical devices are an integral part of patient care, providing vital life support and physiologic monitoring that improve safety throughout hospital care units. The alarms and alerts generated by such devices are intended to warn clinicians about any deviation of physiological parameters from their normal values before a patient can be harmed. Life support devices (e.g., ventilators and cardiopulmonary bypass machines) also employ alarms to alert health care providers to potentially life-threatening failures. These two alarm types (i.e., physiologic and device function) lead to a high frequency of alarms in the clinical setting. For example, in one study of patients undergoing procedures, 8,975 alarms occurred during 25 consecutive procedures. An average of 359 alarms were recorded during each procedure, or approximately 1.2 alarms per minute.1 Equipment manufacturers deliberately set alarm defaults to high sensitivity, so that true events are not missed. The result is that most alarms have low specificity and low positive predictive value and are often ignored.2 This problem is compounded when alarms are implemented across multiple parameters, leading to a cascade of alarms that create a noisy, distracting environment while doing little to improve patient care.
Alarm fatigue refers to an increase in a health care provider’s response time or a decrease in his or her response rate to an alarm as a result of experiencing excessive alarms. Alarm fatigue is common in many professions (e.g., transportation and medicine) when signals activate so often that operators ignore or actively silence them. The organizational and technological aspects of the hospital environment are highly complex, and alarm fatigue has been implicated in medical accidents.2 The Joint Commission, recognizing the clinical significance of alarm fatigue, has therefore made clinical alarm management a National Patient Safety Goal. This article will provide an overview of signaling (alarms, alerts, and warnings) and offer practical solutions to reduce alarm fatigue in the operating room and intensive care unit.
False, Nonactionable, and Nuisance Alarms
Researchers have historically used signaling terms interchangeably, which can complicate attempts to understand and address the problems created by excessive alarms. Bliss and Gilson proposed an early taxonomy of signaling terms that accounts for the timing between a signal and its associated situation.3 They adopted the term “signal” as an umbrella term for all stimuli that serve the general function of emergency notification. This taxonomy defines an “alarm” as a transient sensory signal (usually auditory or visual) that indicates an ongoing danger that requires an immediate corrective action, while an “alert” indicates that an adverse event may occur in the future.3 For example, an alert may occur ten minutes before a patient is expected to deteriorate while an alarm might indicate asystole. Alerts give the operator more time to react, allowing a preemptive response that may allow the problem to be avoided, while a response to an alarm takes place when the danger exists and is reactive or corrective.
The current standard for medical alarms is IEC 60601 1-8, which specifies basic safety and performance requirements, including alarm categories that are prioritized by degree of urgency, and consistency of alarm signals.4 The IEC standard does not, however, address the problems associated with the high sensitivity of sensors and low specificity of alarm conditions. A valid alarm may give the health care provider very little time to react to a life-threatening event. In general, signals should ideally give a health care provider enough time to take an action that will prevent an adverse outcome. The duration of an appropriate time delay is, however, contingent upon operational parameters, most notably the rate at which the situation is expected to deteriorate.
Medical signals can be further subdivided according to their underlying condition. Clinical alarms indicate that the patient requires immediate attention, while technical alarms indicate that the biomedical equipment requires attention. For example, ventricular fibrillation results in a clinical alarm, while a disconnected sensor or a poor-quality blood pressure tracing might cause a technical alarm. Xiao and Seagull have proposed a taxonomy that distinguishes among alarms based on their usefulness for medical personnel who monitored clinical processes (Table 1).5
Table 1: Xiao and Seagull’s taxonomy of alarms:5
|False alarms occur when no danger exists, often because sensor thresholds are set too conservatively.|
|Nuisance alarms may indicate a problem in a specific context, but they have been activated in a different context (e.g., an arterial catheter low pressure alarm that activates when a blood pressure cuff is inflated).|
|Inopportune alarms occur at the wrong time, perhaps as alerts that signal a condition far in the future.|
Actionable alarms indicate a physiologically abnormal state, which requires that the anesthesia professional intervene in order to avoid patient harm.
A mild deviation might require only assessment of the patient and heightened alertness for further change, while others might indicate an urgent, life-threatening problem.6 Nonactionable alarms can be caused by monitoring artifact (e.g., electrocautery causing a “ventricular fibrillation” alarm), or a true deviation from the alarm limits that represents a clinically insignificant abnormality (e.g., a ventilator’s apnea alarm activating while the patient is being intubated).
Failure to respond to an alarm can cause patient harm and may potentially be life threatening. The United States Food and Drug Administration (FDA) reported over 500 alarm-related patient deaths during a five-year period, and many believe that this report significantly underestimates the magnitude of the problem.* The purpose of an alarm is to get the immediate attention of a person when an abnormal event occurs; alarms are therefore designed to be intrusive and distracting. Frequent interruptions from nonactionable alarms can degrade prospective memory, and there is evidence that improving the design of alarms and alerts can prevent errors.7 Health care providers may become desensitized to frequent false alarms; this is called the cry-wolf effect8 and is more likely to occur during periods of high workload.9 The cry-wolf effect may lead users to mistrust and possibly ignore subsequent alarms from the same or similar devices.
The intrusive nature of auditory alarms can increase the stress level during an abnormal event.10 In 2015, one of the authors (KJR) defined alarm flood as a large number of alarms, some of which may be in a different patient care area.11 Further, alarms can disrupt sleep and contribute to ICU delirium. Hall et al. measured the stress response to an “emergency” alarm that required the participants to immediately get dressed and walk briskly to a testing room. They found that the physiologic stress (as indicated by saliva cortisol level) caused by nighttime alarms was significantly greater than those that occurred during the day.12
Solutions: Simple and Complex
Alarm fatigue is a complex problem, and potential solutions include redesigning organizational aspects of unit environment and layout, workflow and process, and safety culture. Technical and engineering solutions, workload considerations, and practical changes to the ways in which existing technology is used can mitigate the effects of alarm fatigue. These changes will ultimately require new approaches to training, clinical workflow, and organizational policies.11 The overarching goals for a comprehensive solution to alarm fatigue should be to clearly and accurately indicate potential hazards while minimizing false or nuisance alarms. Signals should be consistent across all equipment used in the health care environment. Multiple factors, including noise, lighting, competing task demands, distrust, and inattentional blindness or deafness can prevent a health care provider from detecting or responding to an alarm. New equipment should incorporate designs that decrease a clinician’s workload and do not unnecessarily distract him or her from other time-critical tasks. Both increasing workload and high levels of ambient noise can impair subjects’ ability to localize alarms.13
Changes to the alarm processing algorithms of physiologic monitors can reduce the number of nonactionable alarms. Delaying alarm activation for short, clinically-irrelevant violations can improve alarm reliability. One study hypothesized that implementing a short alarm delay for minor threshold violations (which the researchers defined as a deviation less than 4% beyond the threshold) would inhibit alarms caused by brief, clinically irrelevant violations.14 The delay allowed the values to return to normal limits before the alarm was activated. Implementing this delay for alarms that transiently violated limits by a small amount resulted in a 74% reduction in false alarms.14 Srivastava et al. used a machine learning algorithm to simultaneously analyze the electrocardiogram, pulse oximetry, and arterial blood pressure waveforms. Their model was able to suppress 77% of false alarms while improving alarm accuracy to 84%.15 These studies and others highlight the opportunities for medical equipment manufacturers to develop innovative algorithms to increase the positive predictive value of clinical alarms.
Reducing alarm volume can alleviate the level of noise pollution in the operating room and intensive care unit. Conventional wisdom suggests that alarms should be as loud as possible to immediately attract the attention of the operator. In one recent study, however, Schlesinger et al. found that physicians who were required to respond to simulated critical events while completing an auditory speech intelligibility test were able to distinguish alarms even when they were -11 dB below the ambient noise level.16 This could reflect the expertise level of the operators and suggests that it might be possible to reduce alarm volumes and thereby the overall noise level in health care institutions. Although alarms must be audible, this study suggests that reducing volume might be possible, especially for alarms that do not indicate a life-threatening condition. Strategies for doing so should be considered jointly with manipulations of signal wave form, intertemporal interval, and other physical parameters.13
Some simple interventions can be used immediately by nearly any clinician. Clinicians should choose appropriate alarm limits for each patient. Shanmugham et al. found that perceived workload was lower when alarm settings were modified to reflect an individual patient’s physiologic status as compared to an unmodified default clinical alarm setting.17 The simple step of changing clinical alarm limits and disabling nonessential alarms improved the accuracy of alarm response, participants’ experience, and overall satisfaction. A simple way to accomplish this goal is to use specific profiles when available (e.g., use pediatric defaults when caring for a child and use the “paced” mode when a patient has a pacemaker or implantable cardiac debrillator device). Disposable sensors may also be responsible for false alarms caused by artifact, especially when they are repositioned or allowed to dry. A sensor or cable that is not compatible with the monitor in use and electrodes with dried gel or adhesive may also trigger false alarms. A simple solution is to use new electrodes and to replace them rather than attempting to reuse them if they must be moved. Over-monitoring can also increase the number of alarms to which a clinician is exposed. The level of monitoring should therefore be selected to suit the needs of the individual patient.11,18
Alarm fatigue is a multifaceted problem with multiple contributing factors, including false alarms, and nonactionable alarms. Most alarms are triggered when the value of a given parameter violates a preset threshold that is frequently set in anticipation that vital signs that are normal for a given patient will fall within a narrow, predicted range. Although this philosophy might work well when monitoring a single parameter with a well-defined normal range (e.g., oxygen saturation), it can also result in a significant number of false alarms when monitoring patients with multiple comorbidities in an actual clinical environment. Medical equipment manufacturers can help to solve this problem by developing innovative alarm processing algorithms. Clinicians can also make simple changes to their practice that will help to mitigate the effects of alarm fatigue.
Dr. Ruskin is professor of Anesthesia and Critical Care at the University of Chicago, Chicago, IL.
Dr. Bliss is professor and associate chair, Psychology at Old Dominion University, Norfolk, VA.
Neither author has any conflict of interest pertaining to this article.
- Schmid F, Goepfert MS, Kuhnt D, et al. The wolf is crying in the operating room: patient monitor and anesthesia workstation alarming patterns during cardiac surgery. Anesth Analg. 2011;112:78–83.
- Cvach M. Monitor alarm fatigue: an integrative review. Biomed Instrum Technol. 2012;46:268–77.
- Bliss JP, Gilson RD, Deaton JE. Human probability matching behaviour in response to alarms of varying reliability. Ergonomics. 1995;38:2300–12.
- Medical Electrical Equipment: General requirements for basic safety and essential performance. 2015. https://www.iso.org/standard/65529.html. Accessed December 1, 2018.
- Xiao Y, Seagull FJ. An analysis of problems with auditory alarms: defining the roles of alarms in process monitoring tasks. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 1999;43:256–60.
- Karnik A, Bonafide CP. A framework for reducing alarm fatigue on pediatric inpatient units. Hosp Pediatr. 2015;5:160–3.
- Loft S, Smith RE, Remington RW. Minimizing the disruptive effects of prospective memory in simulated air traffic control. J Exp Psychol Appl. 2013;19:254–65.
- Breznitz S. Cry wolf : the psychology of false alarms. Hillsdale, N.J.: Lawrence Erlbaum Associates; 1984.
- Bliss JP, Dunn MC. Behavioural implications of alarm mistrust as a function of task workload. Ergonomics. 2000;43:1283–300.
- Peryer G, Noyes J, Pleydell-Pearce K, Lieven N. Auditory alert characteristics: a survey of pilot views. Int J Aviat Psychol. 2005;15:233–50.
- Ruskin KJ, Hueske-Kraus D. Alarm fatigue: impacts on patient safety. Curr Opin Anaesthesiol. 2015;28:685–90.
- Hall SJ, Aisbett B, Tait JL, et al. The acute physiological stress response to an emergency alarm and mobilization during the day and at night. Noise Health. 2016;18:150–6.
- Edworthy J, Reid S, Peel K, et al. The impact of workload on the ability to localize audible alarms. Appl Ergon. 2018;72:88–93.
- Schmid F, Goepfert MS, Franz F, et al. Reduction of clinically irrelevant alarms in patient monitoring by adaptive time delays. J Clin Monit Comput. 2017;31:213–9.
- Srivastava C, Sharma S, Jalali A. A novel algorithm for reducing false arrhythmia alarms in intensive care units. Conf Proc IEEE Eng Med Biol Soc. 2016;2525–8.
- Schlesinger JJ, Baum Miller SH, Nash K, et al. Acoustic features of auditory medical alarms—An experimental study of alarm volume. J Acoust Soc Am. 2018;143:3688.
- Shanmugham M, Strawderman L, Babski-Reeves K, et al. Alarm-related workload in default and modified alarm settings and the relationship between alarm workload, alarm response rate, and care provider experience: quantification and comparison study. JMIR Hum Factors. 2018;5:e11704.
- Paine CW, Goel VV, Ely E, et al. Systematic review of physiologic monitor alarm characteristics and pragmatic interventions to reduce alarm frequency. J Hosp Med. 2016;11:136–44.
*The Joint Commission Sentinel Event Alert. Medical device alarm safety in hospitals. http://www.jointcommission.org/assets/1/18/SEA_50_alarms_4_5_13_FINAL1.PDF