Alarm FatigueAlarm fatigue is a multifaceted problem, and its presence is influenced by a variety of factors. Preliminary investigation into the intended use environment of a medical device is a crucial step in identifying key desirable, and undesirable features – including auditory cues and warnings.

An inherent functionality of acute and critical care monitoring devices are their alarms – triggered by the device in response to an undesirable change in a patient parameter, to get the attention of a clinician.  An undesirable artifact of this system is the percentage of alarms that are non-actionable. The alarm was triggered due to an event not associated with a compromised patient parameter (i.e. patient movement, sensor displacement, incorrect reading), and therefore required no clinician intervention.

Within an ICU or acute care facility, there are alarms consistently ringing for everything from a loose or detached Sp02 finger probe, to a critically low respiratory rate. Non-actionable (false) alarms happen so often, and sound so similar to critical alarms, that clinicians and healthcare staff become desensitized to this auditory stimulation. This is coined alarm fatigue.

Factor 1: Volume & Tone

There is no synonymous sound, volume or tone used across the medical device industry for a particular type of alarm. Standards outline the accepted ranges for these auditory variables, but there is still room for interpretation of these guidelines between medical device manufacturers. What sounds like the alarm of a low battery on an infusion pump may sound strikingly similar, if not identical, to the bradycardia (abnormally slow heart beat) alarm on an ECG monitoring system. To further this variability, hospital departments always have devices from a range of device manufacturers. So even if there is a standardized alarm chosen for one device brand, there is no coordination with the other products to which it is combined.

One of the 10 human factors heuristic guidelines outlines the importance of the match between the system and the real world. Ideally, the system should be capable of speaking the user’s language with words and concepts familiar to the user, and the system should follow a real world, natural convention. Similarly, as outlined within the ANSI/AAMI HE75 Human Factors guidelines, the signal itself should be compatible with “human perceptual and cognitive abilities” [1].

With respect to auditory cues, the device should use an auditory cue that is intrinsically recognizable to the user, and that triggers an appropriate intensity and level of reaction. Argumentatively, perhaps warnings such as parameter alarms are better suited to an entirely different form of feedback, through tactile (a buzz or vibration) or visual (a triggered light or change in colour) methods.

Procedural investigation can provide insight into the current tones used by other device manufacturers, which auditory and tactile feedback has proven beneficial to users, and which have proven detrimental to device use, and ultimately patient safety. Collecting first hand observations and experiences provides insight required for design considerations that define the usability of the user interface.

Factor 2: Non – Actionable Alarms

Non-actionable alarms are essentially false alarms, which require no clinician intervention. As outlined within ANSI/AAMI HE75, the ideal alarm should be one which generates the appropriate alarm signal in response to the correct alarm condition, but never generates an alarm signal when the alarm condition does not exist [1]. Currently medical device manufacturers favour device designs that more heavily identify false positives, leading to non-actionable alarms and a work environment that inconveniences clinicians with an overload of auditory stimulation.

Procedural investigation can help mitigate this factor by providing designers and engineers the opportunity to identify how false alarms are currently being triggered. Identifying these problems during the preliminary investigation phase of the design process encourages more effort and time spent mitigating these use errors through design development. A firsthand experience of the falsely triggered alarm warnings may identify environmental artifacts that can be corrected, but were previously unidentified. Alternatively, observing how these alarms are triggered during use may spark creativity in identifying this patient parameter through a different, more consistent way.

Factor 3: Alarm Range Management

Another variable associated with alarm fatigue is the front-end alarm management, dependent on individual clinical practices. Parametric alarms included within physiological monitoring systems have the opportunity to be adjusted and personalized through alarm ranges, and set on a per-patient basis. This variable adds use and system errors associated with different workflow environments.

Procedural observation helps identify critical variabilities within clinical workflows which could cause confusion or frustration to the user, and ultimately misuse of the alarm function. Currently, alarm ranges are programmed to be a minimum value (x) to a maximum value (y). The clinician chooses which range best suits the respective patient. Instead of a static range which is set by the clinician based on professional judgement and experience, perhaps a range that is recommended based on known patient characteristics would be more effective.

Imagine the power of a system that takes in patient characteristics, including illness type or injury, compares them to documented cases, and recommends an alarm range tailored to the patient’s needs. Looking deeper, machine learning could be used to adapt this dynamic alarm range following changing patterns within patient parameters. Imagine a device that could identify a patient on the mends, and gently relax alarm ranges as the patient grew stronger. Contrastingly, a device notices a patient in declining health and constricts certain alarm parameters to ensure the quality of monitoring intensifies as the patient’s health declines.

Concepts such as these provoke innovation within product development and design. The center point of such notions are end users and patients, whose interests and behaviours are most identifiable through procedural observation.

[1] American National Standard, ANSI/AAMI HE75:2009/(R)2013, Human Factors Engineering – Design of Medical Devices. Alarm Design p205 – p207.

Image credit: Creative Commons

Hannah Rusak-Gillrie is a Human Factors Engineering Co-op at StarFish Medical. This is her first StarFish blog. Hannah is a 4th/5th year Biomedical Engineering student at UVic who enjoys hiking, cooking and being a proud plant mom!



One response to “Alarm Fatigue – The Alarm that Cried Wolf”

  1. Martin Poulin says:

    Nice article Hannah.

Leave a Reply

Your email address will not be published. Required fields are marked *