The future of sound in patient monitoring: a call to action to embrace the evidence and move forward

Summary

Patient monitoring systems have relied on beeping alarm sounds for the past 70 years, yet evidence suggests that these conventional tones limit alarm effectiveness, informativeness, and user situation awareness. More advanced alarm modalities, such as auditory icons and voice alerts, offer more intuitive approaches, and were found to improve alarm detection and reduce caregiver workload. Made possible by adoptions of international standards, and supported by evidence, the future of patient monitoring alarm sounds might incorporate conventional alarm tones with auditory icons, voice alerts, and advanced visualisations to better support caregivers and reduce alarm fatigue in patients and caregivers. Sound examples for advanced alarm modalities, comparisons with conventional alarm sounds, and alarm sound examples from aviation are provided in an online supplementary video.

In 1957, Dr Karl William Edmark patented the idea of a patient monitor that flashed a light with each heartbeat and sounded an alarm if no heartbeat was detected. This ‘Heartbeat Indicator’ was one of the earliest monitors with an alarm function.¹ The fundamental principles behind this early technology remain in use today. Acoustical alarm systems, which operate within predefined numerical thresholds, signal significant changes in patients' conditions so that healthcare providers can intervene. All systems that endure so long must fundamentally work, and in the case of acoustical alarms, they do. However, the evidence has evolved, and there are clear indications that we are significantly limiting ourselves by not fully using the perceptive abilities of humans and bringing the benefits of advanced, scientifically proven, acoustical alarm modalities to care providers and patients. The International Electrotechnical Commission Standard IEC 60601-1-8 (International Electrotechnical Commission, Geneva, Switzerland) now recommends using advanced sounds in patient monitoring, including auditory icons and voice alerts.² Edworthy and colleagues'³ recent study in the British Journal of Anaesthesia adds another piece to the growing body of evidence supporting auditory icons.³ Here, we outline three modalities for future alarm sounds in patient monitoring and discuss the potential impact of these modalities on situation awareness. Audio examples are provided in an online supplementary video.

Potential future patient monitoring sound modalities

Refinement of existing alarms through sound engineering

Monitoring devices emit various sounds to convey information to healthcare providers. Routine monitor beeping continuously informs clinicians about operational status or minor deviations in patient parameters that do not require immediate action. In contrast, alarms are designed to alert clinicians to changes in a patient's condition that necessitate prompt attention. These alarms are typically more urgent in tone and might be accompanied by visual cues to ensure they capture immediate attention, such as yellow or red blinking lights. But alarms need not to be loud or annoying to achieve this goal. Foley and colleagues^4,5 have shown that quiet alarm sounds are perceivable in background noise, and that percussive sounds are significantly less annoying yet as easily detected as current tonal sounds with flat amplitude envelopes. Philips (Amsterdam, The Netherlands) developed new alarm sounds that fit this category. They engineered alarm sounds with SenSound (Boston, MA, USA), a human-centred sound design firm, to increase perceived sound appeal while maintaining functionality.⁶ A study evaluating the performance of these sounds in clinical practice is currently being conducted at the University Hospital Zurich, Zurich, Switzerland, to provide initial data on the clinical performance of human centred-designed alarm sounds.

Although these developments might represent a step forward, they were not designed to address the most critical limitation of conventional tone-based alarms: their similarity in sound and its consequences: they become difficult to learn, remember, and distinguish. Overlapping alarms cannot be resolved. Sanderson and colleagues,⁷ who evaluated the learnability of melodic alarms, found that alarms were challenging for participants to learn and distinguish accurately, leading to frequent misidentification. Considering that only a small fraction of these alarms result in care interventions, the conventional alarm design can be described as inefficient, although effective at the price of too many insignificant alarms and alarm fatigue.⁸

Adoption of auditory icons

Auditory icons, which are not yet commercially available, represent a paradigm shift in alarm design and efficiency.9, 10, 11 These tones mimic real-world phenomena to convey information in a way that is intuitive and easily understood, for example, the sound of slow, ‘normal’, or quick breathing to indicate respiratory rate. Research suggests that metaphor-based acoustic signals, aligned with natural environmental sounds and human perception, enhance alarm detection accuracy.^9,10

Edworthy and colleagues³ present a comparison between conventional tonal-based alarms and auditory icons in simulated clinical scenarios with participants exposed to a standardised distraction. Their study revealed large effect sizes, demonstrating the clear superiority of auditory icons in enabling faster and more accurate alarm identification in a scenario-based monitoring context. Limitations of auditory icons include the limited number of auditory icons that can be reliably learned and distinguished. Expanding their use, as generally with every new technology, will require finding intuitive ways to represent complex information and continuous evaluation of their performance.

Use of spoken language

Spoken language has proven to be an extremely effective method for conveying information to humans without requiring diversion of visual attention. Voice alerts provided 99% recognition accuracy, compared with 54% accuracy of auditory icons and an estimated 17.5% accuracy with conventional alarm sounds in a study by Roche and colleagues.¹⁰ In clinical settings, spoken alerts could enhance the situation awareness of clinicians and teams by delivering detailed information directly through language.^10,12 Limitations include language barriers. Research by Paterson and colleagues¹² has shown that spearcons, or compressed speech phrases, can convey unambiguous and diverse information faster and in a less frightening way than plain speech messages.

How do these technologies align with situation awareness?

Endsley13, 14, 15 has developed a model of situation awareness, which can be seen as the fundamental principle behind every decision we make as humans. Situation awareness is gained through three stages: (i) perceiving the relevant information around us (e.g. an alarm sound), (ii) interpreting what it means, and (iii) projecting what will likely happen in the very near future. Good situation awareness causes good decisions. High workload, stressors, and complex system design complicate the development of situation awareness.

Traditional alarm systems require clinicians to shift their attention to the monitor, silence the alarm, identify the cause, and assess its relevance. This process is disruptive and time-consuming. Although auditory icons and voice alerts can help by providing more intuitive cues, they do not eliminate the need for interaction with the monitor. Current auditory icons do not convey numerical values or direction of change, and voice alerts typically indicate only ‘high’ or ‘low’ rather than specific measurements.¹⁰ In aviation, voice alarms, as used, for example, in Ground Proximity Warning Systems, provide clear and immediate instructions such as a ‘whoop whoop’ sound to awake attention, followed by, for example, a ‘Sink rate, Sink rate’ voice alert of the cause and a ‘Pull up, Pull up’ instruction to help clear the situation.¹⁶ Examples are given in an online supplementary video. Similarly, studies by Paterson and colleagues¹⁷ have shown that encoding more specific information, such as oxygen saturation ranges into pulse oximeter tones, can enhance alarm effectiveness and help to reduce the need for visual checks.

Endsley and Jones¹³ recognised unnecessary complexity as a major obstacle to achieving situational awareness and advocated for direct information transfer as an effective solution. As an example from the visual domain, we applied these principles in Visual Patient Avatar using simplified visualisations that connect to the real-world physiological values they represent. This approach has proven effectiveness and can similarly enhance the design of voice alerts and auditory icons.^18,19

Breakdowns in situation awareness have been identified as the leading cause of death, undesirable outcomes, and malpractice claims in anaesthesiology,²⁰ just as they have been identified in other high-reliability industries.^13,15 According to James Reasons' ‘Swiss Cheese Model’, multiple layers of defence typically exist between hazards and accidents. Still, flaws or holes in each layer can align, creating a pathway for an accident. The model illustrates how every improvement in situation awareness helps close the holes in the layers of defence.²¹

Alarm fatigue

Alarm fatigue is a serious patient safety issue when healthcare providers become desensitised to the constant sounds of medical device alarms.²² In intensive care units, the overwhelming number of alarms can make it difficult for clinicians to distinguish urgent situations from minor deviations. This can lead to delayed or missed responses to critical alarms, putting patients at risk. Studies show that many clinical alarms are false or non-actionable, contributing to excessive noise and unnecessary disruptions.^8,22 Overcoming alarm fatigue will require a comprehensive approach, combining optimised acoustical alarm design with strategies to reduce bedside noise and transfer alarms directly to clinicians, as Koomen and colleagues²³ have suggested.

Silent alarms

These challenges have also pushed development of the ‘Silent ICU’ concept, aiming to reduce bedside alarm sounds. The Institute of Electrical and Electronics Engineers (IEEE, New York, NY, USA) has published IEEE 11073 SDC (Service-oriented Device Connectivity),²⁴ a standard adopted by most manufacturers, enabling real-time, bidirectional data transfer and remote control between devices from different manufacturers. Integration into practice is underway and will enable the seamless transmission of alarms to central dashboards or mobile devices, reducing bedside noise while ensuring critical alerts remain reliable. Acoustic alarm design might need to be different for diverse care areas. In the OR, direct alarm notification on the monitor can be more frequently required than in the typical ICU workflow. In some settings, ‘Traffic light’ systems might be helpful to monitor ambient noise levels, providing visual feedback to staff when noise exceeds predefined thresholds.²⁵ This could improve patient sleep quality and help reduce alarm fatigue among healthcare providers in the ICU, where the sensory environment has been found to be unnatural and therefore potentially harmful.²⁶ Maintaining reliable alerting systems that promptly notify clinicians of critical patient needs is crucial. International guidelines advise against completely silencing or turning off audible alarms.²⁷

The future of patient monitoring alarm sounds

Why not consider a combination of all three alarm modalities for the future of patient monitoring? A nuanced approach could be used depending on the severity of the vital sign deviation: starting with an auditory icon, escalating to a voice alert if the situation persists, and culminating in the traditional red alarm tonal alarm in an apparent emergency. With the now rapid advances in artificial intelligence, such systems should also support the choice of which sound modality to deploy in a certain situation. As long as humans remain the final decision-makers, our guiding principle in designing such alerting systems must always be to ensure optimal human situation awareness. With the integration of more devices in the future, a syringe pump, for example, might trigger a clear voice alert: ‘The propofol syringe is almost empty’, instead of nonspecific beeping emanating from the general area of the syringe pumps that is common today, which persists until we walk over, identify the problem, then manually silence it. Although research has identified voice alerts as the most effective modality, an environment dominated solely by voice alerts would likely not be ideal for clinicians or patients alike, just as the current use of only conventional sounds is not ideal. It is clear that each sound modality carries limitations and that intelligent integration will be key. Humans thrive in diverse and acoustically rich environments. Patient monitoring could benefit from a soundscape where speech, auditory icons, and refined conventional alarms work together seamlessly. Such a system, akin to the layered auditory experience of the natural world, could replace today's dimensionless, stress-inducing, and often inefficient alarming systems which are built on 70-yr-old sound technology from a vastly different era without human factors input. As in aviation, visual and audio alerts should be integrated with one another to provide optimal situation awareness.

Call to action

We call upon all stakeholders to drive this change. By fostering collaboration, we can ensure cutting-edge technologies are integrated into monitor solutions, enhancing patient care and safety. We ask manufacturers to collaborate with clinicians and researchers to build innovative solutions, clinicians to embrace these innovations, and researchers to continue their evaluation, particularly in real-world applications.

Authors’ contributions

Conception: DWT

Drafting the original manuscript, revising, reviewing, and editing: all authors

Reviewing and approving the final version to be submitted and published: all authors

Declaration of Generative AI and AI-assisted technologies in the writing process

The human authors used generative AI and AI-assisted technologies in the writing process (OpenAI, ChatGPT, various models, San Francisco, CA, USA). These technologies were used to improve the readability and language of the work. Applying the technology was always done with human oversight, and the authors carefully reviewed and edited the results. The human authors take ultimate responsibility and accountability for the content of the work.

Declarations of interest

TRR is an inventor of Visual Patient Predictive and Visual Blood, with intellectual property held by the University of Zurich and Philips. TRR receives research funding, honoraria, and travel support through joint development and licensing agreements from Philips. DWT is an inventor of Visual Patient, Visual Patient Predictive, Visual Blood, Visual Clot, and Visual Hemofilter, with intellectual property held by the University of Zurich and Philips. DWT receives research funding, honoraria, royalties, and travel support through joint development and licensing agreements. Instrumentation Laboratory–Werfen, the Swiss Foundation for Anaesthesia Research, the International Symposium on Intensive Care and Emergency Medicine, and Stiftung Charite have provided additional honoraria and travel support. DWT serves on the Philips Patient Safety Advisory Board. The other authors declare that they have no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: