An alarming problem
Anyone who has spent time in a critical care environment will be familiar with the sound – amongst the backdrop of clinical conversation, busy movement of people and the hum of machinery is a chorus of ‘pings’ and ‘bleeps’ demanding the attention of nearby staff.
We help our customers revolutionise therapy through ground-breaking surgical devices, implants and acute care platforms.
Alarms exist as a risk mitigation strategy and are a crucial aspect of all critical care or life sustaining machinery. Risk analyses highlight potentially hazardous situations for which alarms should be used to warn clinicians of impending dangers. Issues arise however, when the quantity of alarms in hospitals impart high cognitive loads on clinicians and strain their working capacity.
A study of perioperative alarms in surgical settings (1) identified 8,975 alerts and alarms occurring across 124 hours of monitoring yet only 6,386 were classified as serious and life-threatening. Furthermore only 70% were deemed valid, with the remaining 30% caused by artefacts, and of these valid alarms only 39% were classified as clinically relevant! One consequence of false alarms is the intentional ignoring, or hesitation to respond by clinicians to avoid ‘Crying Wolf’ and annoyance may encourage users ignoring signals or even taking risks by changing to alarm thresholds to inappropriate levels. Reaction times also increase when there is low alarm validity; (2).
Overloading has been attributed to ‘alarm fatigue’ (3) – the desensitisation of the brain to persistent stimulations, in a similar manner to the olfactory adaptation which prevents you from smelling your own home or fragrance after a period of exposure.
The ‘heavy traffic’ of multiple tones activating simultaneously mask the critical signals through a wall of noise. Sound levels of between 60-90dB are common within ICUs (4), which can be distressing to patients and relatives alike and lead to a loss of confidence in the care being given. High background noise has been demonstrated to impede patient recovery due to sleep deprivation recovery (5) and is correlated to increased opioid usage and rehospitalisation rates (6). Some evidence even suggests that the constant drone of distressing tones contributes to burn-out in ICU nurses (7).
Ultimately, these phenomena lead to notifications of clinical relevance being ignored or silenced by the health care team, raising risks to patients.
Regulations [including ISO 60601 & IEC 80001] provide guidance on the design and application of alarms to medical devices, and advise assigning priority to alarms dependent upon both the consequences of a failure to respond and the rate of onset for those potential harms. Standards are very specific on the mode of communication, the volume pitch and pulsation of alarms, and that they cannot be permanently disabled (although temporary suppression is allowed); however they are not always applied consistently which convolutes the repertoire of signals that clinicians need to learn.
The guidance also allows for ‘information signals’, which are intended to mitigate issues which may not, in and of themselves, be dangerous or require any medical intervention. However, if not carefully considered, these signals may end up being superfluously applied. In some cases, manufacturers can overestimate the clinical value of some of the alerts they implement. Equally, designers can become so engrossed in the development of their devices that they lose sight of the wider context in which the machines will be used.
A number of strategies exist to try to curb the ineffective application of alarms and their associated impact to patient safety:
Maintain signal integrity
Many false alarms are caused by signalling artefacts – technical disturbances which alter the quality of the readings. For instance, poor adhesion of sensors to the skin can cause signal aberrations. This should be primarily overcome by physical means (e.g. the development of better connectors and bonding agents) but also through clinical protocols (i.e. routinely checking the integrity of machinery, connections and sensors). Within their telemetry trial, Cleveland Clinic demonstrated that an intervention in which nurses regularly checked electrode placement and wire attachment yielded a 29% reduction in nuisance alarm events.
Improve usability and customisation
Often manufacturers provide the means for trigger limits to be set for certain alarm variables. This allows machines to be configured to patient specific settings and potentially reduce the incidence of false alarms; however the usability of these interfaces can be complex, with operators often unable to correctly program custom limits, or moreover even recognise that this functionality is available (8).
A large number of false alarms are caused by only minor threshold violations of short-duration (9). Graduated time delays in which severe deviations are alarmed faster offers a flexible approach to mitigate the likelihood of an insignificant interruption. One validation option is to have patient monitoring performed by humans via remote telemetry. Cleveland Clinic developed an off-site central monitoring unit providing cardiac telemetry for non-critically ill patients, eliminating many audible bedside alarms. Amongst other metrics, the control team monitors blood pressure, heart and respiratory rates, and provide only urgent notifications to ward nurses for rapid response.
Additionally, the amalgamation of data from additional sources can help validate whether an alarm is warranted. For example, ECG, continuous arterial blood pressure monitoring and pulse oximetry are all commonly used in critical care environments to measure different aspects of cardiac function which are individually treated as vital signs. Monitors will enunciate an alarm if any one has a value that lies outside a threshold, which could be due to a genuine emergency or to an unreliable signal. By considering a combination of signals, the clinical feasibility of an alarm state can be assessed and conveyed accordingly - thus giving greater credence to emergency alarm states. Integration of this data will require the adaption of a universal exchange protocol for healthcare data, (i.e. FHIR standard) with the analysis either performed on device or by a centralised clinical informatics solution such as the Philips Intellispace Critical Care & Anaesthesia.
In addition to more conventional statistical approaches for alarm generation, modern techniques such as Deep Learning and Bayesian networks can enable monitoring machinery to anticipate future events on the basis of advance information. This can assist in disseminating between probable physiological events and unintended artefacts. For example, Edwards’ Acumen Prediction Index uses machine learning (upon a data set of 200,000 patient events) to estimate the risk of a forthcoming hypotensive event; facilitating the move from descriptive to predictive real-time patient monitoring. Imagine the value of a critical care monitor that could diagnose issues from the waveform, anticipate an impending deterioration and distinguish between physiological and physical artefacts.
A renewed focus is needed on usability engineering to ensure the interfaces of critical care equipment are not merely safe but comprehensible to a level which encourages clinicians to utilise the full extent of their capabilities. Equally, comprehensive training should be offered by manufacturers to educate users about this functionality, and guidance developed to establish when alarms should be set to bespoke levels rather than the population mean, and what these levels should be.
Superfluous alerts are more than a mere nuisance in critical care; their presence raises risk factors by desensitising clinicians to more serious alarms, and also inhibits patient recovery.
A number of options exist to mitigate false alarms, covering multiple avenues including technical design, software algorithms, human factors engineering and targeted clinical protocols. Ultimately, making critical care a quieter and safer place will require a universal perspective with concerted efforts from hospitals and equipment manufacturers alike.
1. Schmid F, Goepfert MS, Kuhnt D, Eichhorn V, Diedrichs S, Reichenspurner H, Goetz AE, Reuter DA. The wolf is crying in the operating room: patient monitor and anesthesia workstation alarming patterns during cardiac surgery. Anesth Analg. 1, 2011, Vol. 112.
2. Felix Schmid, Matthias S Goepfert, Daniel A Reuter. Patient monitoring alarms in the ICU and in the operating room. Crit Care. 2, 2013, Vol. 17.
3. Bridi, Adriana Carla. Reaction time of a health care team to monitoring alarms in the intensive care unit: implications for the safety of seriously ill patients. 2013.
4. Kam PC, Kam AC, Thompson JF. Noise pollution in the anaesthetic and intensive care environment. Anaesthesia. 1994, Vol. 49, 11, pp. 982-986.
5. Aaron JN, Carlisle CC, Carskadon MA, Meyer TJ, Hill NS, Millman RP. Environmental noise as a cause of sleep disruption in an intermediate respiratory care unit. Sleep. 1996, Vol. 19, 9, pp. 707-710.
6. BB, Minckley. A study of noise and its relationship to patient discomfort in the recovery room. Nurs Res. 3, 1968, Vol. 17.
7. Topf M, Dillon E. Noise-induced stress as a predictor of burnout in critical care nurses. HeartLung. 1988, Vol. 17, 5.
8. Drews, Frank A. Patient Monitors in Critical Care: Lessons for Improvement. Advances in Patient Safety: New Directions and Alternative Approaches. Rockville MD : Agency for Healthcare Research and Quality, 2008. Vol. 3.
9. Görges M, Markewitz BA, Westenskow DR. Improving alarm performance in the medical intensive care unit using delays and clinical context. Anesth Analg. 5, 2009, Vol. 108.