Alarms are still a problem!

In this issue of Anaesthesia, de Man et al. 1 demonstrate that the typically reported problems with alarms during high-risk surgery are encountered also in moderate-risk surgery. They found that alarms sounded very frequently (on average every 2.5 minutes). Many alarms that sounded were not clinically relevant (they were either clinically irrelevant or false), and of those that were assessed as being clinically relevant, most of these were of low, rather than high, priority. The findings illustrate the problems that alarms still cause for the clinician, and pose a range of questions from the relatively simple question as to whether an alarm is useful or not, to more complex questions such as tackling the problem of alarm fatigue and the development of more sophisticated alarms. IEC 60601-1-8 2, an important standard concerned with medical alarms, contains the following statement in one of its annexes: “The alarm system should result in a greater probability that the operator will correctly detect and appropriately respond to the condition that requires their awareness or action than would be the case in the absence of alarm signals”. It seems obvious that the presence of an alarm must add value to a medical monitoring device. Yet, despite their being a major topic both in the research sphere and in the clinical working environment, there isn't as much evidence as we would expect that unequivocally demonstrates that alarms do add value and make the clinical workplace safer. Indeed, many deaths have been attributable to alarm problems. FierceHealthcare, a US-based daily newsletter, draws attention to the fact that over 200 hospital deaths have been linked with alarm fatigue 3. As we typically have few data on lives saved by alarms, the ‘alarm problem’ seems to threaten the usefulness of alarms at the most basic level. The President of the Association for the Advancement of Medical Instrumentation (AAMI) has issued the bold challenge that “by 2017, no individuals will be harmed by adverse alarm events” 4. Many of the deaths have been attributable to simple factors such as alarms' not being tested, or not working at all (indicating clinicians' reliance on alarms), and are not necessarily connected with the more complex questions that revolve around alarms, though there is an assortment of problems with these more complex issues. Assuming that alarms are here to stay (the fear of litigation alone suggests that they will be), then there are many challenges that need to be met in order to create a working environment where alarms are helpful, informative, and correct. Demonstrating that alarms have a direct effect on safety is not a topic that can be easily studied in the clinical environment – it would be difficult to obtain ethical clearance for any studies that removed alarms from a clinical environment, for example. Some laboratory-based research does, however, demonstrate that the addition of alarms to a hazard-detection and judgement task improves safety 5, 6. Thus, alarms can be demonstrated to have utility. However, in the clinical environment the a priori advantages of alarms are often obscured by the practical consequences of a ‘better-safe-than-sorry’ approach to alarm implementation. One of the most obvious ‘better-safe-than-sorry’ issues is that alarms are often shrill, loud, and unnecessarily high-pitched in acoustic quality. This is entirely unnecessary, and indeed such alarms can be less effective than gentler ones. Whilst the motivation to make alarms high-pitched and shrill might be to make them sound ‘alarm-like’, the ability of users to localise such alarms is not usually improved by having them thus. Indeed, the pitch of many alarms is often (unintentionally) placed so that the user is able to use neither of the auditory mechanisms available to localise sound (one of which operates at relatively low pitch, and the other only at very high pitches). Furthermore, alarms are often ‘tinny’ and of poor acoustic quality, meaning that they can be susceptible to masking by other sounds. Thus the acoustic quality of many alarms could be improved. Of course, there are many other cues that the clinician can use in order to assess the situation, such as visual observations, their own knowledge and expertise, expectations, and information from other systems or clinicians. A useful alarm would be one that confirms, or pre-empts, the clinician's view of what is happening, rather than providing an often unwanted, uninformative, and irritating signal. The fluid way in which clinicians interact with alarms, which may be driven in part by their awareness of high false-alarm rates, makes it difficult to determine whether or not an alarm is effective. Research involving observation of the way clinicians interact with alarms in situ 7-9 demonstrates that clinicians use alarms in a variety of ways, depending on the particular process on which they are working 8. Sometimes, though appearing not to have noticed the alarms, they may have in fact made use of the information and respond much later, perhaps after a minute or so 9, 10. One study 7 presents a view of the clinician's behaviour that is on a continuum of activity from self-generated behaviour to that initiated by an alarm, and complementary research has indicated the interplay between various levels of attention in responding to alarms 11, 12. By far the largest problem with alarms in the clinical environment is that of false alarms, and the real problem of alarm fatigue. We know that people will respond to alarms in proportion to the perceived (or actual) reliability of the system, so if alarms are 90% reliable they will respond slightly more than 90% of the time, and if they are 10% reliable they will respond slightly more than 10% of the time 13, 14. Thus immediately we can see that performance will be compromised by high false-alarm rates. Furthermore, the constant sounding of alarms that the clinician knows to be inaccurate adds to the noise of the working environment and can be both irritating and distracting. Finally, many of the alarms provide little or no information other than to attract attention. I would argue that the problem of high false-alarm rates is the most significant current issue concerning alarms, and that whilst immediate and practical measures can be taken to ameliorate the problem, the responsibility for reducing false-alarm rates lies also with manufacturers and the bodies that mandate how alarms are implemented in practice – in particular, the standards bodies. There is a raft of practical measures that can be taken to lower false-alarm rates, many of which can be implemented rapidly. These will have greater success if they become standard practice in the work environment through the use of protocols and best practice guidelines. Issues such as proper electrode placement, skin preparation and alarm setting appropriate to the individual patient can of course be dealt with by clinicians whilst going about their work. For example, properly applied single-use pulse oximeter sensors will be less prone to false alarms than recycled ones, and regular checking of cables can be important in keeping false-alarm rates down 15. The need for protocols for proper skin preparation and placement of electrodes has also been highlighted 16. A further useful practical intervention is to introduce a time delay in the triggering of the alarm. Many events that trigger alarms at their most conservative settings go away within a few seconds; therefore, introducing a time delay between the onset of the trigger and the onset of the alarm will reduce the number of false alarms. This practice will have a significant effect on reducing low-level and low-priority alarms. A study in a medical intensive care unit showed that false-alarm rates go down considerably if a delay of 5–20 s is introduced, with little damage to the integrity of the system 10. Because this study recorded whether each of the alarms observed was actionable or otherwise, it was possible to discern that clinical staff selectively respond to alarms. Staff responded to few of the alarms, but of those that were responded to, over two thirds were clinically relevant. This study also modelled the effect of time delay on different categories of alarm (ignored; ineffective; effective – technical; effective – patient) and demonstrated that introducing a time delay of 19 s had a much more significant effect on the reduction of ignored and ineffective alarms than on those that proved to be clinically or technically significant. Thus, introducing time delays appear to reduce alarm rates significantly, while having only a small effect (other than lengthening response times) on the detection of true alarms. Another study 15 also draws attention to the benefit of introducing time delays. Analysis of actual alarms in SpO2 monitoring indicates that introducing a delay of 5 s would reduce alarm frequency by 32%, a 10-s delay would reduce it by 57%, and a 15-s delay would lower alarm frequency by 70%. Given that many alarms are caused by self-correcting desaturations, introducing delays would seem to be a relatively risk-free way of reducing unnecessary alarms. This study also demonstrates the effect of lowering the threshold settings. For example, reducing the SpO2 alarm threshold to 88% instead of 90% reduces alarms by 45%. Lowering it further reduces the alarm rate more, though the setting of alarm thresholds clearly has to be done on a case-by-case basis. Combining both alarm delays and lowered thresholds can have a huge effect on lowering alarm rates 15, 17 though care needs to be taken to ensure that the settings are appropriate for that patient and that situation. For example, lowering the SpO2 threshold to 85% and introducing a time delay of 15 s would produce a six-fold reduction in alarm frequency, but the consequence is that any periods of less than 15 s where desaturation is just above 85% would not be signalled. With one-to-one monitoring this may be acceptable, but with poorer staff-to-patient ratios this strategy might need to be reviewed. Alarm management of this sort, which might vary from patient to patient, could still be a useful part of the patient's care plan; it certainly seems to be worth the time and attention it requires in order to reduce the effects of alarm fatigue. Many US bodies that are involved with medical devices are aware of the ‘alarm problem’ and both the Emergency Care Research Institute and the AAMI, along with the Joint Commission, list alarm problems close to the top of their lists of top ten safety hazards. A summit was held in Herndon, Virginia, USA in October 2011, recognising the alarm problem, with a view to tackling this issue at a high level (see http://www.aami.org/meetings/summits/alarms.html). The outputs from this summit, the ongoing debate, and the development of successful measures for reducing false-alarm rates should help to reduce this problem. Whilst practical measures can have a dramatic and relatively swift effect on false-alarm rates, and therefore on the clinician, this puts the solution (and therefore the responsibility) in the hands of the end-users rather than the manufacturer of that equipment. At a slower pace, the equipment itself can help to reduce false-alarm rates. The better the signal extraction techniques used (for example, the better any signal interpretation that allows clinical data to be separated more cleanly from ‘noise’), and the more ‘intelligent’ the system that surrounds the alarm, the more likely it is that the resultant alarms will be true, rather than false. Whilst the availability of a range of techniques and considerable research demonstrating that greater system integrity results in lower false-alarm rates have been around for some time, at least since the 1980s (summarised in 18), the uptake by industry has been somewhat slow, though there now seems to be considerable progress in developing systems that reduce false-alarm rates substantially. Manufacturers are of course motivated by their own sales to make their technology as up-to-date and reliable as possible, but a further method of ensuring this takes place is to mandate the use of the best available technology. IEC 60601-1-8 is in a position to do this. IEC 60601-1-8 is an international standard concerning basic safety requirements for medical electrical equipment and seeks to harmonise medical device alarms 2. It is a collateral standard covering a wide range of equipment (more specific standards deal in greater detail with equipment and devices with a narrower application). The standard does contain information about many aspects of better alarm delivery such as intelligent and distributed alarms, distinctions between clinical, technical, and other alarms, and limit setting, but does not require manufacturers to implement any specific recommendations as part of the standard. It also does not specify when alarms should be used, but rather indicates that if alarms are used, they should comply with the standard. More specific guidance on these issues may be the remit of more specific standards, but it seems something of a missed opportunity not to have been a little more directive about some of these important issues. Somewhat ironically, the standard has a great deal to say about the alarm sounds themselves, which are specified in minute detail. These have been found to be problematic in that they are melodies, which are difficult to learn compared with some other types of sounds that might have been used 19, 20, and they are very similar in structure, compromising learning yet further. The standard does, however, allow the use of other alarms, providing certain conditions are met. The practical consequences of high false-alarm rates are multifaceted and the solutions lie in a range of places and responsibilities. In practice, high false-alarm rates hinder performance and cause irritation and fatigue. It is not surprising therefore that clinicians are tempted to turn them off and/or ignore them. The consequences of high false-alarm rates for research is that they hamper progress in areas such as, for example, the ergonomic design of audible alarms and the understanding of how clinicians learn and retain the meanings of alarms. How clinicians interact with alarms is also difficult to discern and it is to the credit of researchers who have looked at this issue that they have managed to elucidate this process. No external funding and no competing interests declared.