Designing Effective Alarm Sounds

Abstract

Alarms are known to be a significant problem in the medical arena, testament to which is that the issue of “alarm management” has become a significant topic in itself. However, many of the problems associated with alarms could be reduced considerably through appropriate application of some important human factors principles together with application of the considerable published body of scientific knowledge specifically concerning alarms. This article will look at key areas and summarize important findings which are of direct application. Three central issues are considered: the alarm philosophy, which is the thinking that underpins how alarms are used in any specific area; the design of the alarm sounds themselves; and the relevance of these findings to the international standard IEC 60601-1-8.Some of the most important issues concerning alarms and their application relate to how alarms are used in the workplace conceptually. For example, many research articles cite that alarms often proliferate, are often ignored, and are often unidentifiable.123 A key cause of this problem is the lack of an alarm philosophy guiding alarm implementation. An alarm philosophy provides a conceptual background for thinking about how alarms will be used and assigned to specific situations.An alarm philosophy should include the following features: that alarms are uniquely assigned to specific situations, or groups of situations; that high-priority situations are signaled by unique, one-to-one matched alarms with lower priority situations priority-coded only; that hazard matching occurs (so that more hazardous situations have more urgent alarms); and that standardization of alarms occurs. Such an approach will in itself reduce the occurrence of many of the problems associated with alarm use, although the actual alarm sounds which are chosen can also compromise the effectiveness of a set of alarms. For example, the alarm philosophy in IEC 60601-1-8: 2006 is generally sound and meets the background requirements of a good alarm philosophy, whereas there has been debate concerning the design of the alarm sounds themselves.Alarm philosophies have been developed in many areas outside medicine, such as aviation4 and within medicine.5 To some extent an alarm philosophy underpins the way alarms are designated in IEC 60601-1-8:2006.6 Here, the situations which require signalling have been specified and, because alarms are designated on the basis of the medical problems that they represent (which will remain largely the same in the future) rather than specific equipment and/or situations (which are likely to change), the standard has some of the important elements of an alarm philosophy which are likely to lead to it being effective. Hazard matching has also been attempted because there are two versions of each alarm, one for highly hazardous situations and one for moderately hazardous situations.To avoid alarm proliferation, the most hazardous situations should have unique alarms; less hazardous situations can be indicated by a single alarm which indicates this lower priority. Logically, if a situation is less than critical then only the priority of the situation needs to be signaled because the receiver is able to check the situation out in a less time-critical way, for example by examining the patient or visual displays. Of course in practice this is hard to achieve for a number of reasons. One reason is that it is not always possible for even intelligent monitoring devices to interpret criticality. Another reason is that new pieces of equipment are being introduced all the time and the manufacturers of those pieces of equipment will use their own alarms which may not easily integrate in a more general fashion.Employing good human factors in alarm design and usage requires thinking at the level of how alarms are integrated. The goal here should be to reduce the number and occurrence of alarms, not to increase them.73Two key aspects to alarm sound design include learnability and types of sound, and urgency of sounds.The days are long gone when medical alarms had to be buzzers, beepers or some other rather basic sound. The amount of computing power available on many pieces of medical equipment, particularly the more elaborate and expensive, means that alarms can be almost any sound of the manufacturer's or standard committee's choosing. This creates a whole new set of problems for the designer, because there are a number of factors which need to be taken into account. Among the most important of these problems are that some types of sound are easier to learn than others; that some types of sound are more acceptable to users than others; and that some classes of sound are easier to implement in a potentially noisy environment than others.In terms of classes of sound, alarms can be abstract (such as beeps, buzzers and more contemporary sounds), tonal/melodic (such as those designated in IEC 60601-1-8: 2006), speech, “auditory icons,” or “earcons.” “Auditory icons” is a term used to describe a large range of sounds, usually real-world sounds, which bear a distinct relationship to the hazard they are indicating. “Earcons,” originally developed by Brewster (1993)8 are tonal alarms with specific and hierarchical relationships with one another. Each of the classes of sound described above, together with indications as to how they might be used in a work environment, are demonstrated in a sound library designed by the author and set up by the UK's Rail Safety and Standards Board.9In terms of alarm learnability, two research projects1011 demonstrated that there are significant differences in how easy different classes of alarms are to learn. Both studies showed that abstract and tonal alarms are difficult to learn, auditory icons are somewhat easier to learn, and speech alarms are very easy to learn (because the message requires no learning). Obviously there is more to alarm use than just learnability, but in the first instance it would seem logical that alarms should be easy to learn. A study by Sanderson, Wee & Lacherez (2006)12 demonstrates that the alarms currently associated with IEC 60601-1-8:2006 are difficult to learn (as the literature would predict) and also that people with musical training tend to fare better with learning those alarms than do those without.Thus the research would suggest that tonal and abstract alarms ought to be used with caution. They are useful if it is necessary to code the alarm (for example, so as to avoid distress to both patient and family) but the cost is that more time and effort will need to be put into learning the meanings of alarms. If abstract alarms are to be used, then urgency-coding them (see below) is the most ergonomic way of cutting down the amount of learning required. The alarms in IEC 60601-1-8: 2006 are urgency-coded to some extent, but there are many more things that could be done in terms of alarm design to aid differentiation between high- and medium-priority alarms.Auditory icons are a useful class of sounds and have been shown to perform better than some other types of alarms.1314 This class of sounds can in principle be almost any sound, but the key requirement is that there is some kind of link (often metaphorical) between the sound used and the hazard it is signalling. There is no such link between abstract alarms and the hazards they are signalling, and therefore the only way to establish a link is to learn it.For example, a possible auditory icon for a ventilator would be the sound of someone breathing or wheezing; a cardiac monitor would be a heartbeat; a more generic signal for a parameter becoming critical might be a screeching animal or some other naturally-occurring sound. In practice, implementing auditory icons in the medical arena might be somewhat problematic as the sounds may be neither acceptable to users nor sufficiently acoustically robust in the noise environment in which they would be used.Speech (or text) also has clear advantages but there may be factors such as discretion and audibility which might also limit their use. Thus there are pros and cons for each of the classes of sound, and thinking about what classes of sound to use for specific application should be part of the alarm philosophy. In any case, much of the problem with alarms could be ameliorated by reducing the total number of alarms and standardising those sounds.In practice an alarm set may consist of a mixture of different types of sound. The more heterogeneous the alarm set, the easier the set should be to learn, and in particular the listener should be more able to distinguish between one alarm sound and another.15 In this author's opinion, the alarms specified in IEC 60601-1-8: 2006 are too homogeneous for adequate discrimination to be readily achieved. In addition to all being tonal, the variation within the set of alarms is small. Much greater variation could be achieved even within a set of tonal alarms. Unfortunately the concepts of non-uniformity and variation do not sit well with the reasonable objective of having a consistent, documented and defensible rationale when specific alarm designs are documented, such as in IEC 60601-1-8: 2006. Nevertheless, variation in design should be encouraged.In the animal kingdom, animals that live in a 3-D environment tend to use predator-specific alarms whereas those who live on the forest floor (and therefore live in a 2-D environment) tend to have urgency-coded alarms only. This is because animals living in a 3-D environment need specific information about what kind of predator it might be (because they need to know whether to go up a tree or go down to the ground, for example) whereas the animals who live on the ground simply need to know how quickly they should run to their burrow. This translates very readily into the way alarms might be used in the workplace; if the receiver needs to know precisely what the problem is, then a specific alarm is needed. If they simply need to know how quickly to respond, then urgency-coded alarms will suffice. This is particularly true for all alarms other than the most urgent, for which specific information might be required.Fortunately the urgency coding of alarms is a topic that has been explored in detail in the literature for both abstract alarms and speech.16–22 There are significant experimental databases showing how the urgency of an alarm can be reduced or increased by altering acoustic parameters such as loudness, pitch, rhythm, repetition and so on. The research also demonstrates how we can predict the change in urgency that will be achieved through specific changes in those acoustic parameters.This database can be applied not only to determining the urgency of a set of alarms associated with situations that might vary in their urgency, it can also be applied to the use of intelligent alarms that might change in their urgency depending on the current status of the parameter or parameters being monitored. It is fair to say that the known methods of designing different urgency levels into alarms have largely not been applied to the alarms specified in IEC 60601-1-8: 2006.Some classes of sound will be easier to manipulate in this way than others. Abstract and tonal alarms, though initially hard to learn, lend themselves to manipulation in this way. Auditory icons, because they tend to be real-world sounds, may lend themselves rather less to this manipulation. Speech can be altered in urgency,2122 though, because speech messages are easier to learn, additional urgency-coding might not be necessary.One of the biggest advantages of urgency-coding is that even if the hearer does not know the immediate meaning of an alarm, the seriousness with which they should take the alarm and the speed at which they should respond will be indicated by the alarm itself. In reality, alarms are often not correctly urgency-coded. For example, Momtahan et al (1993)2 demonstrated that medical staff working in an operating theatre and recovery room could only recognise approximately half of the almost 50 alarms used in that area. In addition, many of the more strident alarms were associated with less urgent situations, so that listeners, in the absence of information as to what the alarm meant, could not use the acoustic information coming from the alarm to determine the relative urgency of the situations being signaled.If the meanings of alarms are known, then urgency coding is not necessary, because the listener can use their knowledge of the meaning of the alarm to judge the urgency of their response. But in situations where listeners do not know the meaning of an alarm (which is fairly frequently) then proper urgency coding will give an indication as to how quickly they should respond. The most ergonomic approach to designing alarms is to have first unique alarm-to-hazard mapping for the most urgent situations—because specific information is required—and to have second urgency-coded lower priority generic alarms —because specific information is not required from the alarm itself; thereby, the total number of alarms will therefore be kept to a minimum.There is some agreement that the alarms specified in IEC 60601-1-8: 2006 are less than optimal. Sanderson, Wee and Lacherez (2006)11 have demonstrated that the sounds are difficult to learn, favor listeners with musical training, and do not necessarily differentiate between high and medium priority. Block (2008)23 has conceded that the alarms are not ideal. The reasons for this include the following. First, melodic alarms, we now know, are difficult to learn and retain and so are at a disadvantage right from the start. There are other classes of sound which are easier to learn, if ease of learning is the aim.Second, the alarms specified in the standard are much more homogeneous than they need to be. They are all the same length and have the same rhythm and general pitch range, are tonal and so on. They will therefore be much more difficult to tell apart than a more varied set of alarms. For example, if there is variation in the number of pulses and the rhythm, the listener will have more cues in order to differentiate between the alarms. Using mnemonics to distinguish between the alarms simply will not work if all of the alarms have the same number of pulses in them, as the single most important cue from a mnemonic in this instance is its number of syllables. If all of the mnemonics have the same number of syllables—which they must do, because the alarms all have the same number of pulses—then the mnemonics will not be an aid to discrimination. In fact, the mnemonic is no more of an aid to remembering the tone sequence than the tone sequence itself is an aid to remembering the mnemonic.Finally, in the current standard, urgency differentiation is achieved by simply adding extra pulses to the more urgent form of the alarm, whereas there are much better and more varied ways of manipulating urgency.