Signaling Patient Oxygen Desaturation with Enhanced Pulse Oximetry Tones

Pulse oximetry is widely used to monitor patient peripheral oxygen saturation (SpO₂) and pulse rate, mostly in the operating room (OR) and postanesthesia care unit. Evidence suggests that pulse oximetry helps clinicians reduce hypoxemic episodes,¹  though evidence for further clinical benefits to patients has been elusive.^1–3 

Since early in their development, pulse oximeters have emitted a tone with every detected pulse. The pitch of the tone is related to the measured saturation, and the repetition rate of the tone conveys the pulse rate, both of which assist clinicians in detecting changes. Although the variable-pitch tone has contributed to the success of the pulse oximeter by allowing eyes-free monitoring, room for further improvement exists.

The objectives of this article are to (1) review what is known about the variable-pitch SpO₂ tone and (2) outline the findings of research directed at making the tone more informative.

The Nellcor pulse oximeter patent, initially filed in 1982,⁴  includes a claim for a variable-pitch tone, and Nellcor pulse oximeters were the first to emit “a tonal signal … having a pitch proportional to the ratio of SpO₂ and a sequential repetition proportional to pulse.” Nellcor made the technique available to the public in the 1992 ASTM F1415 standard, which became ISO 9919, which in turn was superseded by ISO 80601-2-61. Thereafter, some implementation of a variable-pitch tone was included on most clinical pulse oximeters.

In a tribute article to Dr. Takuo Aoyagi, the inventor of pulse oximetry, contributors mentioned the power of the variable-pitch tone.⁵  For example, Jeff Cooper noted, “Perhaps, the most appealing feature was the changing sound of the pulse tone with decreasing saturation” and the impact of “the confidence during intubation on hearing that so recognizable tone indicating the state of saturation.” In the same article, a group from Lifebox Foundation noted that “anesthesia providers often refer to their pulse oximeter as their ‘eyes and ears.'”

In 1982, electrocardiogram (ECG) monitors already emitted a tone with every heartbeat, but with fixed pitch. The idea to map tone pitch to SpO₂ was a well-justified choice. As a systematic review has shown,⁶  pitch is the most commonly used auditory dimension to convey quantity, whether on a continuum or in discrete steps. Moreover, increases in pitch are almost universally associated with increases in quantity, whereas decreases in pitch are associated with reductions in quantity. The pulse oximetry standard ISO 80601-2-61:2019⁷  does not require a tone; however, if a tone is provided, the standard does require that “the pitch change shall follow the SpO₂ reading (e.g., the pitch decreases as the SpO₂ reading decreases)” (section 201.102, p. 23).

The advantage of the variable pitch for conveying desaturation is so self-evident that only one study seems to have been needed to demonstrate it. Craven and McIndoe⁸  showed that with a fixed-pitch pulse oximeter tone, desaturation was detected in an average of 129 seconds, but the variable-pitch tone allowed desaturation to be detected in an average of 32 seconds—a fourfold advantage.

Pulse oximetry was included in the original 1986 American Society of Anesthesiologists (ASA) monitoring standards,⁹  but the standards did not mention the tone. Nonetheless, use of the tone was strongly encouraged by the anesthesia community. For example, in 2004, Goldman and Robertson¹⁰  asked, “Isn't it time that we mandate the use of the pulse oximeter pulse tone for the monitoring of all patients undergoing general anesthesia and incorporate this requirement in the ASA Standards for Basic Anesthetic Monitoring? If so, we must explore related issues, such as the necessity of standardizing the pulse oximeter's pitch-saturation values.”

Around the same time, Stoelting¹¹  noted “the audible presence of the ‘beep’ tone from the pulse oximeter, plus knowing that the audible alarm on at least one physiologic monitor of the anesthesiologists' choice … is active, would provide the desired safety net to the anesthesiologists' eternal vigilance.” After 2005, in response to such observations, the ASA standards for basic anesthetic monitoring were revised to require that “when the pulse oximeter is utilized, the variable-pitch pulse tone and the low threshold alarm shall be audible to the anesthesiologist or the anesthesia care team personnel” (section 2.1.2, p. 2).

Despite the above, the effectiveness of the current pulse oximeter sound is limited. First, when listening to changes in the variable-pitch tone, listeners tend to underestimate how much SpO₂ has changed. Second, in certain clinical situations, knowing the clinical range in which SpO₂ lies is critical, but SpO₂ range is not readily conveyed with the current types of tones. These issues are addressed below. A glossary of terms appears in Table 1.

A few years after the pulse oximeter's introduction in the U.S., Schulte and Block¹²  found that only 67% of their nonclinician participants could detect a single-step drop in the SpO₂ tone of the Nellcor pulse oximeter and 11% of participants required seven successive drops of the SpO₂ tone before they reported a change in saturation. However, as Santamore and Cleaver¹³  and Loeb et al.¹⁴  have shown, the Nellcor tones have a relatively small pitch range and small pitch increments (5 Hz) compared with other monitors, possibly contributing to Schulte and Block's findings. Still, using the more generous step sizes of the Datex AS3 tones, Morris and Mohacsi¹⁵  showed that anesthesiologists systematically underestimate degrees of desaturation below 90% SpO₂, and only 70% of their sample could identify the direction of change for two tones one step apart. Similar findings were reported by Perez et al.,¹⁶  with particularly inaccurate estimates of SpO₂ values when the tone was decreasing.

The participants in the above studies made their judgments in ideal situations away from the distractions and noise of the OR. However, distractions and noise probably would lead to worse performance. Stevenson et al.¹⁷  evaluated anesthesiologists' ability to detect a drop from 99% to 98% saturation on a simulated Philips MP70 monitor, under different controlled conditions of background noise and visual attentional load. Noise reduced the anesthesiologists' accuracy at detecting the drop by around 10 to 15 percentage points and increased their response times by approximately 50 ms. Moreover, high visual attentional load decreased accuracy by around 5 percentage points and lengthened response times by approximately 125 ms.

The above studies of listeners' perceptions of pulse oximeter tones were performed with different pulse oximeter models. Several studies have noted the different mappings of the pitch of tones to SpO₂ levels across brands, which may compound the difficulties that anesthesiologists have with interpreting the variable-pitch tone, especially when moving between pulse oximeters made by different companies.^13,14,18  The study by Loeb et al.¹⁴  is the most comprehensive, showing that of 10 representative models, none had the same mappings. Differences existed in whether there was a linear or logarithmic mapping of pitch increments to saturation increments, whether pitch changed on each saturation change or every second change, whether pitch changes occurred for all saturation levels or stopped changing below a certain level, and in the actual pitch ranges and component frequencies used for the tones. The article revisited the need to consider standardization of pitch mappings to SpO₂, as Goldman and Robertson¹⁰  had advised more than 10 years earlier.

However, there are deeper problems. The enthusiasm for the variable-pitch pulse oximeter tone is based on its ability to signal change and its ability to signal whether SpO₂ is increasing or decreasing. Notwithstanding the above studies, humans have good relative pitch, meaning that except in the closest cases they can tell if a tone has changed and whether its pitch is going up or down, even if they cannot judge by how much. However, most humans do not have good absolute pitch.¹⁹  Translated to the clinical context, good absolute pitch would mean being able to identify the exact SpO₂ number on hearing a single pulse oximeter tone (or repetitions of that tone), without looking at a visual display and without hearing tones at a prior known value.

Knowing the clinical range or even the absolute saturation may be desirable in some situations. For example, hyperoxygenation can be harmful in certain clinical cases (e.g., preterm neonates, long-term critical care); therefore, the target saturation is less than 100% and a higher pitch is not necessarily better. In another example, clinicians may intervene to correct a patient's hypoxemia (desaturation) and cease intervening after they hear the tone indicating that saturation is increasing. However, SpO₂ may still be below target level, which would not be revealed until the clinician checked the visual indicator. Accordingly, researchers have taken steps to improve how effectively clinicians can identify the clinical range or even the absolute SpO₂ number from the sound of the variable-pitch tone alone.

Attempts have been made to improve perception of pulse oximeter pitch changes using multisensory training.²⁰  Anesthesiologists were trained to detect temporal differences in the onset times of a visual and auditory stimulus. After training, the anesthesiologists detected pitch changes 9% more accurately and 72 ms faster than in their pretraining performance, taken across different levels of ambient noise and visual attentional load. The findings were particularly strong with high levels of ambient noise. Training and the test of pitch perception were on separate days, but it is unclear how long the advantage of multisensory training would persist. In addition, as recognized in hierarchy of effectiveness schemes, changing human performance through training or education generally is a less desirable and less effective approach than changing performance through design.²¹ 

As noted above, Loeb et al.¹⁴  demonstrated the wide variability in how frequency is mapped to SpO₂ across current pulse oximeters. Some devices map frequencies linearly to SpO₂, and others map frequencies logarithmically. Logarithmic mappings create approximately equal-appearing pitch intervals between SpO₂ levels, closer to the mel scale²²  of perceived-as-equal pitch intervals. Brown et al.²³  demonstrated that clinicians make more accurate estimates of SpO₂ values when the values are mapped logarithmically, with each step the equivalent of a semitone. They found that when SpO₂ values were mapped linearly, clinicians underestimated the change from a reference tone, as also reported by Morris and Mohacsi.¹⁵  With logarithmic steps, clinicians also identified direction and percentage change from a reference tone more accurately. However, tests were always done relative to an anchoring reference sound.

Some studies have attempted to improve the way sound frequencies are mapped to SpO₂ levels. An initial attempt at signaling acoustically when SpO₂ goes into a nontarget range was done by Hinckfuss et al.²⁴  in a neonatal context. The pitch step sizes in the target SpO₂ range were kept quite small, but as soon as SpO₂ went into a low or high range, pitch step sizes between saturation levels increased. However, this approach did not improve listeners' ability to identify which range SpO₂ was in, compared with a logarithmically mapped scale using consistent step sizes.

Some multiparameter monitors alter the pulse oximeter sound quality when pulse oximetry is not used or the pulse oximeter signal is lost. In this situation, a fixed-pitch tone is emitted with every ECG-detected heartbeat. The ECG tone is given a different sound quality (timbre) from the pulse oximetry tone to prevent it from being mistaken for a pulse oximetry tone. As discussed below, a similar approach can be used to convey other information.

Direct comparisons of enhanced pulse oximetry tones with a version of the current pulse oximetry tones have been motivated by the challenge of avoiding hyperoxemia when providing supplemental oxygen to preterm neonates. According to the NeoPROM trial, SpO₂ of preterm neonates should be kept in a target range of 90% to 95% SpO₂ to optimize survival without hyperoxia- or hypoxia-associated morbidity.²⁵  As noted by Hinckfuss et al.²⁴  an increase in the pitch of the pulse oximetry tone is ambiguous for this population, as it could indicate a return from low levels to the target range or from the target range to high levels. The reverse problem holds for decreases in the pitch of the tone.

Accordingly, a program of research has investigated ways of enhancing the existing pulse oximetry tone to make it clear when SpO₂ is very low, low, target, high, or very high (Figures 1 and 2, and Appendix 1 [available in the supplemental material for this article at www.aami.org/bit]). In each experiment, the control and enhanced conditions used tones with a logarithmic mapping of sound frequency to SpO₂ levels, creating equal-appearing pitch intervals across the range of SpO₂ levels. For the more recent studies in this series, saturation of 100% was 950 Hz and 0% was 150 Hz. These values follow the recommendations given in Loeb et al.¹⁴  Several types of enhancements have been tested. For example, the tone itself might acquire tremolo (rapid repeated changes in volume), a different timbre (manipulation of harmonics to create a sharper or duller sound), or both. Alternatively, an extra tone (a “beacon” or, more formally, a “pointer”) might sound before every nth tone to alert the listener that SpO₂ is not in the target zone and to remind them what pitch the main tone would be if it were in the target range. Examples of these enhancements are provided in sample audio files, which are included in the supplemental material for this article (at www.aami.org/bit).

Figure 2 shows the results of five studies evaluating enhanced pulse oximetry tones designed to enhance neonatal monitoring. The first three studies assumed a consistent target range of 90% to 95% SpO₂.^24,26,27  Participants listened to short scenarios during which the pulse oximetry tone varied. At the end of each scenario, they identified the ending saturation range. In two studies, the participants also responded during the scenario if they heard the pulse oximetry tone go into or out of the target range.^26,27  Substantial improvements occurred in range identification (Figure 2A), target transition identification (Figure 2B), and latency to detect target transitions (Figure 2C) with the beacon and with the tremolo enhancements, though results for the tremolo enhancements were slightly more impressive. With tremolo, range identification improvements varied between 20–42 percentage points, target transition improvements were 25 percentage points in both studies, and the latency for detecting transitions sped up by 5.4 seconds in one study and by 13.5 seconds in a further study with clinical participants.

The two rightmost studies in Figure 2A are from studies examining oxygen targeting in the first 10 minutes after birth, where the target saturation increases each minute from around 60% to above 90%,²⁸  creating a dynamically changing target range. In one experiment, Zestic et al.²⁹  enhanced the variable-pitch pulse oximetry tone with different degrees of tremolo in the high and low SpO₂ range. In a second experiment, they added a sharp-sounding timbre in the high SpO₂ range. The SpO₂ values deemed to be low or high, and therefore subject to the enhancements, changed minute by minute. The tremolo enhancement produced an 18 percentage point improvement in SpO₂ range identification, and the addition of timbre in the high SpO₂ range led to a 29 percentage point improvement in SpO₂ range identification. In another study, Clarke³⁰  used formants to distinguish ranges (formants are vowel sounds such as “ee” for high SpO₂ range, “ah” for target range, and “oh” for low range) and achieved even more dramatic results, with a 33 percentage point improvement in SpO₂ range identification.

Despite minor procedural differences among the studies described above and the different acoustic enhancements tested, the studies consistently showed substantial improvements in performance with acoustic enhancements to the variable-pitch pulse oximetry tone. Further work will include simulation studies and clinical trials.

Other studies that have addressed neonatal pulse oximetry sounds are less relevant for the present context. A study by Janata and Edwards³¹  proposed intermittent earcons (short abstract auditory motifs) rather than continuous sonification and therefore has a somewhat different purpose. A study by Schwartz and Ziemer³²  used continuously ascending or descending Shepard's tones³³  to indicate when SpO₂ was in ranges above or below target. However, these researchers performed no determination of whether participants would confuse SpO₂ range with SpO₂ direction and rate of change.

Acoustic enhancements have also been applied to adult SpO₂ ranges (Figures 3 and 4, and Appendix 1). In a series of studies, Paterson and colleagues^34–37  enhanced the variable-pitch tone to distinguish three ranges: target, warning, and critical. In each study, SpO₂ tones in the warning range were enhanced with four cycles of tremolo, and tones in the critical range were enhanced with tremolo plus a constant level of “bright” timbre. In contrast to the neonatal studies, for the adult studies the unenhanced variable-tone pulse oximetry sound was supplemented with an alarm at the boundary between the warning and critical level, creating a control condition closer to common clinical practice.^34,36,37  Two of the studies included clinician participants, both in a desktop study³⁴  and in a full-scale simulation study.³⁷ 

As Figure 4 indicates, all studies showed marked improvements with the enhanced pulse oximetry sounds for range identification, target transition identification, and latency to detect target transitions. Range identifications improved between 10 and 43 percentage points, target transition identifications improved between 26 and 40 percentage points, and latency to detect target transitions sped up by 7.6 seconds in one study³⁴  and by 22.3 seconds in another.³⁷ 

In two of the studies, participants were asked to estimate the absolute SpO₂ value.^34,36  Responses were considered correct if they were within 1 percentage point of the correct answer. Absolute accuracy performance increased by 17 percentage points in one study³⁶  and 21 percentage points in another.³⁴ 

In all the neonatal and adult studies shown in Figures 2 and 4, the enhancement took the form of a substantial discrete change in the quality of the pulse oximetry sounds in different ranges, rather than small incremental changes with each percentage change in SpO₂. To test whether there was any advantage in using small incremental changes, Collett et al.³⁸  compared enhancements using discrete changes across ranges with enhancements that gradually increased tremolo and the brightness of timbre within ranges as they got further away from the target range. With the discrete changes, range identification was 9 percentage points more accurate compared with the incremental changes, target transition identification was more accurate by 18.5 percentage points, and latency was faster by more than 5 seconds, underscoring the importance of discrete changes in the enhancements.

The benefits of acoustic enhancements appear to overwhelm the impact of distractions, such as noise or time-shared tasks.^17,20  Two of the neonatal studies shown in Figure 2 included a time-shared arithmetic task^29,30  and one included both the arithmetic task and OR noise.²⁷  Three of the adult studies shown in Figure 4 combined time-shared tasks, ambient pop music, and OR noise.^34,36,37  Markedly better performance was seen with the enhanced tones, despite these distractions. A further benefit is that the acoustic enhancements require only familiarization, rather than end user training, to be used effectively. A final benefit is that all participants, including clinicians, rated it easier to use the enhanced pulse oximetry tones than current pulse oximetry tones, and they rated their confidence in their judgments higher. Clinicians' qualitative comments after the simulation study³⁷  underscored the greater effectiveness and potential clinical usefulness of the enhanced tones.

Using acoustic enhancements to signal clinically significant saturation ranges is effective, whereas using them to convey blood pressure (BP) values seems not to be effective. There are no clear demonstrations that listeners monitor BP more effectively either when the pulse oximeter tone is modified,^39,40  when a separate BP-specific sound feature is added,⁴¹  or when multiple vital signs are mapped to different musical instruments.^42,43  Due to the complexity of monitoring three vital signs (heart rate [HR], SpO₂, and BP), responses to changes in HR and SpO₂ may become less accurate in an auditory display that includes all three rather than two vital signs. Even with just the two standard pulse oximetry parameters (HR and SpO₂), attempts to indicate high, target, and low HR ranges with further acoustic enhancements (i.e., in addition to the SpO₂ enhancements) have not led to a benefit.³⁰  Anderson and Sanderson⁴⁴  showed that changes in two or more acoustic dimensions can either reinforce or interfere with each other, depending on how they are mapped to parameters such as vital signs. Therefore, enhancing both pulse oximetry parameters or integrating a third vital sign would require careful design and evaluation.

The current variable-pitch pulse oximeter tone is already a powerful auditory display of clinically important information; therefore, altering it may seem superfluous. However, enhancing the variable-pitch tone to convey clinically relevant ranges and desaturation events can improve healthcare providers' ability to provide safe and effective care. Research has demonstrated substantial improvements in listeners' ability to detect clinical ranges, detect transitions into and out of the target range and other ranges, and identify absolute SpO₂ levels—all without negatively affecting the ability to perform time-shared tasks.³⁷  Moreover, during the COVID-19 pandemic, when personal protective equipment can restrict clinicians' ability to scan visual displays, the need for effective auditory displays is more important than ever.

Manufacturers should consider providing a means for clinicians to distinguish SpO₂ ranges with simple acoustic enhancements to the variable-pitch pulse oximetry tone. The enhancements should differentiate whether SpO₂ is above (where relevant) or below target, and to what degree, using discrete acoustic signatures for different ranges rather than smooth incremental changes across values. We envision that these ranges would be adjustable by clinicians, who would tailor them for individual patients just as they do for alarm thresholds.

Acoustic enhancements are effective in the absence of alarms, potentially reducing contributions to alarm fatigue, but alarms could still sound at selected levels if and as required. Acoustic features such as beacons, changes in timbre, tremolo, or formants could be used as enhancements, and they yield substantial benefits compared with other approaches, such as manipulating pitch step sizes or specialized training. Further, no intellectual property rights would be infringed—the concept is in the public domain. The sounds required can be adequately reproduced on the sound systems in most commercial pulse oximetry systems in use today. The enhancements suggested could avoid misinterpretations of the pulse oximetry signal that others have documented,^12,15  improve accuracy and timeliness of SpO₂ interpretation, and further reduce the workload of clinicians.

The authors acknowledge researchers who conducted studies of pulse oximetry sonification for honors or course credit (Marie-Lys Deschamps, Caitlin Browning, Sam Leav, and Hugh Clarke from The University of Queensland), software engineers (Birgit Brecknell and Ismail Mohamed from The University of Queensland), and collaborators who assisted with the simulation study (Neil Paterson from Queensland Children's Hospital and Isaac Salisbury, Felicity Burgmann, Thomas Davidson, and Garry Mann from The University of Queensland).