Advanced Auditory Displays and Head-Mounted Displays: Advantages and Disadvantages for Monitoring by the Distracted Anesthesiologist

Abstract

BACKGROUND: 

In a full-scale anesthesia simulator study we examined the relative effectiveness of advanced auditory displays for respiratory and blood pressure monitoring and of head-mounted displays (HMDs) as supplements to standard intraoperative monitoring.

METHODS: 

Participants were 16 residents and attendings. While performing a reading-based distractor task, participants supervised the activities of a resident (an actor) who they were told was junior to them. If participants detected an event that could eventually harm the simulated patient, they told the resident, pressed a button on the computer screen, and/or informed a nearby experimenter. Participants completed four 22-min anesthesia scenarios. Displays were presented in a counterbalanced order that varied across participants and included: (1) Visual (visual monitor with variable-tone pulse oximetry), (2) HMD (Visual plus HMD), (3) Audio (Visual plus auditory displays for respiratory rate, tidal volume, end-tidal CO₂, and noninvasive arterial blood pressure), and (4) Both (Visual plus HMD plus Audio).

RESULTS: 

Participants detected significantly more events with Audio (mean = 90%, median = 100%, P < 0.02) and Both (mean = 92%, median = 100%, P < 0.05) but not with HMD (mean = 75%, median = 67%, ns) compared with the Visual condition (mean = 52%, median = 50%). For events detected, there was no difference in detection times across display conditions. Participants self-rated monitoring as easier in the HMD, Audio and Both conditions and their responding as faster in the HMD and Both conditions than in the Visual condition.

CONCLUSIONS: 

Advanced auditory displays help the distracted anesthesiologist maintain peripheral awareness of a simulated patient's status, whereas a HMD does not significantly improve performance. Further studies should test these findings in other intraoperative contexts.

In Brief

IMPLICATIONS: Study tests whether auditory displays or head-mounted displays in combination with standard patient monitoring displays help distracted anesthesiologistsv to detect clinically significant events in simulated patients. Anesthesiologists were more likely to detect clinically significant events only when auditory displays were present.

The Anesthesia Patient Safety Foundation's alarm summit in 2004¹ emphasized that the variable-pitch pulse oximeter tone and the capnography auditory alarm must be on and audible if they are to prevent patient incidents. Recognizing the unique alerting ability of the auditory modality, several researchers have suggested that the advantages of the continuous variable-tone pulse oximetry signal could be extended to other vital signs.^2–7 Further researchers have noted that lightweight head-mounted displays (HMDs) of vital signs could also keep anesthesiologists continuously informed about the patient's state.^8–13 In the simulator study to be reported, we investigated advantages and disadvantages of advanced auditory displays versus HMDs for helping anesthesiologists maintain awareness of patient status and detect potentially adverse events when distracted from normal visual scanning.

Whether a display is visual or auditory may have a strong impact on monitor effectiveness.¹⁴ Most research on patient monitoring ignores the potential for auditory displays to inform rather than to alert or alarm. Auditory displays can reduce competition for the anesthesiologist's visual attention, letting him or her monitor patient vital signs in the background. Continuous sound may move into focal awareness if it signals an unexpected state but recede into peripheral awareness if it signals an expected state.^15,16

Several research groups have used desktop simulators to investigate patient monitoring with continuous auditory displays (sonification) of heart rate (HR), oxygenation (SpO₂), blood pressure (BP), respiratory rate (RR), tidal volume (V_T), and end-tidal carbon dioxide (ETCO₂).^4–6 Results indicate that participants' ability to report patient vital signs is either more accurate^4,5 or equally accurate⁶ with numerical visual displays than with sonification alone, but participants can perform time-shared tasks more effectively when monitoring with sonification.^5,6

In the above research, BP was either mapped to continuous auditory displays^4,5 or was not used.⁶ A continuous mapping of noninvasive BP (NIBP) to sound would be misleading because NIBP readings quickly become outdated.¹⁷ In an initial test with nonmedical participants, discrete auditory displays (earcons) signaling systolic and diastolic BP after each NIBP reading led to high accuracy, with the modal error size only one point on a nine-point scale.¹⁸

Surprisingly, no study has reported whether sonification or earcons draw attention to the start of potentially adverse events sooner than visual displays do, although this happens with auditory alarms.¹⁹ Achieving early but nonintrusive awareness of changes in patient state is a key goal of using sonification rather than auditory alarms. Such results have been reported for HMDs, which keep patient variables in the forward field of view and so reduce the workload of visual scanning. Prior simulator-based research shows that a HMD with a transparent monocular eyepiece reduced the average time anesthesiologists took to detect each of a series of four to six critical events embedded in an anesthesia scenario²⁰ and was well-accepted.²¹ Anesthesiologists using a HMD while handling major incidents switched gaze between visual monitors and a patient less often, spent more time looking towards the patient, and completed small procedures faster.^8–11,17 However, concerns that HMDs may cause unexpected events not to be seen have been noted in other domains and may carry over to the operative environment.^14,22

Current levels of patient safety in anesthesia are due to the multiple, partially redundant means that anesthesiologists use to maintain awareness of patient physiologic status, such as visual monitors, clinical observation, auditory and visual alarms, equipment checks and the like.²³ Demonstrating prospectively that any one monitor improves patient outcomes is difficult in both large-scale clinical trials^24,25 and in smaller simulator studies. In his well-known Swiss cheese model of accident causation, Reason²⁶ argues that adverse outcomes usually result from the rare alignment of “holes” in multiple layers of defense. Accordingly, we will test how well each display prevents unsafe situations when combined continuously with a risk factor. The conditions under which we test are therefore deliberately unrepresentative of normative or typical monitoring practices.

The risk factor we manipulate is demand for the anesthesiologist's attention that makes it difficult to scan visual monitors. Anesthesiologists occasionally experience strong demands on visual attention, as when inserting lines, performing blocks, or reading patient notes. To achieve standardized and reliable measurements to compare performance across displays, in our experiment we maintain the challenge to visual attention consistently within and across scenarios. We do not test the displays with undistracted anesthesiologists as a control condition because anesthesiologists' usual defensive monitoring strategies would be active and the unique contribution of the displays we wish to test could not be evaluated.

Based on the research reviewed above, we conjecture that further support for monitoring in the form of advanced auditory displays or HMDs may improve the distracted anesthesiologist's vigilance. In the following study, first we hypothesized that distracted anesthesiologists will detect a greater number of clinically significant events when visual monitors and variable-tone pulse oximetry are supplemented by further auditory displays and/or a HMD. Second, we hypothesized that distracted anesthesiologists will detect clinically significant events faster when monitoring is supplemented by auditory displays and/or a HMD.

METHODS

Participants

The study received Human Research Ethics Committee clearances from The University of Queensland and from Royal Adelaide Hospital. Participants were 16 anesthesiologists (seven consultants and nine residents) recruited from Royal Adelaide Hospital (14 males and two females). Participants had between 3.5 and 36 yrs of anesthesia experience after basic medical training (average 12 yrs). Participation was voluntary and participants were rewarded with a gift worth around AUD$50 that they chose from a small selection.

Design

Each participant experienced all four display conditions in a repeated measures design:

• Visual: Visual monitor plus variable-tone pulse oximetry, with visual and auditory alarms suppressed (condition serves as baseline for comparisons).
• HMD: As for Visual, plus HMD.
• Audio: As for Visual, plus respiratory sonification and BP earcons.
• Both: As for Visual, plus HMD, respiratory sonification, and BP earcons.

The order of display conditions was counterbalanced using a Latin Squares design. Scenarios were mapped to create 16 display orders so that each display condition was tested with all four scenarios (Table 1). As a result, across the experiment, participants' performance for each display condition was observed in Scenarios A, B, C, and D, and in each of the four serial positions, so there was no confounding of display condition with scenario or with serial position.

Table 1: Counterbalancing Table Used for Experiment

Apparatus and Stimuli

Simulator and Software

The study was conducted using a full-scale simulator at Royal Adelaide Hospital. The participant sat at a computer on which a distractor task was presented with his or her back to the resident (Fig. 1). The resident was played by one of two actors who were practicing anesthesiologists. The surgeon, anesthetic nurse, scrub nurse, and scout nurse were all played by actors trained for these roles. The scout nurse was introduced as a support person for the participant. She stood near the participant to aid recording of patient events, to answer any questions, and to take sound level readings.

Figure 1.: Layout of the operating room simulator, showing representative locations of equipment and actors.

To achieve a realistic operating room environment with tight control over the timing of scenario events, a METI ECS™ simulator was used alongside the Body 2003™ dynamic link library simulator. The patient's vital signs were initially generated using the Body™ simulator. Custom software manipulated the Body™ outputs to meet the scenario profiles and the result was displayed on a custom-designed patient monitor. Finally, matching scenarios were developed on the METI™ to complement the above-mentioned scenarios with physical changes on the manikin.²⁷ The simulated patient's vital sign information was broadcast over a TCP/IP protocol to a METI Waveform Display™ style standard visual patient monitor; auditory displays for variable-tone pulse oximetry, respiratory sonification, and BP earcons (Csound 5 scripts); the HMD, driven by a Sony U50™ palmtop PD; and the abstract classification distractor task.

Scenarios

Five 22-min scenarios were developed: one for training and four to use in the experiment itself (Table 2). Each scenario included three principal events (1, 2a, and 3a) which ran for between 1:08 and 2:17 (min:s) and which were usually first detectable from respiratory vital signs. Each event started when first indications of it appeared in the monitoring displays (Table 2) and ended either when the resident resolved the event by taking action or the end of the scenario was reached. In each scenario, the registrar made an unsuccessful attempt at resolution resulting in a further event, 2b or 3b.

Table 2: Scenarios Used. Each Scenario Lasts for 22 Minutes

Scenarios were completely deterministic and were closely controlled from the simulator control room (Fig. 1). Actors wore lightweight walkie-talkies with earpieces and their interventions with the manikin and simulated operating room equipment were cued by a scenario controller. Only the resident made interventions to the simulated patient, and did so strictly according to the scenario script. No participant indicated awareness that the scenarios were deterministic and noninteractive. Visual and auditory alarms were suppressed on all equipment so that the relative effectiveness of the auditory displays and the HMD could be fully assessed.

Displays

The respiratory sonification combines information about RR, I:E ratio, Vt, and ETCO₂ into one sound stream.^6,7 Flow of gas is represented by relatively pure rather than breath-like tones. As the patient's chest rises with inspiration and Vt accumulates, the sound pressure level (SPL) of the sonification increases to a maximum (at Vt of 500–600 mL, maximum SPL is usually around 55 dB(A)). As the patient's chest falls with an expiration, SPL decreases to zero. The inspiration sound is mapped to a musical pitch representing the most recent ETCO₂ level (when normal, between 349 and 358 Hz, or around F4) and the expiration to a pitch a musical minor third interval below the inspiration pitch (at approximately 0.841 of the first pitch, around D4 for normal). As ETCO₂ increases or decreases, the inspiration and expiration pitch together increase or decrease (maximum inspiration was 566 Hz for ETCO₂ of 69 mm Hg and minimum inspiration was 166 Hz for ETCO₂ of 0 mm Hg).

The NIBP earcons sound just after a NIBP reading has been taken and are scaled in pitch and duration to represent nine levels of systolic and diastolic BP.^18,28 Two initial tones (at 503 Hz, just above B4) provide a normal reference level (beacon) for systolic and diastolic BP. Two subsequent tones indicate systolic and diastolic BP by playing at one of four higher pitches (905 Hz and higher) and slightly shorter duration if hypertensive and at one of four lower pitches (304 Hz and lower) and slightly longer duration if hypotensive (Video clips 1 and 2; please see video clips available at http://www.anesthesia-analgesia.org).

The HMD was a Microvision Nomad™ with a transparent monocle. In the lower right part of the monocle there was a display of HR, SpO₂, BP, RR, Vt, and ETCO₂. To enhance equivalence with the auditory information for this experiment, no waveforms were provided. Figure 2 approximates what was seen through the HMD from the participant's location.

Figure 2.: Representation of the information displayed on the Microvision Nomad™ HMD monocle, taken from the perspective of the participant. Information appeared as luminous red characters in the right visual field.

Distractor Task

To test display effectiveness when the anesthesiologist is distracted, we developed a distractor task (Fig. 3). Abstracts of around 200 words from Anesthesia & Analgesia were displayed on a laptop screen. Participants classified each abstract according to (1) the kind of anesthesia application it covered, (2) the evidence class of the research, (3) the 2-year period of publication of the paper, and (4) the likely impact of the paper on anesthesia practice. Participants had up to 40 s to read each abstract and to enter their four classifications by clicking the corresponding buttons on a computer screen, but within the 40 s the task was self-paced. At 35 s, a bell rang to remind the participant to complete the classification task before the next abstract appeared.

Figure 3.: Abstract classification distractor task interface. Abstract is top left, the four classification category responses buttons are bottom left, performance feedback to participant is bottom right, and patient response buttons are top right.

Auditory distraction was not formally controlled, but consisted of normal conversation between operating room actors on a set of topics unrelated to the patient. In addition, certain surgical sounds such as diathermy were emulated. Visual, auditory, and cognitive distraction were therefore present throughout each scenario.

Questionnaires

A background questionnaire asked participants about their medical training, anesthesia experience, and musical training. Postscenario questionnaires asked participants how easy it was to monitor the patient in the condition they had just experienced, how quickly they believed they detected abnormal changes and returns to normality, how easy it was to do the distractor task, how much the distractor task interfered with patient monitoring and vice versa.

A postexperiment questionnaire asked participants to compare their experience of the four display conditions, ranking ease of monitoring with the four displays, preference for monitoring, and satisfaction with amount of training received for the HMD, the respiratory sonification, and the BP earcons. Participants also provided open-ended comments about what they liked and dis-liked about the HMD, the respiratory sonification, and the BP earcons. Further personality questionnaires were administered, whose results are not reported here.

Procedure

The experiment ran for around 4 h and had four phases: consent and orientation; training; scenarios; and final questionnaires.

Consent and Orientation

After reading an information sheet about the goals of the study, participants gave written in-formed consent to participate. They completed the background questionnaire. Then they were oriented to the simulated operating room and the functioning of its equipment for the purposes of the study.

Training

In a briefing room, the experimenter trained each participant on the tasks and displays they would experience. First, a PowerPoint™ presentation introduced the respiratory sonification, the BP earcons, and the information that would be seen on the HMD. The mapping of sound to vital signs and the range of values used were explained. The above introduction lasted around 20 min. Second, participants watched and listened to nine 2-min video clips of the visual patient monitor during patient monitoring, supplemented by variable-tone pulse oximetry, respiratory sonification, and BP earcons. The first two clips showed stable vital signs during spontaneous and mechanical ventilation. The remaining seven clips showed changes in vital signs associated with commonly recognized anesthesia incidents. Third, participants learned about the abstract classification task software and practiced classifying abstracts with the software for 5 min.

The experimenter and participant then moved to the simulated operating room for a 22-min training scenario (Table 2) using the respiratory sonification, BP earcons, and HMD in an approximation to the Both condition. The participants sat at the location indicated in Figure 1 with the anesthesia machine and visual monitor behind them, and performed the abstract classification task. Because there were no actors in the simulated operating room during training, the experimenter read out the events as they occurred to provide context for the vital signs. Half-way through the training scenario the experimenter removed the HMD so that the participant could practice monitoring without the vital signs in forward view.

Scenarios

Participants then completed the four 22-min scenarios. In each scenario, participants used a different display configuration. Instructions were as follows. Participants were told they were supervising a resident while the resident conducted anesthesia. Because the resident might be slow to respond to problems, the participant was asked to note if there was any clinically significant event, but to remain seated. The participant could turn to look at the patient monitor or make a comment or direct the resident any time they wished. However, they were asked not to neglect the abstract classification task and to give each abstract a considered response. The participant could note a clinically significant event or a return to a normal clinical state either by (1) clicking the relevant button on the computer screen, (2) noting it to the resident, or (3) noting it to the experimenter/scout nurse standing nearby.

During each scenario the actors' activities were directed by the controller over walkie-talkies. The actors conversed on a set of standard topics, but comments about the patient were confined to scripted comments cued by the controller and delivered at the same time in each scenario. The experimenter/scout nurse did not participate in general conversation. After each scenario, the participant and experimenter returned to the briefing room where the postscenario questionnaire was completed.

Final Questionnaires

After the final postscenario questionnaire, participants completed the postexperiment questionnaire and further inventories.

Statistics

Timestamped responses by participants to the abstract classification task and their identification of significant changes in patient state were captured by the abstract classification task software and stored in databases. Video data were captured in quad video display format (Fig. 4) onto DVD or hard drive and VHS tape. Digitized video data were analyzed using MacSHAPA²⁹ to determine event detections and detection times.

Figure 4.: Sample of video data captured in quad format. Within each view, top left is view of resident's activities from video camera 1 in Figure 1, top right is view from anesthesia machine to participant from video camera 2 in Figure 1, bottom left is view of participant from video camera 3 in Figure 1, and bottom right is a capture of the video monitor display at video source 4 in Figure 1. Upper figure: participant concentrates on the abstract classification task. Lower figure: participant checks activity of resident after hearing indications of a gas embolism event.

Detection and questionnaire responses were tested for significance using the Friedman Rank Test for k correlated samples, followed by tests of the HMD, Audio, and Both conditions against the Visual condition with Wilcoxon Signed Ranks tests with α = 0.05, using a Bonferroni correction for multiple tests (P = 0.0167). Detection speed and distractor task performance were tested for significance with Statistica™ 7 using repeated-measures ANOVAs followed by two-way Tukey HSD tests with α = 0.05. Detection and speed of detection for the unexpected resolution on either the second (2b) or third (3b) event (Table 2) were not included because participants often handled the unexpected resolution as a continuation of the previous event, creating a dependency and making it impossible to score separately. To test the effect of display on event detections, the proportion of the three major patient events detected in each scenario was calculated for each scenario and each participant. For detection times, two measures were used. First, the average detection time for the three major patient events in each scenario was calculated for each scenario and each participant, using data from detected events only (detection time for events detected). Second, the maximum event duration was added for undetected events to yield the minimum time in which a detection might have occurred (estimated minimum detection time).

For the distractor task, correct answers were determined as follows. (1) Abstract classification followed Anesthesia & Analgesia's classification for the six selected categories. (2) Research evidence class was rated by two anesthesiologists prior to the study. Abstracts were used in the study only if there was agreement or near agreement between the two raters on research evidence class. Participant responses were scored as correct if they matched the response of at least one of the raters. (3) Year of publication was determined from journal records and participants were given half points if their responses were in the neighboring period to the correct one. (4) Likely impact was considered too subjective to score, but was used to keep participants distracted.

Ambient Noise Levels

For eight of the 16 participants, readings of ambient SPL were taken before the onset of the first event and after the onset of each event. Readings were taken near the participant by the experimenter/nurse in the corner of the room (Fig. 1). Readings were taken with a hand-held SPL meter set to the A-weighted scale on the slow setting.

RESULTS

Event Detection

Participants' detection of the three principal events of each scenario was analyzed. Preliminary analyses determined that there were no main effects or interactions of participant expertise or musical training on detections or detection speed.

Proportion of Detections

Compared with their performance in the Visual condition ( = 52%, Md = 50%, event detection) participants detected significantly more events in the Audio condition ( = 90%, Md = 100%, P = 0.004) and in the Both condition ( = 92%, Md = 100%, P = 0.011) but, under the Bonferroni correction, not in the HMD condition ( = 75%, Md = 67%, P = 0.036). The direction of results was consistent across scenarios A, B, C, and D, with the HMD condition always better than the Visual condition, and the Both and Audio conditions always better than or equal to the HMD condition, and thereby always better than the Visual condition.

Detection Speed

In the first analysis, only the times for events that the participant detected were used (Table 3). No condition was significantly faster than the Visual condition, although the overall ANOVA shows a trend consistent with event detections, P = 0.09. In the second analysis, if a participant did not detect an event then the time until the resident resolved the event was used. Compared with performance in the Visual condition (Table 3), the estimated minimum detection time for all events was faster in the Audio condition (P = 0.003) and in the Both condition (P = 0.0005) but not in the HMD condition (P = 0.14).

Table 3: Summary of Basic Findings for the Four Display Conditions, Across all 16 Participants. Percentages in Parameter Column are Percentage Correct with Chance Responding to Distractor Task Questions. All Self-Reports are on 7-Point Likert Scales

Distractor Task Performance

The number of abstracts completed and the rate of correct answers are given in Table 3. Accuracy was greater than chance in all display conditions for abstract classification and research evidence class but not for year of publication. Performance did not differ significantly across display conditions for number of completed abstracts, accuracy of categorizing each abstract, or accuracy of estimating year of publication. For research evidence class classification, however, the Tukey HSD test indicated that performance was worse in the Both condition than in the Visual condition (P = 0.031).

Questionnaire Findings

Self-Report of Performance

Compared with their ratings for the Visual condition, participants rated it easier to monitor in the HMD condition (P = 0.002), the Audio condition (P = 0.004) and the Both condition (P = 0.003) (Table 3 for means). Participants believed they responded faster to patient events in the HMD condition (P = 0.011) and in the Both condition (P = 0.0002) than in the Visual condition. However, under the Bonferroni correction, their ratings for their response speed to patient events in the Audio condition just failed to be significantly higher than for the Visual condition (P = 0.021). Participants believed they detected the patient returning to normal faster than in the Visual condition only for the Both condition (P = 0.003). No further display differences were found for self-reports.

Preferences, Likes and Dislikes

Participants' postexperiment preferences were almost evenly divided across the advanced display conditions (Table 3). Only one participant preferred to monitor with the Visual condition. Participants' written comments on likes and dislikes were classified post hoc into categories. Counts were made for each category and the results are shown in Table 4.

Table 4: Likes and Dislikes About the Displays for the 16 Participants

Ambient Noise Levels

Average SPL across the Visual and HMD conditions was 57.5 dB(A) whereas across the Audio and Both conditions it was 59.5 dB(A). Minimum sustained SPL for the Visual and HMD conditions was 50 dB(A) and for Audio and Both it was 51.3 dB(A). Maximum sustained SPL for Visual and HMD conditions was 68.3 dB(A) and for Audio and Both it was 72.5 dB(A). Occasional peaks were observed into the 80–93 dB(A) range in all conditions. When compared with monitoring using Visual displays plus audible pulse oximetry, respiratory sonification increased sustained minimum SPL by + 1.3 dB(A), average SPL by + 2.0 dB(A), and sustained maximum SPL by + 4.2 dB(A). Due to the lower statistical power and high variability in readings, a statistical test of these data was not attempted.

DISCUSSION

First, anesthesiologists detected a greater number of clinically significant events when advanced auditory displays were present, but not when the HMD was used. Adding the HMD when the advanced auditory displays were already being used did not lead to further gains. Second, detection time for events successfully detected was not significantly faster with the advanced auditory displays when using only the events that anesthesiologists successfully detected. If the maximum times allowed for undetected events were included, however, then estimated minimum possible detection time was significantly faster when advanced auditory displays were present.

These results extend previous findings^4–6 by showing that even when used alongside visual displays and variable-tone pulse oximetry, respiratory sonification and BP earcons improve detection of significant events when the anesthesiologist is distracted. In contrast to previous findings²⁰ the HMD does not significantly increase event detection, probably because it does not signal changes as effectively as auditory displays do when the anesthesiologist is distracted. Rated ease of use and preference were higher for the HMD than performance with it would suggest, possibly because users are unaware of events they missed.

Advanced auditory displays dramatically increased the percentage of detected events from a median of 50% in the Visual condition to a median of 100% in the Audio and Both conditions respectively. Any failure to detect an event produced a detection delay of anywhere from 1:08 to 2:17 (min:s) (the range of event durations in the scenarios). Such delays are equivalent to those found with use of fixed-tone versus variable-tone pulse oximetry (1:37 delay)³⁰ and they exceed the average delay anesthesiologists showed in detecting an anoxic gas supply in the absence of pulse oximetry and visual capnography (0:53 delay).³¹ The significant differences in the frequency of the delays observed indicate that informative audible information can improve the response of the distracted anesthesiologist by minutes rather than seconds.

When detection time for events detected (Table 2) was analyzed, there was only a trend in the direction predicted. Participants sometimes waited to complete the current abstract before noting an emerging patient event and initiating an exchange with the resident. This strategy may have attenuated real differences between display conditions in event detection times.

The abstract classification distractor task results suggest that participants were fully engaged in the task almost equally across display conditions. Improvements in detecting events were therefore not due to differing degrees of prioritization of the distractor task. The slightly better performance for research evidence class classification in the Visual condition compared with the Both condition suggests that identifying research evidence class was a resource-limited task³² that yielded better results with greater effort. In the Both condition, the displays made participants more aware of the patient so they could not give as much effort to the distractor task as in the Visual condition. Clinically, it is desirable for a display to make it harder to block out awareness of the patient.

The results suggest that, when combined continuously with the risk factor of distraction, advanced auditory displays may draw attention to unsafe situations when the anesthesiologist is otherwise distracted. The results do not indicate that auditory displays lessen the need for visual monitors, clinical observation, alarms, or equipment checks, or that anesthesiologists might permit higher levels of distraction. Such conclusions would lessen the benefits of the displays and open new areas of risk.^26,33

Tolerance of Advanced Displays

Participants felt adequately prepared to use the advanced displays and saw benefits in all of them. Concerns with the weight and bulk of HMDs will reduce as technology improves. Concerns with the added noise of auditory displays require further investigation. Given the overall variation in SPL noted, these are not large increases. Sound patterning, social effects, and perceived control are usually just as critical to tolerance.^34,35 We are investigating the advantages and disadvantages of optional earpieces that can bring the use of the advanced auditory displays completely under the anesthesiologist's control.

Limitations and Future Research

First, the advanced displays were all unfamiliar and the relatively short exposure may not have produced a fully representative view of their effectiveness in practice. Longer simulator studies and clinical evaluations are needed. Second, although our results demonstrate the effectiveness of advanced auditory displays when anesthesiologists are distracted, the displays should be tested for usefulness and tolerance during normal monitoring. Third, tests with visual and auditory alarms enabled would compare benefits of the continuous informing property of advanced displays with the discrete alerting property of alarms. Fourth, the scenarios are not fully representative of situations in the operating room in which advanced displays would be used. At 22 min each, the scenarios were relatively short and eventful, and they were oriented towards respiratory events. Fifth, performance in the HMD condition might have been significantly better than in the Visual condition with greater statistical power and if waveforms had been used. Sixth, participants' performance may have been affected by awareness of the treatment conditions and their awareness of being observed.

Finally, performance with the HMD may have been better, and performance with auditory displays worse, if the participant had been distracted solely by auditory information. However, preliminary results of a laboratory study using the scenarios and displays used here, but testing with either a visual or an auditory distractor task, suggest no selective benefits of intermodal distraction and no selective costs of intramodal distraction. Further, prior research indicates that auditory distraction in the form of music imposes no costs on monitoring with advanced auditory displays³⁶ due to auditory streaming effects exploited in the design of such displays³⁷ and that music can even improve detection of trends.³⁶ Our next step is to test these displays in the clinical context.

ACKNOWLEDGMENTS

We acknowledge Queensland Health's Skills Development Centre for access in our preparatory work, with special thanks to Lucas Tomczak, Daniel Host and Andi Thompson. Our grateful thanks to the Royal Adelaide Hospital staff members who participated in the study. We also thank Philippe Lacherez for help with preparation of materials for the abstract classification distractor task and Matt Thompson for help with video analysis.