| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
 
|
|
|
||
|
|
© 2002 International Anesthesia Research Society
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Encouraged by the popularity of the pulse oximeter pulse-tone, we developed and tested an auditory display of six physiologic variables. The display consisted of a cardiovascular sound triggered by every heartbeat (conveying heart rate) and a respiratory sound triggered by every breath (conveying respiratory rate). Attributes of the cardiovascular sound were modulated to convey hemoglobin saturation and blood pressure, and those of the respiratory sound were modulated to denote end-tidal CO2 and tidal volume. Three display formats (auditory, visual, and combined) were compared. Fourteen anesthesia residents monitored dynamic displays of 6 variables to detect and identify 6 predefined events during 21 trials. An event occurred during each trial and the subject’s task was to detect when it started and then identify the type of event. Subjects detected every event. They detected events more rapidly with the combined display (10.4 s) than with the visual (12.8 s) or auditory (13.0 s) displays. Subjects correctly identified events least often with the auditory display (60% versus visual 88% and combined 80%). They correctly identified events more quickly with the combined display than with the visual display. We conclude that, with little training, clinicians can successfully detect and identify simulated clinical events using an auditory display of six variables.
IMPLICATIONS: We developed and tested an auditory display of multivariable clinical data. With little training, clinicians successfully used the display to detect and diagnose simulated critical events. This suggests that a multivariable auditory display could enhance intraoperative monitoring.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Continuous patient monitoring is one of the anesthesiologist’s most important intraoperative functions. Increasingly, electronic monitors automatically collect quantitative patient data and present them to the anesthesiologist through visual displays. Unfortunately, anesthesiologists do not reliably detect data from visual displays in the operating room (OR) (1). One reason is that anesthesiologists have other responsibilities besides watching monitors. Anesthesiologists spend less than one-third of their overall time looking at monitors, and this limits when they can receive information from visual displays (2–4).
Auditory displays may overcome this inherent limitation of visual displays. The auditory system is omnidirectional. Even when otherwise occupied, humans continue to hear, and analyze, auditory stimuli at a preattentive level (5). Thus, an anesthesiologist could continuously monitor information coming from an auditory display, even while performing other duties.
The variable-pitch pulse oximeter tone is an auditory display of arterial hemoglobin saturation and heart rate that has been widely accepted in the practice of anesthesia (6). Its efficacy has been demonstrated; anesthesiologists react faster to changes in hemoglobin saturation when using a variable-pitch pulse oximeter than when using a fixed-pitch oximeter (7).
Other cardiovascular and respiratory variables deserve to be monitored as closely as are saturation and heart rate. Therefore, we used a patented technique (8) to develop an auditory display of six important physiologic variables. The purpose of this investigation was to evaluate whether anesthesiologists could use this display to detect and diagnose critical events.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Overview
After institutional approval and with written informed consent, we evaluated 14 anesthesia residents’ ability to monitor vital signs as presented on a simulated physiologic monitor. Three monitor display formats (auditory, visual, and combined) were compared in a randomized within-subjects design. Subjects monitored dynamic displays of six physiologic variables to detect and identify predefined patterns of change (events). Their accuracy and speed in detecting and identifying these events were measured as indicators of display efficacy.
In each study session, a single subject was trained to use the three display types and to recognize six event types (Table 1). After training, the subject was tested during 21 trials, each lasting a maximum of 2 min. Subjects monitored six variables: heart rate, blood pressure, hemoglobin saturation, respiratory rate, tidal volume, and end-tidal CO2. At the beginning of every trial, each variable was set to a baseline value. These values were held constant for a random period of 15 to 80 s until an event began, at which time all affected variables transitioned synchronously from baseline to abnormal over 30 s. An event occurred during each trial. The subject’s task was to detect when the event started and then to identify the type of event.
|
Before training, the subject completed a questionnaire and received a screening hearing examination. After testing, each subject was asked to offer subjective evaluations of the displays. Study sessions lasted approximately 90 min and were conducted in a quiet office. Each subject received a gift certificate for his or her participation.
Displays
The visual display replicated a conventional physiologic monitor, with five waveforms and six numeric values (Fig. 1). Subjects watched it on a 17-in. monitor (1024 x 768 resolution) and input their responses using a mouse and on-screen controls. They heard the auditory display through stereo speakers placed at either side of the computer monitor.
|
The auditory display consisted of two independent auditory streams: the cardiovascular signal and the respiratory signal. These signals were "self-labeling" in the sense that they were designed to mimic real physiologic sounds: the cardiovascular signal was a low-pitched, repetitive, thudding sound reminiscent of a heartbeat; and the respiratory signal was a higher-pitched amplitude-modulated filtered noise, similar to the sound of breathing. A cardiovascular signal was triggered by each heartbeat and a respiratory signal was triggered by every breath. Thus, heart rate and respiratory rate had natural mapping onto the display. Respiratory rate also had a second effect, in that the respiratory sound became longer in duration as the respiratory rate slowed. This maintained a constant inspiratory-to-expiratory ratio as the respiratory rate changed. The other four variables had more arbitrary mappings to the auditory display. However, to enhance learning, cardiovascular variables modulated attributes of the cardiovascular signal and respiratory values modulated attributes of the respiratory signal. For example, blood pressure modulated the timbre of the cardiovascular signal and hemoglobin saturation modulated its pitch. End-tidal CO2 modulated the pitch of the respiratory signal, whereas tidal volume modulated both the bandwidth of a band-pass filter on the respiratory signal, and its duration. Thus, as tidal volumes increased, the respiratory signal became longer and more broadband and "noisy." Table 2 summarizes the mapping of physiologic variables to attributes of the auditory display. An audio clip and a more detailed description of the auditory display are available on the Internet at http://www.ahsc.arizona. edu/anesth/AuditoryDisplay.html.
|
Training
The training process was accomplished in three stages. In the first stage of training, subjects learned to use the displays. First, they interactively explored the three different display formats (auditory, visual, and combined). Using on-screen controls, they changed the values of each of the six variables, and saw and/or heard the changes in the visual and/or auditory displays. When they were ready, subjects attempted to identify the variable that was changing, as the computer randomly selected a display format and slowly changed a single variable. They received feedback on each response and were tested until they correctly identified the changing variable 12 consecutive times.
In the second stage of training, subjects learned to recognize the clinical events. First, they were asked to memorize a table showing the manifestations of each event (i.e., the bold-typeface items in Table 1). Then, they took two self-paced computerized tests to ensure that they could identify each event from its manifestations. Subjects received feedback on each response and were tested until they correctly identified 12 consecutive events.
During the third stage of training, subjects practiced the whole-task during 21 training scenarios. Each of seven event types (six "trained" and one "unfamiliar") was presented in random order, once each using each type of display. Subjects monitored vital signs and pressed an on-screen button when they first noted a change in any variable; this indicated their detection of an event. They then pressed a button, labeled with the name of the event, as soon as they deciphered the event type; this indicated their identification of the event. Subjects were instructed that the premature detection of an event or the misidentification of an event would be scored as an error; and that their first priority was to minimize error rate and their second priority was to minimize response time. To simplify event detection during training, variables were displayed without superimposed random noise factors (see next section). At the end of each trial, the subject received feedback regarding the true identity of the event.
Testing
In the testing phase of the study, subjects performed the experimental task during 21 testing scenarios. The experimental task, instructions to subjects, and data collection techniques were identical to those of the whole-task training exercise, with two exceptions. First, subjects did not receive feedback about the correctness of their responses. Second, a random noise factor (i.e., heart rate ± 3 bpm; blood pressure ± 10 mm Hg; hemoglobin saturation ± 0.5%; respiratory rate ± 1 breath/min; tidal volume ± 30 mL, end-tidal CO2 ± 1 mm Hg) was added to each variable on every update to prevent numerical fields on the visual display from remaining static while in their baseline state. The latter modification made it impossible for subjects to detect the onset of an event by simply detecting any change in a numerical field.
Statistical Analysis
Detection and identification scoring systems are summarized in Table 3. Detection and identification accuracy rates were calculated for each subject and display type. The effects of display type on detection accuracy rate and identification accuracy rate were analyzed by using one-factor within-subjects analysis of variance. Where appropriate, post hoc comparisons were made by using the Scheffé F test.
|
Two multivariate within-subject analysis of variance models were constructed: one for detection response time and the other for identification response time. These 14 x 3 x 7 models incorporated the following factors: subject, display type, and event type. Both models had cells of variable sample size, because response times were only calculated for correctly detected or identified events (see Table 3). Therefore, where appropriate, post hoc comparisons were made by using the Games-Howell test, which is robust with cells of unequal sample size (9). In all analyses, a criterion value of P < 0.05 was considered significant.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Demographics
Nine subjects were men and five were women. Their average age was 34.4 ± 5.1 yr (mean ± SD) with a range of 29 to 45 yr. Seven of the subjects were in their first year of clinical anesthesia training, three were in their CA2 yr, and four were in their CA3 yr of training. Subjects reported an average of 6.1 ± 1.5 h of sleep per night during the previous 2 nights; 4 admitted to feeling tired before starting the study. Seven of the subjects stated that they had not had any caffeine on the day that they were studied, whereas the remainder stated that they had had one to two caffeinated beverages. Twelve of the subjects stated that they wore prescriptive lenses and all but one of these wore them during the study. Hearing deficits above 25 decibels were detected in 6 of the subjects. No subject wore hearing aids.
Training
Display Training.
In the display training exercise, it took subjects between 12 and 50 attempts to record 12 consecutive correct responses. Subjects had no trouble identifying a change in a single variable when vital signs were presented visually (141/143 changes correctly identified when displayed visually, and 142/144 changes correctly identified when displayed visually and audibly). However, they did have difficulty identifying the changing variable when vital signs were only presented audibly (85/131 changes correctly identified when displayed audibly). Variables within the breath signal were more difficult to correctly identify than were those within the heart signal (52/63 heart sound changes correctly identified, versus 33/68 breath sound changes correctly identified). Subjects often confused blood pressure changes with saturation changes. They also often confused CO2 concentration, respiratory rate, and tidal volume changes with one another. However, they rarely confused cardiac changes with respiratory changes.
Whole-Task Training.
Subjects’ performances tended to improve over the course of the 21 whole-task training trials (Fig. 2). All performance measures tended to stabilize by the end of the training session, suggesting that the subjects reached a plateau in the learning curve.
|
During whole-task training, subjects performed better (i.e., higher accuracy and shorter latency) with the visual and combined displays than with the auditory display. This was an expected result, because the variables did not have superimposed random noise factors, which made it easy to visually detect any variable that was changing.
Testing
Subjects detected every critical event within its 2-min scenario. They prematurely detected the event in approximately 20% of the scenarios; this was not affected by display type. When the event was correctly detected, it was detected 20% more quickly with the combined visual and auditory display than with either display alone (Table 4).
|
Subjects were 30% less likely to correctly identify an event when vital signs were presented using the auditory display than when they were presented using either the visual display or the combined display (Table 4). When subjects correctly identified an event, they did so 20% more rapidly when using the combined display than with the visual display alone (Table 4).
Subjects’ performances tended to improve over the course of the 21 testing trials (Fig. 2). Detection and identification accuracy were better in the last seven trials than in the first seven trials with all display types. Detection latency was shorter in the last seven trials than in the first seven trials with the auditory and combined displays.
The bronchospasm event was detected more quickly than all other event types (8.6 ± 5.8 s vs 12.6 ± 6.6 s, P = 0.002).
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Our subjects did not diagnose critical events as accurately when using the auditory display as they did when using the visual or combined displays. This may be attributed, in part, to the relatively short duration of auditory display training. A number of findings support this contention. First, although the protocol was designed to provide thorough and extensive training, our subjects received approximately 1 hour of auditory display training, in contrast to their average of 20 months experience using clinical visual displays. In a previous study, clinically naive subjects who had no experience with clinical visual displays identified events more accurately when using an auditory display than they did when using a visual display (10). Second, subjects’ performances improved during the testing period, and the most consistent improvement, over the course of training and testing, occurred with the auditory display. Third, subjects detected the bronchospasm event more quickly than all other event types. Because the bronchospasm event was the only event in which saturation changed, this finding suggests that our subjects’ prolonged experience with pulse oximeters left them better trained at hearing pitch changes in the cardiovascular signals than at hearing other changes in the auditory display.
The auditory display was designed to be easy to learn. Where possible, physiologic variables were mapped to sound qualities in a manner that we predicted would be compatible with clinicians’ natural expectations. For instance, oxygen saturation was linked to the pitch of the heart signal because clinicians have already learned that mapping by using pulse oximeters. The timbre (ringing quality versus dullness) of the heart signal was chosen to represent blood pressure because the timbre of Korotkoff sounds decreases as a blood pressure cuff deflates. We designed the breath signals to become closer together and shorter in duration when the respiratory rate increased, to mimic how natural breath sounds change when breathing becomes faster (at a fixed inspiratory-to-expiratory ratio). Finally, we made the breath signals longer in duration and more raucous sounding as the tidal volume increased to emulate the natural sound of taking larger breaths. Only one variable, end-tidal CO2, did not have an obvious natural mapping to any sound. But, inasmuch as end-tidal CO2 is somewhat analogous to saturation, we linked it to the pitch of the respiratory signal.
Even though the auditory display was designed to be intuitive, after the initial training period, some subjects had difficulty interpreting what they heard. Overall, they misidentified the variable that was changing in 35% of the presentations, and the error rate was more frequent with the respiratory signal than it was with the cardiovascular signal. This suggests that the design of the display can be improved in a few ways. First, changes in sound property could be made more perceivable. For instance, several subjects commented that timbre changes were difficult to hear. Second, some mappings could be modified to make them more compatible with clinicians’ expectations. For instance, four subjects suggested that the pitch of the respiratory signal should increase as end-tidal CO2 concentration decreases, because they associate lower pitch with central nervous system depression and decreased respiratory drive. Future studies will address these ergonomic issues.
Certain differences between the test conditions used in this study and a real clinical situation limit the clinical applicability of the results. In this study, subjects had no concurrent responsibilities to distract them from the primary task whereas, in the clinical setting, anesthesiologists perform many concurrent technical and cognitive tasks. On average, anesthesiologists look at monitor displays <30% of the time during an operation. They look at monitors even less during high workload periods, such as induction and intubation (2,4). Because they do not constantly watch visual displays, anesthesiologists react more quickly to auditory stimuli than they do to equivalent visual stimuli in the OR. For instance, Morris and Montano (11) found that anesthesiologists detected 90% of auditory alarms in 3 seconds, whereas it took them 40 seconds to detect the same proportion of visual alarms. Thus, we anticipate that when loaded with other tasks, anesthesiologists will react more quickly to critical events with an auditory or combined presentation than with a visual presentation. However, this prediction was not tested in the current study because our subjects were able to look at the visual display without interruption.
This study was conducted in a quiet setting that was not representative of an OR. ORs are noisy. Their average sound level has been compared with that of a freeway (12). To be useful, an auditory display must be clearly audible above the background noise. This can be achieved by increasing the volume. But, it is preferable to select sound frequencies and patterns that are not masked by the sounds of an OR (13,14). Because the auditory display’s audibility was not evaluated in this study, it cannot be predicted how well it would be perceived in a noisy OR.
In this study, subjects were asked to diagnose events based solely on information coming from the auditory display. However, this is not how anesthesiologists typically use auditory displays, such as alarms, in the OR. Anesthesiologists cannot reliably identify OR alarms based on their sound (15). In responding to an auditory alarm, anesthesiologists often perform a visual search of the environment and look at visual information displays to identify and verify the alarm condition. The predominant function of alarms is to capture attention. The same can be said for the pulse oximeter’s auditory display. Its primary function is to alert the clinician to a change in heart rate or saturation. Thus, it was somewhat of a contrived task to have subjects diagnose events based solely on information coming from the auditory display. Although it is comforting to know that subjects usually made a correct diagnosis based solely on auditory information, in a natural setting, they would more likely refer to visual displays before making a diagnosis and performing interventions.
In summary, this study demonstrates that: 1) multivariable clinical data can be encoded in an auditory display, 2) with little training, clinicians can decipher and use an auditory display to detect and diagnose simulated critical events, and 3) detection and diagnosis of critical events is accelerated when an auditory display is added to a visual display of the same data. These results suggest that a multivariable auditory display could enhance intraoperative monitoring. However, such displays need to be tested under conditions more representative of the real clinical environment before they can be accepted for routine clinical use.
|
Acknowledgments |
|---|
Roche Laboratories provided gift certificates for the subjects’ participation as an educational gift.
The authors gratefully acknowledge the technical assistance of Liza Kantor.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
- Loeb RG. A measure of intraoperative attention to monitor displays. Anesth Analg 1993; 76: 337–41.[Web of Science][Medline]
- Loeb RG. Monitor surveillance and vigilance of anesthesia residents. Anesthesiology 1994; 80: 527–33.[Web of Science][Medline]
-
Allard J, Dzwonczyk R, Yablok D, et al. Effect of automatic record keeping on vigilance and record keeping time. Br J Anaesth 1995; 74: 619–26.
[Abstract/Free Full Text] - Weinger MB, Herndon OW, Zornow MH, et al. An objective methodology for task analysis and workload assessment in anesthesia providers. Anesthesiology 1994; 80: 77–92.[Web of Science][Medline]
- Wickens CD. Engineering psychology and human performance. 2nd ed. New York: HarperCollins, 1992.
- Schulte GT, Block FE Jr. Can people hear the pitch change on a variable-pitch pulse oximeter? J Clin Monit 1992; 8: 198–200.[Web of Science][Medline]
-
Craven RM, McIndoe AK. Continuous auditory monitoring: how much information do we register? Br J Anaesth 1999; 83: 747–9.
[Abstract/Free Full Text] - Fitch WTS. Sonification system using synthesized realistic body sounds modified by other medically important variables for physiologic monitoring. US patent number 5730140, March, 1998.
- Games PA, Keselman HJ, Rogan JC. Simultaneous pairwise multiple comparison procedures for means when sample sizes are unequal. Psychol Bull 1981; 90: 594–8.[Web of Science]
- Fitch WT, Kramer G. Sonifying the body electric: superiority of an auditory over a visual display in a complex, multivariate system. In: Kramer G, ed. Auditory display: sonification, audification, and auditory interfaces. Reading, MA: Addison-Wesley, 1994: 307–26.
- Morris RW, Montano SR. Response times to visual and auditory alarms during anaesthesia. Anaesth Intensive Care 1996; 24: 682–4.[Web of Science][Medline]
- Hodge B, Thompson JF. Noise pollution in the operating theatre. Lancet 1990; 335: 891–4.[Web of Science][Medline]
- Momtahan K, Hetu R, Tansley B. Audibility and identification of auditory alarms in the operating room and intensive care unit. Ergonomics 1993; 36: 1159–76.[Medline]
- Wallace MS, Ashman MN, Matjasko MJ. Hearing acuity of anesthesiologists and alarm detection. Anesthesiology 1994; 81: 13–28.[Web of Science][Medline]
-
Loeb RG, Jones BR, Leonard RA, Behrman K. Recognition accuracy of current operating room alarms. Anesth Analg 1992; 75: 499–505.
[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
 |
|
  B. Donmez, M.L. Cummings, and H. D. Graham Auditory Decision Aiding in Supervisory Control of Multiple Unmanned Aerial Vehicles Human Factors: The Journal of the Human Factors and Ergonomics Society, October 1, 2009; 51(5): 718 - 729. [Abstract] [PDF]
|
|
|
|
 |
|
 P. M. Sanderson, M. O. Watson, W. J. Russell, S. Jenkins, D. Liu, N. Green, K. Llewelyn, P. Cole, V. Shek, and S. S. Krupenia Advanced Auditory Displays and Head-Mounted Displays: Advantages and Disadvantages for Monitoring by the Distracted Anesthesiologist Anesth. Analg., June 1, 2008; 106(6): 1787 - 1797. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. O. Watson and P. M. Sanderson Designing for Attention With Sound: Challenges and Extensions to Ecological Interface Design Human Factors: The Journal of the Human Factors and Ergonomics Society, April 1, 2007; 49(2): 331 - 346. [Abstract] [PDF]
|
|
|
|
|
|
P. M. Sanderson, M. O. Watson, and W. J. Russell Advanced Patient Monitoring Displays: Tools for Continuous Informing Anesth. Analg., July 1, 2005; 101(1): 161 - 168. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Watson and P. Sanderson Sonification Supports Eyes-Free Respiratory Monitoring and Task Time-Sharing Human Factors: The Journal of the Human Factors and Ergonomics Society, January 1, 2004; 46(3): 497 - 517. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
