Current physiological monitoring technologies collect large amounts of data at the bedside, which are typically conveyed to the anesthesiologist by auditory and visual displays. The exponential growth in the number of monitored and visually displayed variables has dramatically increased the demands on a clinician's attention. Patients are placed at risk if surveillance fails to identify a visual cue or if an auditory alarm is missed or ignored.
The use of a vibro-tactile display offers one possible solution to the problem of visual and auditory overload.1,2 Vibro-tactile displays harness the body's skin to convey information from physiological monitors to the anesthesiologist. The tactile modality has communication attributes that make it suitable for the operating room environment: it is personal (sensed only by the person wearing it); obligatory (unable to prevent stimulation unless device removed); and does not interrupt other individuals in the environment.3 Devices are lightweight and do not require a cable connection that could present an inconvenience in the operating room. Visual displays permit fixation on one variable to the potential detriment of the emergency management of the patient. Therefore, a display system that combines both tactile and visual input may reduce visual fixation by providing a second source of information.2
The potentials for vibro-tactile displays in the field of anesthesiology are 2-fold. First, tactile communication can provide subtle cues, rather than outright alarms, to indicate physiological changes. It does not distract from other forms of communication or patient interaction, nor does it disturb other individuals in the clinical environment. Only the anesthesiologist receives this information. Second, vibro-tactile communication may facilitate attention-switching between monitored variables.4
We have previously developed a vibro-tactile display, assessed optimal vibro-tactile patterns and positions, and shown that vibro-tactile communication provides greater accuracy in recognizing an array of physiologically meaningful alarms in comparison with standard auditory communication.5 Studies in a human patient simulator have shown that a vibro-tactile display is capable of reducing the time needed to diagnose and treat a critical incident.6 In this investigation, we evaluated the ability of anesthesiologists to recognize vibro-tactile cues in a real-time clinical environment. We hypothesized that anesthesiologists would correctly identify the location (variable), level, and direction of change communicated to them through the vibro-tactile belt 90% of the time; that anesthesiologists would find the vibro-tactile belt to be usable in a clinical environment; and that anesthesiologists would find the vibro-tactile belt to be wearable in a clinical environment.
Approval for this study was obtained from the University of British Columbia Clinical Research Ethics Board and from The British Columbia Children's and Women's Hospital Research Review Committee (both, Vancouver, B.C., Canada). Informed written consent was obtained from all anesthesiologist subjects. Clinical Research Ethics Boards determined that patient consent was not required for this study. We used a prospective observational design. Anesthesiologists were required to have a minimum of 3 years of clinical experience before participating in this study.
The vibro-tactile belt prototype used in this investigation (Fig. 1) was developed by the Electrical and Computer Engineering in Medicine group at The University of British Columbia (Vancouver, B.C., Canada) in conjunction with the Product and Process Applied Research Team at the British Columbia Institute of Technology (Burnaby, B.C., Canada). It consists of 6 vibro-tactile motors operated using a 9-V DC source, vibrating at a frequency of 100 Hz. A switch on the belt buckle provides 3 options for vibro-tactile intensity. In this study, anesthesiologists could choose the desired level of vibro-tactile intensity by adjusting the switch on the buckle. We used 4 of the 6 possible tactor locations around the waist (Fig. 2) to encode 4 physiological variables: mean noninvasive arterial blood pressure (NIBPmean), peak airway pressure (Ppeak), end-tidal carbon dioxide partial pressure (EtCO2), and expired minute ventilation (MVexp). The 2 directions of change (increasing or decreasing) and 2 possible levels of change7 (level 1: “minor change,” and level 2: “major change”) were represented by 4 rhythms (Fig. 2). The overall encoding scheme was based on the results of previous studies using spatial locations8–10 and patterns of vibration.11 The schema was selected to achieve a priority for identification of physiological variable (location of the tactor) and direction of change (duration of initial stimulus), followed by level of change (pattern of stimulus). The vibro-tactile belt was connected with a Bluetooth device to a software tool using the change-point detection algorithms12–14 to identify changing trends in NIBPmean, Ppeak, EtCO2, and MVexp. This type of monitoring, dependent upon changes in physiological trends, is known as context-sensitive monitoring (see below).
To reduce the rate of false alarms associated with fixed threshold alarms and to improve accuracy in detecting trend changes in physiological signals, we have previously proposed the use of context-sensitive change detection in physiological monitoring.15,16 In this monitoring method, an alert is based not only on the instantaneous value, but also is dynamically dependent on all the previously recorded physiological data available before an observation. Predictions of the next observation or observations further in the future are calculated on the basis of information from all the previous data. The prediction is compared with new observations to identify when a significant change in trend is initiated (called a change-point).
Change-Point Detection Algorithms
The 2 key elements in a change-point detection algorithm are the signal model used for generating predictions and the technique used for evaluating the difference between the predictions and the new observations. The cumulative sum (CUSUM) test, a method widely adopted in process control and monitoring, was used to determine whether the signal measurements significantly deviated from the predictions. In the CUSUM test, the differences between successive observed values and a target value (a predicted value in this case) are accumulated; the cumulative sum (CUSUM) is then compared with a threshold to detect change points. The CUSUM test is an optimal change detection method in the sense that, given Gaussian white noise, the test can reach the highest sensitivity and specificity within an allowable detection delay.17
The different physiological variables were described using different predictive models based on their temporal characteristics. NIBPmean and Ppeak were modeled using an autoregressive regression integrated moving averaging (ARIMA) (0,1,1) model as described in detail elsewhere.14 The model predicts the next observation using an exponentially weighted moving average (EWMA) of the historical data. A standard CUSUM test is used with this model to detect changes in the NIBPmean and Ppeak.
The trend signals of EtCO2 and MVexp were described using a dynamic linear model.12,13 In this model, both the signal level and the incremental rate of trend change are treated as dynamic processes; the uncertainties caused by measurement noise, artifacts, and physiological variations (e.g., respiration) were represented as additive Gaussian noise.
We used an adaptive Kalman filter7 to estimate the true instantaneous level and the incremental rate and generate predictions in the presence of noise. The adaptive Kalman filter extracts the level of the noise and the degree of signal variability. The target values and the thresholds of the CUSUM test were updated according to the estimates of the Kalman filter with each new observation. In this way, the change detection process was adapted to the context of historical observations.
Each algorithm is based on the data collected from a single sensor (e.g., NIBPmean). The algorithms are implemented in the Java® programming language and have been tested in both simulated and clinical studies.15,16
Before participating in the clinical evaluation, anesthesiologists were to undergo a training phase to familiarize themselves with the tactile belt and the complete set of 16 tactile stimuli (4 physiological variables, 2 directions, and 2 levels of alerts). Using a graphical user interface (Fig. 3), subjects were to be administered different stimulation patterns mapped to the 16 buttons on the interface. When the anesthesiologist felt familiar with the stimulation patterns and mapping and had activated each button at least once, they were to proceed to a post-training quiz. The training phase was to end when a subject obtained a minimum accuracy rate of 75% on the quiz. Subjects were to be permitted to repeat the quiz as necessary, until a maximum of 30 minutes passed from the start of the training period. Subjects requiring more time were to be excluded from further testing.
After training, the software tool running the signal processing algorithms was to be activated in real-time during anesthesia to enable context-sensitive monitoring of the 4 variables, displayed on a touch screen monitor alongside current Datex S/5 physiological monitors (Datex-Ohmeda, Helsinki, Finland) using a serial RS-232 interface. Once a patient's physiological state was considered stable (usually 5 to 10 minutes after induction of anesthesia), the anesthesiologist was to activate the vibro-tactile belt. When the algorithms detected change points, the belt would vibrate with the appropriate tacton (waist location and corresponding rhythm). Using the touch screen monitor, the anesthesiologist was to then enter the vibro-tactile message that he or she had received by first identifying the variable, and then identifying the level and direction of change. Subjects were to discontinue the evaluation once the surgical procedure had been completed. Each participant was to evaluate the vibro-tactile belt for a maximum of 5 cases.
Usability and Wearability
A modified Post-Study System Usability Questionnaire (PSSUQ)18 was to be administered to participants after their first use of the vibro-tactile belt. The PSSUQ rates the usability of the application across 4 factors: overall, system usefulness, information quality, and interface quality. The questionnaire used in this study was modified to exclude 3 not applicable items that stated “Whenever I make a mistake using the system, I recover easily and quickly”; “It is easy to find the information that I need”; and “The information provided with the system is easy to understand.”18 Participants were to complete a paper version of the survey, responding to each question by ranking their agreement with 16 statements on an 8.5-cm-long numerical scale, with possible responses ranging from 1 (strongly disagree, poor usability) to 7 (strongly agree, good usability). Each question also provided a “not applicable” option and a freeform text section for detailed feedback. The score for each factor was to be determined by averaging the responses to the appropriate questions.18 In addition, a 6-item wearability questionnaire (developed in discussion with users during previous studies) was to address the comfort of the vibro-tactile belt and provided space for freeform text responses. Trends were to be identified from the subjective feedback provided.
The total sample size projection for this study was based on a targeted accuracy of 90% in decoding the correct variable, level, and direction of change. To predict the accuracy within 5%, we required 250 events. Assuming independent random sampling of cases and anesthesiologists and an average of 5 events per case, we anticipated studying 50 surgical cases. The results are reported as mean (SD) and percentages. Participant responses were classified with a graphical confusion matrix representation.
Identification of Vibro-Tactile Stimuli
Seventeen anesthesiologists (Table 1) evaluated the display during 57 cases of minimum scheduled length of 1 hour. All cases were performed in pediatric patients undergoing general anesthesia. All anesthesiologist subjects successfully completed the training quiz within the allotted 30-minute timeframe. The belt was operational for a mean (SD) duration of 75 (41) minutes per case. Seven cases were excluded from analysis: 1 because the anesthesiologist subject accidentally interrupted the software, and 6 because of disconnection of the Bluetooth device.
A total of 530 stimuli occurred during this study. Four hundred twenty-nine (81%; confidence interval [CI], 77% to 84%) of all stimuli were decoded by the anesthesiologists (Table 2). The mean (SD) number of decoded stimuli per case was 8.6 (6.7). Subjects accurately identified the complete vibro-tactile pattern in 384 (89.5%; CI, 86%–92%) of these decoded stimuli. The physiological trend, the direction of change, and the level of change were correctly identified for 419 (97.7%; CI, 96%–99%), 407 (94.9%; CI, 92%–97%), and 401 (93.5%; CI, 91%–96%) of decoded stimuli, respectively. Errors in decoding the physiological variable were similar for all 4 variables (NIBPmean, Ppeak, EtCO2, and MVexp) (Fig. 4), but errors in decoding the vibro-tactile pattern occurred most frequently when distinguishing between level 1 and level 2 alerts (Fig. 5). Nineteen percent of all stimuli were not decoded by the anesthesiologists. Unrated stimuli occurred a median (range) of 15 (0.6 to 740) seconds from another tactile alert. There were 11 false positive events.
Usability and Wearability
Fourteen of the participating anesthesiologists completed a modified poststudy usability questionnaire (PSSUQ).18 The mean (SD) overall usability rating was 4.8 (1.4) of a maximum score of 7 (Table 3). Thirteen questionnaires were scored on the basis of the complete set of 16 questions. One questionnaire was missing a response to 1 item and was scored only on the basis of the remaining 15 items. Thirteen of the participating anesthesiologists completed the wearability questionnaire. Subjective feedback was positive, with 12 subjects reporting overall satisfaction with wearing the tactile belt (Table 4). Most criticism listed in the freeform text was directed toward the fit of the belt. Clinicians subjectively reported that decoding of messages became easier, with less mental effort, the longer the device was used.
In this prospective clinical evaluation of a vibro-tactile belt, we have shown that anesthesiologists were able to decode all 3 components (physiological variable, direction of change, and level of change) of the vibro-tactile messages with a high degree of accuracy (89.5%) in a real-time clinical environment.
In a previous preclinical study by our group,19 subjects interpreted tactile patterns with an overall accuracy of 79.9%. In these experiments, 6 tactor locations were used to encode 6 physiological variables and 4 rhythms (which differed from those used in the current study) for a total of 24 patterns (16 here) under simulated workload (subjects were requested to monitor 2 physiological variables, heart rate and systolic blood pressure, on a visual display as a secondary task). Even though the workload of participants was probably higher in our study (real-life situation in comparison with a simulated one), the overall identification rate was higher. This was likely due to a reduced number of physiological variables monitored by the belt (4 rather than 6); improvement of the location identification rate (97.7% in comparison with 91.0%); and a better design of the tactile rhythms improving the direction and level of change (94.9% and 93.5%, respectively, in comparison with 86%).
The vibro-tactile belt also received positive responses from anesthesiologists who completed usability and wearability questionnaires. Participant feedback indicated that half of participating clinicians would be comfortable wearing a waist-mounted belt throughout their time in the operating room. On the basis of the freeform feedback, this number would increase greatly provided that the belt could be adjusted to accommodate varying waist sizes.
The high level of decoding accuracy in this investigation occurred despite only 30 minutes or less of preclinical training with the belt, vibro-tactile patterns, and user interface. The error rates for the identification of the physiological variable using location was low, consistent with our previous findings.19 As expected, confusion with the identification of direction and level of change increased with the use of patterns utilizing duration and frequency of stimulation.11 It is likely that increased familiarity with the vibro-tactile display would further improve the clinicians' understanding of the vibro-tactile messages. Clinicians subjectively reported that decoding of messages required less mental effort the longer the device was used.
The sense of touch has a set of affordances that render it suitable for information display in the operating room. As with audition, touch is obligatory and unidirectional, such that any person equipped with a tactile display can receive its signals regardless of the spatial orientation of attention. Unlike audition, touch more easily supports privatization of information because it is a proximal sense; only those who are in physical contact with tactile display devices have the ability to receive its messages. Tactile messages can be reliably communicated to the anesthesiologist in a fashion that does not unnecessarily distract or annoy other operating room personnel. Touch has a superior ability to process stimuli that require discrimination in both the spatial and temporal signal domains. The spatial acuity of tactile perception is at least as high as audition (depending on body location), and touch has an advantage in judging the temporal durations, rates, and rhythmic patterns of stimuli in comparison with vision.20 This makes it an ideal channel for relating information that naturally maps to directions (increase/decrease), or for information that has a temporal dimension such as physiological monitoring. Although small changes in a repetitive pattern are more easily detected with audition5 (e.g., the change in pitch in pulse oximetry sound), habituation is less marked with tactile stimulation.21 A recent study used a tactile-based monitoring system to measure anesthesiologists' mental workload.22 Although this display was primarily used as a research tool, this application may provide a useful method to adjust the performance of a monitoring system on the basis of the workload of the user.
The process of using a tactile display to communicate events and the subsequent identification of the events by the anesthesiologist is a complex process. When a patient event occurs, the physiological monitor sends a stream of physiological data to a processing unit that must extract the key information (i.e., identify an abnormal condition) and encode this into a tactile pattern. This pattern must be transmitted through a wire or wireless connection to the tactile display. The display then transmits (displays) the appropriate tactile stimulus to the anesthesiologist to alert him to the abnormal patient condition. Upon detecting the tactile stimulus, the anesthesiologist then must decode the information embedded in it and respond to the event. To successfully complete the full communication process, it is necessary to design stimuli that are easy to identify and easy to correlate with their associated physiological meanings during the decoding phase. Previous studies have investigated different tactile display modalities (vibration and electrical) at different body locations (forearm,9,23 wrist,9 and abdomen8,10). In previous work,11 we have examined different encoding methods and their effects on the reception process, and have proposed an optimal method to design rhythm-based stimuli. We also have evaluated the ability to associate different stimuli with physiological events with different tactile display prototypes. A display using tactons to represent alphanumeric characters proved easier to learn and less affected by distraction; however, the display required a larger array of stimulating locations.19 Further studies will be necessary to determine the optimal design of a clinically useful display.
This study has a number of limitations. We have evaluated the multiple steps in the tactile communication process simultaneously, including event detection (context-sensitive monitoring), message encoding, communication, decoding, and interpretation (Figure 1 in our previous preclinical article19). Although some of these individual steps have been investigated in the laboratory, they have not been studied in the clinical environment. Future work will be directed toward optimizing each step in the communication process in the clinical environment.
In addition, 19% of all alerts were not rated by the anesthesiologists. This lack of response likely is a result of user burden when different alerts were generated simultaneously. Future work aimed at integrating the information from different sensors related to the same clinical event would reduce the number of alerts. For example, an increase in ventilation may trigger an alert for an increase in MVexp, a decrease in EtCO2, and an increase in Ppeak. A single alert that ventilation has increased would avoid multiple alerts. A further limitation to this investigation was the frequent failure of the Bluetooth device. This was due primarily to the inability of the system to automatically reconnect if the clinician strayed outside of the geographical range of the device. More success may be achieved with an alternative Bluetooth device or a more robust wireless communication technology such as ZigBee (ZigBee Alliance, San Ramon, CA). This high failure rate in communication is a significant safety concern. Although the anesthesiologist was advised both by a message on the display and by an indicator on the belt that communication had been lost, a more robust solution would be more desirable. Finally, adoption of novel technology should ideally be preceded by evidence of improved clinical outcomes. Randomized controlled trials provide an objective and comprehensive measure of clinical outcomes, but require large numbers of participants and long study periods.24 In this initial investigation, we have instead evaluated the subjective assessment by clinicians of usability and the usefulness of the tactile belt.
Last, the relatively small sample size and assumption of independence of variables (such as type of surgery and anesthesiologist) must be considered when interpreting the results of this investigation. It is important to emphasize that the tactile belt tested in this investigation was not designed to replace the clinicians' need to be continuously present and attentive to the patient, but instead to assist the anesthesiologist in rapidly processing the vast amount of information available from monitoring equipment and to convey this information in a meaningful manner so that rapid intervention can occur. Information transferred via tactile modalities does not require diversion of visual attention from the patient and, unlike auditory alarm systems, does not pollute an already noisy operating room environment.25 The tactile display evaluated in this study showed a high degree of acceptance by anesthesiologists and accuracy in the decoding of clinical events.
Name: Maryam Dosani, BSc.
Contribution: This author helped conduct the study, analyze the data, write the manuscript, and data collection.
Attestation: Maryam Dosani has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Kate Hunc, BA, MA.
Contribution: This author helped analyze the data, write the manuscript, and data collection.
Attestation: Kate Hunc has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Guy Dumont, PhD.
Contribution: This author helped design the study, conduct the study, and write the manuscript.
Attestation: Guy Dumont has seen the original study data and approved the final manuscript.
Name: Dustin Dunsmuir, MSc.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Dustin Dunsmuir approved the final manuscript.
Name: Pierre Barralon, PhD.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Pierre Barralon approved the final manuscript.
Name: Stephan K. W. Schwarz, MD, PhD, FRCPC.
Contribution: This author helped design the study and write the manuscript.
Attestation: Stephan Schwarz approved the final manuscript.
Name: Joanne Lim, MASc.
Contribution: This author helped design the study, conduct the study, and write the manuscript.
Attestation: Joanne Lim has seen the original study data and approved the final manuscript.
Name: J. Mark Ansermino, MB, BCh, MSc, FFA.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Attestation: J. Mark Ansermino has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
This manuscript was handled by: Dwayne R. Westenskow, PhD.
1. Jones LA, Sarter NB. Tactile displays: guidance for their design and application. Hum Factors 2008; 50: 90–111
2. Ferris TK, Sarter NB. Cross-modal links among vision, audition, and touch in complex environments. Hum Factors 2008; 50: 17–26
3. Sanderson P. The multimodal world of medical monitoring displays. Appl Ergon 2006; 37: 501–12
4. Calvert GA, Brammer MJ, Iversen SD. Crossmodal identification. Trends Cognit Sci 1998; 2: 247–53
5. Ng JY, Man JC, Fels S, Dumont G, Ansermino JM. An evaluation of a vibro-tactile display prototype for physiological monitoring. Anesth Analg 2005; 101: 1719–24
6. Ford S, Daniels J, Lim J, Koval V, Dumont G, Schwarz SKW, Ansermino JM. A novel vibrotactile display to improve the performance of anesthesiologists in a simulated critical incident. Anesth Analg 2008; 106: 1182–8
7. Yang P, Dumont G, Ansermino JM. A Cusum-based multilevel alerting method for physiological monitoring. IEEE Trans Inf Technol B 2010; 14: 1046–52
8. Barralon P, Ng G, Dumont GA, Schwarz SKW, Ansermino M. Development and evaluation of multidimensional tactons for a wearable tactile display. Proceedings of the 9th Conference on Human-Computer Interaction with Mobile Devices and Services: MobileHCI 2007. Singapore, September 11–14, 2007: 186–9
9. Ng G, Barralon P, Dumont G, Schwarz SKW, Ansermino JM. Optimizing the tactile display of physiological information: vibro-tactile vs electro-tactile stimulation, and forearm or wrist location. Conf Proc IEEE Eng Med Biol Soc 2007; 4202–5
10. Cholewiak RW, Brill JC, Schwab A. Vibrotactile localization on the abdomen: effects of place and space. Percept Psychophys 2004; 66: 970–87
11. Barralon P, Ng G, Dumont GA, Schwarz SKW, Ansermino JM. Design of rhythm-based vibrotactile stimuli around the waist: evaluation of two encoding parameters. IEEE Trans Syst Man Cybern A 2009; 39: 1062–73
12. Yang P, Dumont GA, Lim J, Ansermino JM. Adaptive change point detection for respiratory variables. Conf Proc IEEE Eng Med Biol Soc 2005; 780–3
13. Yang P, Dumont G, Ansermino JM. Adaptive change detection in heart rate trend monitoring in anesthetized children. IEEE Trans Biomed Eng 2006; 53: 2211–9
14. Yang P, Dumont G, Ansermino JM. An adaptive CUSUM test based on a hidden semi-Markov model for change detection in non-invasive mean blood pressure trend. Conf Proc IEEE Eng Med Biol Soc 2006; 1: 3395–8
15. Ansermino JM, Daniels JP, Hewgill RT, Lim J, Yang P, Brouse CJ, Dumont GA, Bowering JB. An evaluation of a novel software tool for detecting changes in physiological monitoring. Anesth Analg 2009; 108: 873–80
16. Dosani M, Lim J, Yang P, Brouse C, Daniels J, Dumont GA, Ansermino JM. Clinical evaluation of algorithms for context-sensitive physiological monitoring in children. Br J Anaesth 2009; 102: 686–91
17. Basseville M. Detecting changes in signals and systems—a survey. Automatica 1988; 24: 309–26
18. Lewis JR. IBM computer usability satisfaction questionnaires: psychometric evaluation and instructions for use. Int J Hum Comput Interface 1995; 7: 57–78
19. Barralon P, Dumont G, Schwarz SKW, Magruder W, Ansermino JM. Comparison between a dorsal and a belt tactile display prototype for decoding physiological events in the operating room. J Clin Monit Comput 2009; 23: 137–47
20. Geldard FA. Some neglected possibilities of communication. Science 1960; 131: 1583–8
21. Barralon P, Dumont G, Schwarz SKW, Ansermino JM. Autonomic nervous system response to vibrating and electrical stimuli on the forearm and wrist. Conf Proc IEEE Eng Med Biol Soc 2008; 931–4
22. Byrne A, Oliver M, Bodger O, Barnett WA, Williams D, Jones H, Murphy A. Novel method of measuring the mental workload of anaesthetists during clinical practice. Br J Anaesth 2010;: 767–71
23. Brown L, Brewster SA, Purchase HC. Multidimensional tactons for non-visual information presentation in mobile devices. Proceedings of the 8th Conference on Human-Computer Interaction with Mobile Devices and Services: MobileHCI 2006. Finland, September 12–15, 2005: 231–8
24. Winberg H, Nilsson K, #x00C5;neman A. Paediatric rapid response systems: a literature review. Acta Anaesthesiol Scand 2008; 52: 890–6
25. 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