Presentation
MDD16 - Exploring Visual, Auditory, and Touchless Haptic Interfaces for Alarm Monitoring in Healthcare Systems
DescriptionBackground and Objectives
Alarms play a vital role in medical device utilization within many healthcare settings, particularly when there is a requirement to observe patients. Studies have explored the importance of the use of alarms in hospitals and have found it to improve clinician awareness of patient’s state, even when monitoring multiple patients (Pruitt et al., 2023; Rayo et al., 2019). These alarms are typically presented as either visual or auditory signals, and occasionally, are combined to improve alert efficiency and redundancy gain (Seagull et al., 2001). However, there has been a frequent issue of alarm fatigue observed in clinicians, leading to vigilance decrement and patient harm (Harris et al., 2011). A possible reason for this could be the overutilization of the current alarm modalities, a situation that could be mitigated by the application of multiple resource theory.
Consequently, there is a need to investigate other alarm modalities that could complement the current alarm modalities and improve information offloading. Touchless haptic alarm modality – a concept that has not been critically explored in hospital environment – is currently being used in diverse domains and has proven to be an effective modality (Colombo et al., 2022; Fink et al., 2023; Stoddard et al., 2022). This study therefore aims to investigate the performance of touchless haptic alarm modality in comparison with visual and auditory alarm modalities during alarm monitoring operation. The objectives of the study include (i) investigating the effect of alarm modality (e.g., auditory, visual, touchless haptic) on monitoring performance, and (ii) investigating the effect of alarm modality (e.g., auditory, visual, touchless haptic) on the user’s cognitive load.
Methods
A between-subjects experimental study was conducted to investigate the differences between visual, auditory, and touchless haptic displays. Participants (N=36) were selected and randomly assigned into either the visual, auditory, or touchless haptic alarm modalities and underwent a screening process that assessed their ability to carry out designated tasks peculiar to their allocated group (e.g., visual, auditory, or haptic). Subsequently, they provided their written information consent and filled out demographic questionnaires. In each modality group, participants were required to sit at a monitoring screen and push a spacebar whenever an alarm stimulus associated with their designated modality was observed. For the visual modality task, the experiment was conducted following the Multiple Vigilance Test (Hirshkowitz et al., 1993). In the auditory group, sound signal of 1000 hertz was played at 60 dB and the tasks to be performed followed the Wilkinson Auditory Vigilance Test (Wickens, 2008). Both visual and auditory stimuli were presented using Microsoft PowerPoint. The haptic stimuli were presented by using an Ultraleap haptic system (www.ultraleap.com/haptics/). The haptic system uses ultrasound waves to create touchless tactile displays.
The study spanned 40 minutes, in which a total of 7 alarms and 233 non-alarms were displayed per time block of 10 minutes, for a total of four experimental blocks. In all modality groups, signals were presented every 2 secs, with non-alarm stimulus and alarm stimulus lasting for 500 msec and 300 msec respectively. Participants’ reaction times, measured by their space bar clicks following an alarm, were recorded using Microsoft Excel VBA. Number of hits (H), false alarms (FA) and misses (M) were evaluated by the researcher. To analyze performance, correct detections and false alarms were recorded for each block. Because the data was counts, two univariate negative binomial mixed-effects models were used. The assumption for a Poisson distribution where the mean and variance were equal was violated for both models; therefore, a negative binomial model was fit. The fixed effects were modality and block, and the random effect was participant ID. Tukey contrasts were performed to analyze pairwise comparisons. For reaction time, a mixed effects linear regression model was fit to all times where the participant reacted to a stimuli (i.e., either a FA or hit). All of the data analysis was performed in R (v4.2.1, R Core Team 2022) using the lme4 package (Douglas et al., 2015), multcomp package (Hothorn, Bretz, & Westfall, 2008), influence.ME package (Nieuwenhuis, te Grotenhuis, & Pelzer, 2012).
Results and Discussion
The study included 36 participants (M = 20.92, SD = 1.32) – 17 of them identified as males, 18 as identified as female, and 1 as non-binary. Participants were equally distributed across the visual, auditory and haptic groups. For the average hits per modality, visual modality group had the highest value (M = 6.70), as compared to the auditory (M = 6.35) and haptic groups (M = 5.63). Average FA was highest in haptic modality group (M = 15.92), as compared to the auditory (M =15.02) and visual groups (M =0.31). Furthermore, average misses were highest in haptic modality group (M = 1.39), as compared to the auditory (M =0.65) and visual groups (M =0.29).
Findings from the negative binomial mixed effect model for the performance metrics indicated that there were no significant differences for block nor alarm modality type for hits. No pairwise comparisons for alarm modality nor blocks were significant for hits. For the FA model, the auditory group had significantly more false alarms than the visual group (p < 0.001). Also, block 1 had a significantly lower number of false alarms compared to block 2, block 3, and block 4 (p<0.001). Tukey contrasts also indicate a significant difference in FA of the haptic and auditory groups where there were significantly less false alarms for the haptic group compared to the auditory group (p < 0.001).
The mixed effects linear regression model of the reaction time suggested a significantly higher reaction time in the auditory group as compared to haptic (p < 0.001) as well as the visual group (p < 0.001). Tukey contrasts did not show a significant difference in false alarms for visual compared to haptic. Blocks 2, 3, and 4 all had significantly longer reaction times than block 1 (p<0.001). For block Tukey contrasts, all pairwise comparisons were significantly different (Table 5). Block 2 had significantly longer reactions times than block 1 (p<0.001), block 3 had significantly longer reaction times than block 2 (p<0.001), and block 4 had significantly longer reaction times than block 3 (p<0.001).
Performance was measured in terms of correct detections and false alarms. Correct detections were when the participant correctly reacted to a critical signal in the vigilance task. False alarms were counted when the participant reacted to a non-critical stimulus. A high number of false alarms or low number of correct detections can be an indicator of low performance. For correct detections, all of the modality groups and blocks performed the same. Therefore, no modality performed better than another. Also, over time and across blocks, performance did not significantly change in terms of correctly identifying signals.
However, in terms of FAs, the auditory and haptic group performed significantly worse. For the auditory and haptic groups, there were a large number of FAs in the first block. Block 1 had significantly more FAs overall in comparison to each individual block. There may have been a learning curve in the first block since there was no practice session for the study.
Cognitive load was measured by reaction times in seconds. Longer reaction times can be an indicator of higher cognitive load (Horsky et al., 2003). The auditory group had significantly longer reaction times than the haptic group and significantly longer reaction times than the visual group. It is clear in the experiment that the visual modality performed the best in terms of performance and reaction times; however, it is interesting that the haptic modality performed better than the auditory modality.
Alarms play a vital role in medical device utilization within many healthcare settings, particularly when there is a requirement to observe patients. Studies have explored the importance of the use of alarms in hospitals and have found it to improve clinician awareness of patient’s state, even when monitoring multiple patients (Pruitt et al., 2023; Rayo et al., 2019). These alarms are typically presented as either visual or auditory signals, and occasionally, are combined to improve alert efficiency and redundancy gain (Seagull et al., 2001). However, there has been a frequent issue of alarm fatigue observed in clinicians, leading to vigilance decrement and patient harm (Harris et al., 2011). A possible reason for this could be the overutilization of the current alarm modalities, a situation that could be mitigated by the application of multiple resource theory.
Consequently, there is a need to investigate other alarm modalities that could complement the current alarm modalities and improve information offloading. Touchless haptic alarm modality – a concept that has not been critically explored in hospital environment – is currently being used in diverse domains and has proven to be an effective modality (Colombo et al., 2022; Fink et al., 2023; Stoddard et al., 2022). This study therefore aims to investigate the performance of touchless haptic alarm modality in comparison with visual and auditory alarm modalities during alarm monitoring operation. The objectives of the study include (i) investigating the effect of alarm modality (e.g., auditory, visual, touchless haptic) on monitoring performance, and (ii) investigating the effect of alarm modality (e.g., auditory, visual, touchless haptic) on the user’s cognitive load.
Methods
A between-subjects experimental study was conducted to investigate the differences between visual, auditory, and touchless haptic displays. Participants (N=36) were selected and randomly assigned into either the visual, auditory, or touchless haptic alarm modalities and underwent a screening process that assessed their ability to carry out designated tasks peculiar to their allocated group (e.g., visual, auditory, or haptic). Subsequently, they provided their written information consent and filled out demographic questionnaires. In each modality group, participants were required to sit at a monitoring screen and push a spacebar whenever an alarm stimulus associated with their designated modality was observed. For the visual modality task, the experiment was conducted following the Multiple Vigilance Test (Hirshkowitz et al., 1993). In the auditory group, sound signal of 1000 hertz was played at 60 dB and the tasks to be performed followed the Wilkinson Auditory Vigilance Test (Wickens, 2008). Both visual and auditory stimuli were presented using Microsoft PowerPoint. The haptic stimuli were presented by using an Ultraleap haptic system (www.ultraleap.com/haptics/). The haptic system uses ultrasound waves to create touchless tactile displays.
The study spanned 40 minutes, in which a total of 7 alarms and 233 non-alarms were displayed per time block of 10 minutes, for a total of four experimental blocks. In all modality groups, signals were presented every 2 secs, with non-alarm stimulus and alarm stimulus lasting for 500 msec and 300 msec respectively. Participants’ reaction times, measured by their space bar clicks following an alarm, were recorded using Microsoft Excel VBA. Number of hits (H), false alarms (FA) and misses (M) were evaluated by the researcher. To analyze performance, correct detections and false alarms were recorded for each block. Because the data was counts, two univariate negative binomial mixed-effects models were used. The assumption for a Poisson distribution where the mean and variance were equal was violated for both models; therefore, a negative binomial model was fit. The fixed effects were modality and block, and the random effect was participant ID. Tukey contrasts were performed to analyze pairwise comparisons. For reaction time, a mixed effects linear regression model was fit to all times where the participant reacted to a stimuli (i.e., either a FA or hit). All of the data analysis was performed in R (v4.2.1, R Core Team 2022) using the lme4 package (Douglas et al., 2015), multcomp package (Hothorn, Bretz, & Westfall, 2008), influence.ME package (Nieuwenhuis, te Grotenhuis, & Pelzer, 2012).
Results and Discussion
The study included 36 participants (M = 20.92, SD = 1.32) – 17 of them identified as males, 18 as identified as female, and 1 as non-binary. Participants were equally distributed across the visual, auditory and haptic groups. For the average hits per modality, visual modality group had the highest value (M = 6.70), as compared to the auditory (M = 6.35) and haptic groups (M = 5.63). Average FA was highest in haptic modality group (M = 15.92), as compared to the auditory (M =15.02) and visual groups (M =0.31). Furthermore, average misses were highest in haptic modality group (M = 1.39), as compared to the auditory (M =0.65) and visual groups (M =0.29).
Findings from the negative binomial mixed effect model for the performance metrics indicated that there were no significant differences for block nor alarm modality type for hits. No pairwise comparisons for alarm modality nor blocks were significant for hits. For the FA model, the auditory group had significantly more false alarms than the visual group (p < 0.001). Also, block 1 had a significantly lower number of false alarms compared to block 2, block 3, and block 4 (p<0.001). Tukey contrasts also indicate a significant difference in FA of the haptic and auditory groups where there were significantly less false alarms for the haptic group compared to the auditory group (p < 0.001).
The mixed effects linear regression model of the reaction time suggested a significantly higher reaction time in the auditory group as compared to haptic (p < 0.001) as well as the visual group (p < 0.001). Tukey contrasts did not show a significant difference in false alarms for visual compared to haptic. Blocks 2, 3, and 4 all had significantly longer reaction times than block 1 (p<0.001). For block Tukey contrasts, all pairwise comparisons were significantly different (Table 5). Block 2 had significantly longer reactions times than block 1 (p<0.001), block 3 had significantly longer reaction times than block 2 (p<0.001), and block 4 had significantly longer reaction times than block 3 (p<0.001).
Performance was measured in terms of correct detections and false alarms. Correct detections were when the participant correctly reacted to a critical signal in the vigilance task. False alarms were counted when the participant reacted to a non-critical stimulus. A high number of false alarms or low number of correct detections can be an indicator of low performance. For correct detections, all of the modality groups and blocks performed the same. Therefore, no modality performed better than another. Also, over time and across blocks, performance did not significantly change in terms of correctly identifying signals.
However, in terms of FAs, the auditory and haptic group performed significantly worse. For the auditory and haptic groups, there were a large number of FAs in the first block. Block 1 had significantly more FAs overall in comparison to each individual block. There may have been a learning curve in the first block since there was no practice session for the study.
Cognitive load was measured by reaction times in seconds. Longer reaction times can be an indicator of higher cognitive load (Horsky et al., 2003). The auditory group had significantly longer reaction times than the haptic group and significantly longer reaction times than the visual group. It is clear in the experiment that the visual modality performed the best in terms of performance and reaction times; however, it is interesting that the haptic modality performed better than the auditory modality.
Event Type
Poster Presentation
TimeTuesday, March 264:45pm - 6:15pm CDT
LocationSalon C
Digital Health
Simulation and Education
Hospital Environments
Medical and Drug Delivery Devices
Patient Safety Research and Initiatives