VISUAL CUEING WITH CONTEXT RELEVANT
INFORMATION FOR REDUCING CHANGE
BLINDNESS
Jacqueline M. Tappan, BASc1, Jeremy Daniels, BASc2,
Brad Slavin, BSc2, Joanne Lim, MASc2, Rollin Brant,
PhD3 and J. Mark Ansermino, MBBCh, MSc, FRCPC2
Tappan JM, Daniels J, Slavin B, Lim J, Brant R, Ansermino JM. Visual
cueing with context relevant information for reducing change blindness.
J Clin Monit Comput 2009; 23:223–232
ABSTRACT. Objective. Physiological monitoring is a requisite
for optimal care to ensure that the condition of a patient is
maintained within safe levels. Monitoring can be jeopardized by
the inability of a clinician to recognize important changes in the
visual display of data throughout the duration of the monitoring
task. We hypothesized that the addition of a visual cue imparting
contextual information to a physiological display would improve
the detection ability and response time of a clinician to a change in
a patient variable. Methods. Contextual information based on
trend information was added to a physiological display in the form
of a visual cue. Following IRB approval, the resulting enhanced
display was evaluated by 22 anesthesiologists in a simulated
operating room, through the observation of six simulated
scenarios using a standard anesthesia display and the enhanced
display. Demographic information, response time, accuracy of
detection, and usability data were collected. Results. The
enhanced display reduced the detection time to a change in the
simulated scenarios by 14.4 s (95% CI: -26.4 to -2.38), and
reduced the expected number of missed events per scenario by
0.23 (95% CI: -0.439 to -0.0203), based on the repeated
measures analysis (Poisson model). Conclusions. The data
collected and analyzed in this study supports the addition of a
visual cue to future physiological monitors. The graphic
representation and the context relevant information that it
transmits appears to aid clinicians. While the results indicate that
enhanced visualization of context relevant information can lead
to a significant improvement in event recognition and
identification, further evaluation in clinical settings is required.
KEY WORDS. Visual cues, physiological monitoring, change
blindness, context relevant information, visual display.
INTRODUCTION
Physiological monitoring is a requisite for optimal patient
care, and strives to ensure that the physiological condition
of a patient is maintained within safe levels. Monitoring
can be jeopardized, however, by the inability of a clinician
to recognize important changes in patient data throughout
the duration of the monitoring task [1, 2]. When the
sustained attention of a clinician is diverted from the
monitoring task, change blindness can occur [3]. Tech-
nology designed to assist clinicians in improving and
maintaining their attention level during monitoring tasks
represents a potential mechanism by which patient safety
can be improved [2, 4]. By both sustaining clinician
attention and directing this attention to relevant infor-
mation, small departures from normal physiological states
could be detected at the early stage of evolution, reducing
Presented at the Society for Technology in Anesthesia Annual Meeting,
San Diego, CA, USA; January 16–19, 2008.
From the 1Engineering Systems Division, Massachusetts Institute of
Technology, Cambridge, MA, USA; 2Department of Pediatric
Anesthesia, British Columbia Children’s Hospital, 1L7 – 4480 Oak
St., Vancouver, BC V6H 3V4, Canada; 3Department of Statistics,
The University of British Columbia, Vancouver, BC, Canada.
Received 17 April 2009. Accepted for publication 4 June 2009.
Address correspondence to J. M. Ansermino, Department of
Pediatric Anesthesia, British Columbia Children’s Hospital,
1L7 – 4480 Oak St., Vancouver, BC V6H 3V4, Canada.
E-mail: anserminos@yahoo.ca
Journal of Clinical Monitoring and Computing (2009) 23:223–232
DOI: 10.1007/s10877-009-9186-8
Springer 2009

the time to rescue interventions and the occurrence of
serious adverse events [5, 6].
During a long monitoring period, patient variables
constantly fluctuate. The use of distinctive objects mini-
mizes change blindness by allowing an observer to rec-
ognize these variable fluctuations more easily [3]. The use
of visual cues to sustain attention has previously been
implemented, most aggressively in the air traffic control
environment, with findings suggesting that cueing results
in a high rate of signal detection that can be maintained
throughout a monitoring task [7, 8]. Although these
studies were carried out in fields that are indisputably
different from that of patient monitoring, we felt that such
techniques could be successfully applied to the medical
environment. Specifically, we hypothesized that the
addition of a visual cue imparting contextual information
to a physiological display would improve the detection
ability and response time of a clinician to a change in a
patient variable.
METHODS AND MATERIALS
Following IRB approval1 from the Children’s & Wo-
men’s Hospital Research Review Committee and the
University of British Columbia Clinical Research Ethics
Board, 22 expert subjects, consisting of anesthesiologists
and anesthesia residents (in years 3, 4, or 5), provided
written informed consent and were remunerated $5 for
their participation. Subjects were recruited from British
Columbia Children’s Hospital (BCCH) in Vancouver,
British Columbia, Canada.
Display design
Graphical representations of contextual information were
designed and added to a traditional anesthesia display as
visual cues, providing an indication of the trend of each
variable over time. These graphical cues were governed
by the probability that a trend was increasing or
decreasing. The probability that a trend was changing was
estimated from the relationship between a set of predic-
tions (estimate of present patient state based on case
trends) defined by algorithms described in previous work
and observations (present parameter values) [9, 10]. When
a prediction, based on the information from all previous
observations during a scenario, differs from the present
observations, it suggests that the most current observations
do not follow the previous pattern and a change in trend
has occurred. The accumulation of differences between
predictions and actual observations is used to identify the
probability that the trend or trend segment has changed.
This probabilistic information was included in the
display in addition to the standard numeric display. Four
measured patient variables were developed into visual
cues: heart rate (HR), mean blood pressure (MBP), blood
oxygen saturation (SpO2), and end-tidal carbon dioxide
(ETCO2). These univariate cues were created using the
Macromedia Flash (Adobe Systems Inc., San Jose, CA,
USA) environment. Three graphical prototypes were
designed as possible visual cues, with a participatory design
approach being implemented to assist in the selection of
the graphical prototype most appropriate for the display
[11]. Through this session, a triangle design was chosen
and incorporated into the enhanced monitor design.
The appearance and behavior of the triangle cue is
governed by the probabilistic data. When the probability
of a change in trend is below 0.25, no visual cue is dis-
played. If the probability of a change in trend is above
0.25, a visual cue is displayed, attracting the attention of
the clinician to the changing patient variable. As the
probability value increases past 0.25, the graphical cue
becomes proportionately larger. At all times when a cue is
present, an outline of the maximum size is also displayed
to allow the clinician to contrast the current cue size to
this maximum size (see Figure 1).
Experimental setup
Simulations of both standard [display identical to that
currently used in the operating room (Aestiva/5 anes-
thesia delivery system; GE Healthcare, Madison, WI)] and
enhanced (with visual cues) monitor displays were
developed using the Macromedia Flash environment
(see Figures 2 and 3). In these simulations, the sonification
of heart rate and blood pressure were present, but no
alarms were activated. HR, MBP, ETCO2, SpO2, respi-
ratory rate (RR), anesthetic agent level (AA), and tem-
perature (T) data were displayed to simulate six anesthesia
monitoring scenarios. The enhanced display included four
univariate visual cues alongside the numeric value of the
four patient variables previously described, with the size of
the triangle graphic indicating the probability that a
change in trend for that variable had occurred, and ori-
entation of the triangle graphic indicating direction of this
change in trend. The six scenarios depicted familiar
occurrences during surgery: light anesthesia (Scenario 1),
pulmonary embolism (Scenario 2), anesthetic overdose
(Scenario 3), hypovolemia (Scenario 4), malignant
hyperthermia (Scenario 5), and blood pressure fluctuation
(significant decrease followed by an increase in blood
pressure; minimal changes in other parameters) (Scenario
6). Each scenario was developed from real clinical data1IRB approval on July 6, 2007.
224 Journal of Clinical Monitoring and Computing