Anesthesiologists use data presented on visual displays to monitor patients' physiologic status. Although studies in nonmedical fields have suggested differential effects on performance among display formats, few studies have examined the effect of display format on anesthesiologist monitoring performance.
Methods: A computer-based clinical display simulator was developed to evaluate the efficacy of three currently used display formats (numeric, histogram, or polygon displays) in a partial-task laboratory simulation. The subjects' task consisted solely of detecting any changes in the values of the physiologic variables depicted on a simulated clinical display. Response latency and accuracy were used as measures of performance.
Results: Thirteen anesthesia residents and five nonmedical volunteers, were enrolled as subjects. Use of either the histogram or polygon displays significantly improved response latencies and allowed greater accuracy compared with the numeric display in the anesthesia residents. Neither response latency nor accuracy improved with additional exposure to these displays. In contrast, display format did not significantly affect response latency or accuracy in the nonmedical volunteers.
Conclusions: The results of this study suggest that graphic displays may enhance the detection of acute changes in patient physiologic status during anesthesia administration. This research also demonstrates the importance of assessing performance on clinical devices by studying actual users rather than random subjects. Further research is required to elucidate the display elements and characteristics that best support different aspects of the anesthesiologist's monitoring tasks. (Key words: Computers: simulation. Equipment: monitors. Human performance: human factors. Visual displays: graphic.)
ANESTHESIOLOGISTS gather critical information from devices that monitor patient physiologic status. The values of these measured variables are frequently presented on visual displays. A recent task analysis study demonstrated that anesthesiologists spend at least 20% of their time intraoperatively observing these displays. Although monitoring is a critical component of the anesthesia task, very few studies have examined the effect of the format and manner of clinical data presentation on the ability of anesthesiologists to efficiently and accurately detect changes in the patient's condition.
The ability to correctly detect an unanticipated signal while observing a stream of data (e.g., detection of a clinically relevant change in a clinical variable while monitoring the patient under anesthesia) is influenced by a number of factors including the signal's frequency, salience, and duration as well as the state of the observer (the anesthesiologist)and the other task(s) being performed concurrently. On the basis of laboratory research, it might be predicted that as the frequency of relevant signals increases (i.e., a less stable anesthetized patient), the percentage of signals detected would also increase (increased vigilance). However, the relationship of signal dynamics to anesthesiologist monitoring performance has not been studied.
With the advent of microprocessor-controlled displays and the capability to present data in a wide variety of formats, human factors research has focused on the optimal method for data presentation in complex, dynamic domains. Research in the design of nuclear power plant control rooms and aircraft cockpits suggests that the method in which information is presented can affect the performance of complex monitoring tasks. Traditionally, displays have been divided into two types, numeric and graphic. Studies have demonstrated that the use of certain display formats improves performance of specific types of tasks. For example, numeric displays may be preferable for monitoring tasks that require operators to focus on individual data values. Graphic displays, which provide a pictorial representation of the data, may enhance performance if the monitoring task requires integration of information from many different sources. .
Anesthesiologists must detect acute changes in physiologic variables such as sudden decreases in the patient's blood pressure. This is a focused task that requires the anesthesiologist to know the specific value of the relevant physiologic parameter (blood pressure). However, in many situations, the clinician must have a more global perspective and integrate relevant clinical data to decide whether the state of the anesthetized patient is appropriate. No data are yet available that identify the optimal display elements that enhance the information gathering requirements of the anesthesiologist's job.
A variety of integrated displays have been marketed for use by anesthesiologists, some with claims of decreased user workload or improved clinical efficiency (e.g., only having to look in a single location for multiple sources of information). However, objective published evidence of beneficial effects of integrated graphic displays on clinical performance are thus far absent.
A computer program, the Clinical Display Simulator (CDS), was developed to study how the way clinical information is displayed affects anesthesiologist monitoring performance. The CDS incorporated three design objectives:(1) to present physiologic variables on a display screen in a manner that simulated the physiologic responses of patients under anesthesia;(2) to emulate display formats used by existing anesthesia monitors; and (3) to record accurately and automatically subjects' responses. The primary goal of this initial study was to examine the effect of different display formats on the ability of anesthesia providers to detect acute changes in physiologic variables. Based on the human factors literature, ,* it was hypothesized that the presentation format of clinical data would affect the ability of anesthesia providers to detect changes in the displayed values during a simulated monitoring task. In addition, it was theorized that an increased frequency of abnormal values would result in more rapid detection of subsequent changes in monitored variables. To test these hypotheses, anesthesia residents and control subjects without medical training were studied while using the CDS in an undisturbed laboratory setting.
Materials and Methods
After obtaining approval by the University of California, San Diego Human Subjects' Committee, 13 anesthesia residents at the San Diego Veterans Affairs Medical Center gave written informed consent to be experimental subjects. Four of the subjects were in their first year of clinical anesthesia (CA1 residents with at least three months of anesthesia training), two were in their CA2 year, five were CA3 residents, and two were in their CA4 year. A separate (control) cohort of five nonmedical volunteers also consented to be studied. These latter subjects were all college-educated employees of the Anesthesia Department. All experimental sessions were conducted in a single quiet location between the hours of 10 AM and 7 PM.
During the study period, an Ohmeda Modulus CD anesthesia machine was placed in one of the operating rooms at the San Diego Veterans Affairs Medical Center so that the medical subjects would have the opportunity to use the actual clinical displays that were emulated by the CDS. While all of the subjects used this anesthesia machine during actual cases, no attempt was made to require the use of other than the default (digital) primary display (Figure 1).
The Clinical Display Simulator
The CDS 1.0 is a "C sup ++" computer program, which runs on any Apple Macintosh II series computer (minimum 2 megabytes of random access memory; Apple Computers, Cupertino, CA). The CDS allows the simulation of any clinical scenario. The appearance and content of the simulated displays are controlled by scripts constructed in a tab-delimited text file. These scripts specify the display format, the values of the physiologic variables, and the timing of changes in the values of the variables.
The CDS used a 13" high-resolution RGB monitor (AppleColor, Apple Computers) to present the three display formats to the subjects. Only one display format was presented at a time. The background of each simulated display was dark green and the digital values and graphs were amber.
Physiologic variables were presented in one of three different display formats: the numeric, histogram, and polygon displays (F1-9(A-C)). These displays accurately replicated the primary displays used on the Ohmeda Modulus CD anesthesia machine (Ohmeda, Madison, WI). The physiologic variables displayed included heart rate, arterial blood pressure, tidal volume, oxygen saturation, maximum airway pressure, the expired (end-tidal) partial pressure of carbon dioxide, and the percentage of inspired oxygen. The numeric display presented the values of these physiologic variables plus ten other standard clinical variables in a numeric (digital) format only (F1-9(A)). The histogram display depicted the aforementioned seven physiologic variables in the form of scaled linear tapes where a bobbin indicated the value of each variable on the vertical scale (F1-9(B)). The bobbin moved up and down proportionally on the linear scale as the value of the variable changed. The corresponding digital value for each variable was also displayed directly below each linear scale. Eight other physiologic variables appeared only as digital values on the bottom of the screen (but did not change during the simulation; see Stimulus Dynamics later).
The polygon display represented each variable as one vertex of a single geometric figure (F1-9(C)). The shape of the polygon changed proportionally as the value of each variable changed. Thus, as a variable increased or decreased in value, the corresponding vertex of the polygon either enlarged outward or contracted inward, respectively, and thereby distorted the overall shape of the polygon. A digital value for each variable was displayed adjacent to each vertex. Digital values for 12 other physiologic variables appeared in a row below the polygon (but did not change during the simulation; see later).
Subjects sat in a comfortable chair and observed, without distraction, the values of physiologic variables presented on the computer screen, placed approximately an arm's length away. Subjects indicated when they detected a change in any of the variables by touching directly the appropriate variable via a touch screen (MouseTouch, Richardson, TX) installed on the face of the monitor. After the subject indicated that a variable had changed, a side bar appeared on the display that offered selections:(1) one indicating that the subject thought that the value of the variable had just increased;(2) another indicating a decrease in the variable's value; or (3) a third that was selected if the subject was unsure of the direction of the variable change. During this choice query, the simulation was temporarily halted and then resumed immediately after the subject made one of the side-bar selections.
The CDS recorded the response latency and accuracy of each subject's responses and automatically stored this information. Response latency was defined as the time from the actual occurrence of the variable change until the time the subject first touched the touch screen. Timing was accurate to within 50 ms. If a variable change was not detected within 5 s (a predefined response latency cutoff), then the response was scored as inaccurate and a response latency of 5 s was assigned. The accuracy of each response was determined at two levels; whether the subject chose the correct variable that changed, and whether the subject identified the correct direction of the change. Incorrect responses included both false-negatives (failure to detect a change within 5 s) and false-positives.
Six different clinically realistic scripts were composed to simulate the maintenance of anesthesia during a major surgical procedure. Each scripted simulation was designed to last approximately 6 min. Three of the scripts contained a low frequency of variable changes (stimuli)--only four changes occurred during a 2-min period for each of the three display formats. The three remaining scripts incorporated a high stimulus frequency--9 changes occurred over 2 min per display format. The stimulus interval was randomized between 9 and 26 s for the low frequency scripts and between 6 and 14 s for the high-frequency scripts. These stimulus frequencies were selected based on a review of the literature followed by pilot studies.
The response to each variable change (stimulus) was called a trial. Thus, the low frequency scripts contained four trials on each of three display formats giving a total of 12 trials per experimental session. The high frequency scripts had a total of 27 trials per experimental session--nine trials in each display format (Figure 2).
Ten physiologic variables were chosen as stimuli because they were common to all three display formats and provided clinically important information about the patient's physiologic state under anesthesia. In the graphic displays, only the values of those variables that were actual components of the graph were allowed to change. However, the subjects were not informed as to the subset of variables that would change value. The ten dynamic variables and the respective range of individual variable changes were heart rate (10–43 bpm), oxygen saturation (4–16%), systolic noninvasive blood pressure (12–45 mmHg), diastolic noninvasive blood pressure (10–18 mmHg), systolic arterial blood pressure (11–30 mmHg), diastolic arterial blood pressure (10–28 mmHg), maximum airway pressure (1–15 mmHg), tidal volume (105–350 ml), and the expired (end-tidal) partial pressure of carbon dioxide (4–13 mmHg). Systolic and diastolic blood pressures changed independently and the mean blood pressures changed appropriately to reflect the new value.
Anesthesia residents participated in two separate experimental sessions each lasting approximately 6 min. A crossover design was employed. Subjects were presented with either a high- or low-frequency simulation for the first session and then, during the second session, were presented with one of the opposite frequency. General demographic data and information on recent sleep patterns, caffeine and alcohol intake, and medication use were acquired before each experimental session.
At the first session, the experimenter provided standardized verbal instructions to each subject explaining the nature of the task. The subjects were shown graphic representations of all three display formats and each display was explained in detail. The experimental protocol was described including the underlying rationale (but not the hypothesis under test). Each subject then completed a 2-min scripted training session to become familiar with the CDS, the touch screen, and the experimental task. During the training session, all three display types were presented for 40 s each. There were two randomly occurring variable changes (trials) during each display format presentation. After completion of the training session, questions regarding the CDS, the task, or the display formats were answered. Additional training sessions were offered as needed or desired. The subjects were permitted to begin the actual experiment after successfully detecting at least two thirds of the variable changes during a training session. The first experimental session was undertaken immediately after the subject successfully completed the training. The second session followed 1–10 days later depending on subject availability.
Five of the original subjects were studied on four additional experimental sessions to assess the impact of further exposure to the displays (i.e., training effects). A randomized, blinded, crossover, Latin-Square design was employed. High- and low-frequency scripts were randomized in pairs. If a subject received a high-frequency script at the first session, then a low-frequency script was presented at the second session and vice versa. For the third session, the subject could receive either the high- or low-frequency script; but the script for the fourth session was always the opposite frequency of that used in the third session, and so on.
Control Group (Experiment 3)
As a control, five nonmedically trained subjects with 3–5 years of college education were studied during two experimental sessions. This experiment was performed to analyze whether anesthesiology training and/or prior monitoring experience affected performance on this content specific signal detection task. It was hypothesized that because the simulated displays contained information that was clinically relevant to anesthesiologists, but not to the nonmedical volunteers, there would be differences in performance among the display formats between the two groups of subjects. The initial training and subsequent experiments were conducted in a manner identical to that of experiment 1.
Demographic data and prestudy variables (e.g., sleep, caffeine intake, etc.) were analyzed using one-way analysis of variance (ANOVA) to evaluate differences between experimental groups and sessions. Caffeine intake (in milligrams) was calculated for each subject based on the average caffeine content of various beverages. Linear regression analysis was performed to examine the relationship between response latency and sleep, caffeine intake, or age. In all analyses, a criterion value of P < 0.05 was considered statistically significant. Data are presented as mean plus/minus SEM.
Effect of Stimulus Frequency
The effect of stimulus frequency on average subject response latencies was evaluated. The data from subjects participating in the first and second sessions were analyzed separately using a (2 x 3) mixed ANOVA with one between-subjects variable (stimulus frequency) and one within-subjects variable (display type). .
Because no effect of stimulus frequency on response latency was observed (see Results), the data from the first and second sessions were combined for subsequent analyses creating a new data set, called the first session pair (F2-9). Each subject's session pair contained one high-frequency script (n = 9 trials) and one low-frequency script (n = 4 trials). The session pair data were balanced and randomized with respect to trial and frequency. These data were subjected to a two-way ANOVA using two within-subjects variables (display type and frequency).
Effect of Resident Training Level
The effect of resident training level on response latency was analyzed using a three-way ANOVA with one between-subjects variable (training level) and two within-subjects variables (display type and trial).
Effect of Display Format (Experiment 1)
After frequency and resident training level effects were considered (see Results), the effect of display format on response latency was analyzed using two-way ANOVA with two within-subjects variables (display type and trial number). Significant main effects and interactions were explored using Newman-Keuls a posteriori tests.
Analyses of accuracy were performed on two sets of data. In one data set, a response was considered accurate if the subject identified correctly both the variable that changed and the direction of change. The other data set considered accurate responses as those that identified correctly the changed variable independent of the direction of change. These separate analyses were performed because it was hypothesized that different display types may support better the two different tasks of correctly detecting the variable versus correctly identifying the direction of variable change. Contingency table analyses were performed to examine the effects of display type on response accuracy. The relationship between response latency and accuracy was investigated using linear regression analysis.
Effect of Additional Exposure (Experiment 2)
After excluding the effects of stimulus frequency (see Results), the data from sequential pairs of sessions were combined to create three separate session pairs. Each session pair was balanced and randomized with respect to trial and stimulus frequency. Experimental sessions one and two corresponded to session pair 1, experimental sessions three and four corresponded to session pair 2, and sessions five and six corresponded to session pair 3. The effect of display format across session pairs was then analyzed using a two-way ANOVA with two within-subjects variables in a manner analogous to that described for experiment 1.
Effects in the Control Group (Experiment 3)
The effect of display type on nonmedical volunteers' response latencies was analyzed in a manner identical to that described earlier. In addition, the response latencies of nonmedical volunteers were compared with those of anesthesia residents using a three-way ANOVA with one between-subjects variable (subject group) and two within-subjects variables (display type and trial).
The average age of the 13 subjects participating in Experiment 1 was 31 plus/minus 1 years. On the night before participating in an experimental session, subjects slept an average of 6.8 plus/minus 0.2 h. There were no significant differences in the amount of sleep between sessions. Subjects consumed 65 plus/minus 10 mg of caffeine (approximately 1.5 cups of coffee) during the day of the study before participating in a session. No difference in the caffeine intake between sessions was detected. Neither caffeine consumption, amount of sleep, nor age correlated with response latency or accuracy for any of the display formats. None of the subjects reported consumption of any alcohol-containing beverages in the 12 h preceding any experimental session. Subjects admitted to over-the-counter medication use (e.g., antihistamines, antitussives, etc.) before only 4 of the 46 sessions. There was no effect of prior medication use on response latency.
In all experiments, the majority of incorrect responses were due to a failure to detect a change within the 5-s cutoff latency. In a significant number of incorrect responses, the subject detected that some kind of stimulus had occurred but was unable to correctly identify which variable had changed and/or its direction of change. False-positive responses were extremely rare.
Effect of Stimulus Frequency
Using the data from the thirteen subjects who participated in the first two experimental sessions (the first session pair), no effect of stimulus frequency on response latency was observed (Figure 3). In addition, there was no effect of frequency during the first or the second sessions, individually. Finally, there were no effects of stimulus frequency on response accuracy. Thus, for subsequent analyses, the response latencies from the first and second sessions, consisting of one low- and one high-frequency script, were combined to yield a session pair consisting of 13 trials.
Effect of Resident Training Level
There were no significant differences in either response latency or accuracy between the different years of anesthesia residency training (Table 1). However, when CA1 residents were compared to all of the more experienced (CA2, CA3, and CA4) residents, a significant effect of training level on response latency was observed (F(1,11)= 5.2, P < 0.05). Residents with more than 1 yr of clinical anesthesia experience detected changes in the displayed physiologic variables more rapidly (2.3 plus/minus 0.1 s) than did the CA1 residents (2.7 plus/minus 0.1 s).
Effect of Display Format (Experiment 1)
Using the data from subjects completing one session pair, a significant effect of display type was demonstrated (F(2, 24)= 5.4, P = 0.01). Anesthesia residents detected variable changes significantly more rapidly when using either the histogram or polygon displays compared with the numeric display (Table 2). When accuracy was defined as the correct identification of both changed variable and direction of change, overall response accuracy correlated with response latency (R = 0.61; P < 0.0001). Correct responses occurred more rapidly (1.9 plus/minus 0.0 s) than did incorrect responses (3.6 plus/minus 0.1 s). Subject responses were significantly more accurate with the histogram and polygon displays than with the numeric display (total chi-square = 9.1, P = 0.01).
When correct identification of the changed variable was evaluated independent of the direction of change, accuracy correlated strongly with response latency (R = 0.78). Correct responses occurred more rapidly (1.9 plus/minus 0.0 s) than did incorrect responses (4.6 plus/minus 0.1 s). Again, subjects performed more accurately with the histogram (88%) or the polygon display (83%) than with the numeric display (72%)(total chi-square = 13.6, P = 0.001). In those instances where subjects had correctly detected which variable had changed, they identified correctly the direction of the change in 85% of the trials. There was no difference between displays.
Effect of Additional Exposure (Experiment 2)
The effect of display type on response latency was similar to Experiment 1 when the data from the five subjects who completed three session pairs were analyzed. A significant difference between display types was observed (P < 0.0005;Figure 4). Response latencies for the histogram and polygon displays were 20% faster than for the numeric display. Subjects' performance did not improve with additional session pairs when using any of the three display types.
Analyses of accuracy yielded results similar to those in experiment 1. Subjects' response accuracy did not improve after participating in three session pairs. Noteworthy is that once the correct variable change had been detected, subjects were significantly better at identifying the direction of the change when using the polygon display (91%) compared with the histogram display (75%; P < 0.05).
Nonmedical Control Group (Experiment 3)
The average age of the nonmedical volunteers was 34 plus/minus 6 yr. No significant differences between amount of sleep or caffeine consumption was demonstrated between the first and second sessions. However, in this small group of subjects, age and caffeine consumption did correlate with response latency (R = 0.59, P < 0.001; and R =-0.51, P < 0.005, respectively). As age increased, response latency increased and with increased caffeine consumption, response latency decreased. Multiple regression analysis indicated that the predominant effect on response latency was due to subject age. There were no statistical differences in the age, amount of sleep, or caffeine consumption between the nonmedical volunteers and the anesthesia residents.
There was no effect of display type on response latency or accuracy in the nonmedical volunteer group (T2-9). While there was no difference in overall response latencies between the nonmedical volunteers and the anesthesia residents, there was a significant difference in response accuracy between the two groups (F(1,16)= 9.0, P < 0.01). The nonmedical volunteers responded with appreciably less accuracy (46%) than the residents (69%) across all three display types.
The histogram and polygon displays were more effective than the numeric display at supporting the simulated clinical detection task studied in this experiment. Anesthesia residents responded both faster and with greater accuracy when using the two graphic displays. Subjects did not benefit from additional exposure to any of the display formats suggesting that only modest exposure was adequate training to attain the benefits of the graphic displays. Even though anesthesiologists are more familiar with numeric displays, their performance was still inferior using this type of display compared with the two graphic displays.
In contrast to the anesthesia residents, the nonmedical volunteers performed equally well with all three display formats. However, the responses of the nonmedical volunteers were less accurate than those of the anesthesia residents. A possible reason for the differences between the two groups could be medical training and/or experience. This is supported by the positive effect of anesthesia training within the resident group. It is possible that the information content in the displays was more meaningful to the anesthesia providers than to the nonmedical volunteers. These hypotheses will require explicit testing in future experiments. Nevertheless, the current results appear to suggest that the value of different display formats should be evaluated by the actual users because inaccurate conclusions may be drawn if naive subjects are studied.
The Role of Graphic Displays
The primary purpose of a display is to provide "coherent representation and effective communication" of critical information from the task environment. Graphic displays, also known as object or configurational displays, are intended to improve detection and recognition of visual patterns thereby allowing the observer to determine more rapidly the system's overall state. Object displays, like the polygon display used in this study, appear to be processed "holistically" whereby the perception of the whole takes priority over perception of individual parts. ,** Laboratory data support the assertion that object displays are superior to bar graph (histogram) displays when the task involves the integration of several individual datum values. In contrast, other studies suggest that histogram-type displays may provide for faster responses when a change in a single variable must be detected. The current study contributes to the existing human factors literature on this topic because data are more limited on the relative effectiveness of graphic displays compared with numeric displays on complex signal detection tasks.
The determination of patient status is a time-consuming task in which failure to detect perturbations promptly may lead to serious adverse consequences. Laboratory studies suggest that properly designed object displays may permit more rapid situational assessment and thus enhance performance under time stress, particularly when a unique display configuration has specific diagnostic and/or therapeutic implications. In process (system) control tasks, especially those involving system uncertainty, graphic displays appear to be superior to numeric displays. The relative disadvantage of numeric displays in these situations may be a consequence of slower serial processing of each individual display element. The application of this body of knowledge to the results of the current study suggest the hypothesis that whereas the nonmedical subjects responded to the simulation as a straightforward signal detection task, the experienced anesthesia residents treated the partial-task simulation more like an integrated system control task. Further experiments will be required to validate this hypothesis.
Although this experiment was designed to simulate the monitoring of anesthesia displays, the experimental task represents only one of the many tasks performed concurrently by the anesthesiologist in the operating room. Anesthesiologists are bombarded with many sources of information, demands, and distractions. For example, they must also interact with the surgical staff, keep records, administer intravenous medications, and position the patient. Thus, the anesthesiologist observes clinical displays intermittently and rarely has the luxury to watch them for prolonged periods. In contrast, in this experiment, subjects stared intently at the display screen, free of distractions or other tasks. Hence, the experiment was essentially a classical signal detection task placed in a real-life context using expert (vs. nonexpert) subjects.
The study's experimental design may have led subjects to be more concerned with simply detecting changes in the clinical variables rather than with their information content (the clinical implications of the actual values or how they were changing). However, the different results in the anesthesia residents versus the nonmedical volunteers suggests that some job-specific information was embedded in the simulated displays. Furthermore, an effect of clinical anesthesia experience was demonstrated within the resident group. Although the task used in the experiment did not completely mimic the full responsibilities of the anesthesiologist, it was still a job-relevant task. In fact, signal detection remains a critical component of the anesthesia task; sudden, and sometimes subtle, changes in one or more clinical variables can be early harbingers of acute clinical events.
The frequency and pattern of stimuli (variable changes) chosen for study may have been different than that typically observed in the actual clinical environment. In the partial task simulation, variable changes occurred randomly every 6–26 s (depending on the script). The values of these changes also varied in their potential clinical significance from simply notable to ones that could have required immediate therapy in actual patients. During actual cases, patients' vital signs can similarly change frequently, rapidly and, sometimes, unpredictably. Finally, the subjects only observed each simulated display for approximately 2 min whereas, in the operating room, surgeries may last for many hours. Given that the subjects had no other tasks to perform during the brief study intervals, one might predict that the results obtained would be biased in favor of more rapid and accurate signal detection. Conversely, providers in a real clinical environment might be more highly motivated to detect deviations in patient status.
Although differences in response latency to a change in a clinical variable of 1 s or less may not seem clinically important, these results were obtained in a low workload signal detection task. Yet, even under the relatively low task demands of this study, at best, experienced anesthesia providers detected accurately the variable change and the direction of the change in only 80% of the trials. In a real-life situation, in which workload is much higher and a wide variety of stimuli are vying for the clinician's attention, it is expected that differences in response latency among different display formats could be magnified appreciably. This hypothesis is supported by a study of aviation displays in which during high workload situations, greater differences between display types emerged and by a recent intraoperative task analysis study. .
The results obtained in this study are consistent with data in other, nonmedical fields. For example, in simulations of nuclear power plant control systems, graphic or object displays have been associated with superior performance on integrated monitoring tasks. The ability to obtain objective data on the impact of display format on monitoring performance could provide an impetus to the development and implementation of more effective anesthesia displays. Well-controlled studies employing computer-based clinical simulations may be useful for the objective evaluation of novel displays before their commercial implementation in anesthesia workstations.
Anesthesia providers detected changes in physiologic variables more quickly and accurately using the polygon and histogram displays compared with the numeric display under the simulated experimental conditions of this study. These results are consistent with the hypothesis that graphic displays may better support some complex signal detection tasks. However, additional experiments are required to further elucidate optimal display formats for anesthesiologist performance under actual operating room conditions and the many factors that can potentially influence vigilance and monitoring performance.
The authors thank Jim Callan and Michael Quinn of Pacific Science & Engineering, Inc., for their assistance in the development of the Clinical Display Simulator. They also thank Chuck Boiler, for his statistical insight, and Holly Forcier and Ron Scott, for their technical support.
*Human factors engineering guidelines and preferred practices for the design of medical devices. ANSI/AAMI HE48. Arlington, Association for the Advancement of Medical Instrumentation, 1993, pp 60–62.
**Munson RC, Horst RL: Evidence for global processing of complex visual displays, Proceedings of the Human Factors Society Annual Meeting. Santa Monica, Human Factors and Ergonomics Society, 1986, pp 776–780.