142 708 THS "T“ LIBRARY 1 £63?) l Michigan State L University J This is to certify that the thesis entitled APPLYING CHANGE DETECTION TO TEST THE NOTICEABILITY OF COMPONANTS OF MEDICAL LABELS presented by CARLY JEAN DEHENAU has been accepted towards fulfillment of the requirements for the Master of degree in Packaging Science ../3 T}. r] LL/L/L, /$2-.,.-~ WK V 3;; Major Professor’s Signature L/ Canal :0 Date MSU is an Affirmative Action/Equal Opportunity Employer PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE JAN 2 O 201 f ’l a—h 5/08 KIProi/AccsrPres/CIRCIDaleDue.indd APPLYING CHANGE DETECTION TO TEST THE NOTICEABILITY OF COMPONANTS OF MEDICAL LABELS By Carly Jean DeHenau A THESIS Submitted to Michigan State University In partial fuifillment of the requirements For the degree of Master of Science Packaging 2010 ABSTRACT APPLYING CHANGE DETECTION TO TEST THE NOTICEABILITY OF COMPONANTS OF MEDICAL LABELS By Carly Jean DeHenau Perception is an active, computationally demanding process that requires cognition as well as perception and attention involves looking at specific features of the environment in a more detailed and focused manner. lnattentional blindness occurs when a stimulus that is not attended is not perceived, even though a person is looking directly at it. Given the documented prevalence of medication error and the criticality of the information contained on the labeling that accompanies medication, research was conducted using change detection to measure the ability of TALL Man lettering to garner attention in labels. Two populations were targeted for this study, those who were employed as healthcare professionals and those who were not. There was a significant interaction between graphic presentation (TALL Man vs traditional) and profession (P=0.0243). Time to change detection was decreased for all professions when the change was presented in a TALL Man presentation as compared to the traditional text. However, the magnitude of this difference and its significance was greatest for nurses (P<0.0001). A main effect of ordered group was evident on time to detect the change (P<0.0001), a significant positional effect of change was detected on time to detect the (P < 0.0001), and there was marginal evidence (P=0.0821) for a difference between word pairs when the dependent variable was time to change detection. ACKNOWLEDGEMENTS I would like to thank Dr Laura Bix for being a wonderful and understanding mentor for this project. I could not have come this far without your guidance and wisdom. Joseph Patton and Linda Williams thank you for recruiting participants, coordinating dates with me to travel to your place of work, and for having a place setup for me to conduct my research. I would also like to thank several people who have helped me pull together, categorize, and manipulate this large amount of data. Nora Bello, for the statistical expertise, you have gone above and beyond anything I ever asked for. Anne Giordano, you helped with all the participants input, a lot of data, and you were an awesome traveling companion. Sunitha Nair, for the help categorizing the survey information, your time was appreciated. Finally, thank you to my committee members, Dr. Diana Twede and Dr. Bill Corser (College of Nursing), for your time and expertise you have given me during my thesis journey. iii TABLE OF CONTENTS LIST OF TABLES ................................................................................... v LIST OF FIGURES ................................................................................ vii CHAPTER 1 - LITERATURE REVIEW ....................................................... 1 Introduction ........................................................................................... 1 Attention .............................................................................................. 7 Theories Explaining Perception of Objects ................................................... 7 Categories of Study Related to Visual Perception and Attention ..................... 11 Look-Alike, Sound-Alike Drug Names and Medication Error .......................... 15 TALL Man lettering as a Solution ............................................................. 18 CHAPTER 2 - METHODS ...................................................................... 23 Statistical Methods ................................................................................ 34 CHAPTER 3 - RESULTS AND DISCUSSION ............................................. 37 Demographic Statistics .......................................................................... 37 Positioning of Word Pairs ....................................................................... 38 Dependent Variable- Time to Detect ......................................................... 39 Binary Response — Successful Detection .................................................. 48 CHAPTER 4 — CONCLUSIONS AND RECOMMENDATIONS ....................... 55 CHAPTER 5 - FUTURE RESEARCH ...................................................... 57 APPENDICES ..................................................................................... 58 APPENDIX A — lRB DOCUMENTS .......................................................... 58 APPENDIX B — SURVEY ....................................................................... 62 APPENDIX C — RANDOMIZATION PROCEDURE AND WORD PAIR COUNTS ............................................................................................ 67 REFERENCES .................................................................................... 71 iv LIST OF TABLES Table 1 - Look alike sound-alike drug name pairs ....................................... 28 Table 2 - Example of a Single Group ........................................................ 36 Table 3 - Number of participants; Male/female, healthcare/non healthcare, nurses ................................................................................................ 36 Table 4 - Number of participants; Age Groups, healthcare/non healthcare ........ 38 Table 5 - Male/Female in each Age Group ................................................ 38 Table 6 - Type 3 Tests of Fixed Effects ..................................................... 40 Table 7 - Covariance Parameter Estimates ................................................ 40 Table 8 - Time in Seconds and 95% Confidence Intervals ............................ 42 Table 9 - Time in Seconds and 95% Confidence Intervals ............................ 44 Table 10 — Time in Seconds and 95% Confidence Intervals .......................... 45 Table 11 - Word pair time to detect, Ranked 1-8 shortest to longest ............... 48 Table 12 - Type 3 Tests of Fixed Effects for the Dependent Variable Successful Detection (yes/no) ................................................................................ 49 Table 13 — Percentages of Iikeliness to Detect ........................................... 54 Table 14 - Circle the names of the drugs that are familiar to you .................... 66 Table 15 - Circle the names of the drugs that you have worked with .............. 66 Table 16 — Position quadrant combinations for each word pair ....................... 67 Table 17 - TALL Man and traditional word pairs by quadrant ......................... 68 Table 18 - TALL Man and traditional word pairs by group ............................. 68 Table 19 - TALL Man by group and quadrant ............................................. 69 Table 20 - Traditional by group and quadrant ............................................. 69 Table 21 - TALL Man and traditional wordpairs by quadrant and position ......... 70 vi LIST OF FIGURES Figure 1 - Cycle of Perception .................................................................. 2 figure 2 - Cross Section of the Eye adapted with permission by de la Fuente 4 Figure 3 - Hermann Grid .......................................................................... 5 Figure 4 - Lobes in the Brain .................................................................... 6 Figure 5 — Timing of Change Detection Software ......................................... 12 Figure 6 - Timing of Change Detection Software ......................................... 25 Figure 7 - Demo Pair; Standard Image and Test Image ................................ 26 Figure 8 - Two graphic Presentations of Word pair - 1A flashes against 1B - 2A flashes against 2B ................................................................................ 29 Figure 9 - standard image of one image pair - Labels were presented in a 4 x 4 configuration ....................................................................................... 31 Figure 10 — Noncritical Image Pair Changes .............................................. 32 Figure 11 - Layout of Experimental Design ................................................ 33 Figure 12 - Quadrants and Positions for Randomization- Quadrant and Position designations added for reader clarification and were not present on test materials ............................................................................................ 35 Figure 13 - Graphic presentation by profession of time to detect a change with standard error bars and significance; Different letters indicate statistical significance at a=0.05; TM = TALL Man, Tra = Traditional, N = Nurse, NHCP = Nonhealthcare Provider, HCP = Healthcare Provider ................................... 42 Figure 14 — Groups of time to detect a change with standard error bars and significance; Different letters indicate statistical significance at a=0.05 ............. 43 Figure 15 — Quadrant by Position Time and 95% Confidence Intervals ............ 46 Figure 16 - Comparison of 16 positions with dependant variable time to detect ................................................................................................ 47 vii Figure 17 — Percentage of Iikeliness to Detect Graphic Presentation...... ...50 Figure 18 - Percentage of Iikeliness to Detect between Professions with standard error bars and significance; Different letters indicate statistical significance at o=0.05 ............................................................................................... 51 Figure 19 — Percentage of Iikeliness to Detect between Groups with standard error bars and significance; Different letters indicate statistical significance at o=0.05 ............................................................................................... 52 Figure 20 - Percentage of Iikeliness to Detect between Quadrant and Position .............................................................................................. 53 viii CHAPTER 1 - LITERATURE REVIEW Introduction Most people believe perception to be a straightforward process, that our perceptual systems are passive agents that automatically transfer the world into an internal representation. Under this view, all that is required for visual perception to occur is for a visually resolvable stimulus to impinge on the eye. In reality, perception is an active, computationally demanding process that requires cognition as well as perception. If you have ever had your mind wander as you. drove a car and missed an important road sign signaling an exit, despite the fact that it was highly visible, then you have experienced the complexities of cognition and perception. Perception is a process that involves many steps which includes: environmental stimulus, attended stimulus, stimulus on the receptors, transduction, processing, perception, recognition, and action. These steps occur in a continuous cycle and do not have a beginning or an end[2] (See Figure 1). The term environmental stimulus includes all things in the environment that we may, or may not, see. Once something becomes the center of our attention it is referred to as an attended stimulus. When there is stimulus on the receptors it creates an image on the back of the eye, called the retina (See Figure 2). At this point, the energy is transformed from one type of energy to another through a process called transduction. The energy is then processed in a way that affects electrical responses in neurons of the brain. Perception is the conscious sensory experience of any stimulus. Recognition is the ability to categorize the object and give it meaning, which influences the viewers’ actions, which includes any type of movement. Knowledge is any information brought to the situation by the viewer. Knowledge Perception Recognition Processing Action Transduction Environmental Stimulus Stimulus on Attended Stimulus the receptors Figure 1 — Cycle of Perception Although perception can be studied with any of the five senses, this discussion is limited to the sense that is most commonly used to gather information from product labeling, vision. To properly understand some of the complexities of the process of perception, an understanding of the basic physiology of the eye is needed. The anatomy of the eye and brain, and how they work together, is an important part of how we perceive stimuli. The retina in the back of the eye has two different kinds of receptors for receiving information. The fovea is in the center of the retina and only has cone receptors. The rest of the retina has both cone receptors and rod receptors, but many more rod receptors than cone receptors. Cone receptors are responsible for collecting specific and detailed data, such as defined shape and color. They are less sensitive to stimuli than rod receptors are. Rod receptors are responsible for general perception of stimuli. While rod receptors are unable to detect color or definition, they are more sensitive to the stimuli in the perception process[2]. Vitreous Humor Refina Aqueous Humor Cornea Fovea ....... Lens :. ,, . Pu . il "i‘ Iris K Optic Nerve Figure 2 - Cross Section of the Eye adapted with permission by de la Fuente [1] After the initial reception by the retina, the signals are transmitted from the cone receptors and rod receptors to the brain. The rod receptors are better able to get information to the brain because of convergence, which is the term used when many neurons send signals to a single neuron. A neuron has to reach a certain threshold to be activated, and once it is activated it can send the information on to the brain. Because cone receptors do not converge, many times, they do not reach the threshold necessary to activate. A phenomenon called inhibition occurs when one neuron decreases the activity in another neuron next to it. The intersections of the white stripes between nine black squares (See Figure 3), called the Hermann grid [3], appear to have gray dots in them, but when you cover up the surrounding area to the intersection the gray dots are no longer there. Inhibition can explain some visual illusions, but there are also some illusions that cannot be explained. This means that vision is more complex than just neurons firing next to each other on the way to the brain. III Figure 3 - Hermann Grid There are different pathways information travels in the brain that respond to what and where (See Figure 4). The “what” pathway is the route that information, in the form of electrical signals, travels so that the stimulus is identified. It is also called the ventral pathway, and leads to the temporal lobe. The “where” pathway is the route information goes to tell where the object is located in space, also called the dorsal pathway, and leads to the parietal lobe. More recently it has been suggested that the “where” pathway should be called the “how” pathway, because it is for taking action. Parietal Lobe Dorsal Pathway where/how Ventral Pathway what Temporal Lobe Figure 4 - Lobes in the Brain These pathways in the brain pass by other areas in the brain that influence perception. The neurons in these other areas are specialized to recognize different features, such as faces, places, and bodies; this is called feature detecting. These specialized neurons can only respond to that stimulus feature. Neurons can be adapted to view a certain type of stimulus if presented to it for a prolonged amount time. Whether these neurons are shaped by evolution or by the experiences of a particular person is yet to be agreed upon. Another point of disagreement is if there are single neurons that identify a certain stimulus, called specificity coding, or if there are groups of neurons that identify a stimulus, called distributed coding. Attention Attention involves looking at specific features of the environment in a more detailed and focused manner. Attention can be divided between multiple stimuli or attention can be selective. During selective attention, focus is directed toward certain objects, while others are ignored. The visual system is designed to operate on selective attention, because we cannot possibly focus on everything at once. Our brains would become too overwhelmed, which is why we have the fovea, which is small, to capture detailed information and eye movement to help facilitate focus on a larger area. When observers are shown a scene for a fraction of a second they perceive the scene as a whole and get the gist of it, but not the specifics. Attention is necessary to get specific and detailed information. lnattentional blindness occurs when a stimulus that is not attended to is not perceived, even though it is in plain sight[2]. Theories Explaining Perception of Objects There are several theories that attempt to explain the complexities of object perception: The Gestalt laws and principals; Perceptual segregation; Feature integration theory; Theory of unconscious inference; Likelihood principle. The Gestalt Laws The Gestalt laws of perceptual organization represent a series of rules that aid researchers in understanding how people organize small parts into wholes. There are six of these rules. 1. The law of good figure- “every stimulus pattern is seen in such a way that the resulting structure is as Simple as possible.” The law of similarity -“similar things seem to be grouped together." The law of good continuation -“points that , when connected, result in straight or smoothly curving lines are seen as belonging together and the lines tend to be seen in such a way as to follow the smoothest path.” The law of proximity or neamess - “things that are near each other appear to be grouped together.” The law of common fate - “things that are moving in the same direction appear to be grouped together.” The law of familiarity - “things are more likely to form groups if the groups appear familiar or meaningful.”[2] The Gestalt laws were written in the early 19003, so more recently, psychologists have developed additional principals. The principal of common region is “elements that are within the same region of space appear to be grouped together.” The principal of uniform connectedness suggests that “connected regions of visual properties such as lightness, color, texture, or motion are perceived as a single unit.” The principal of synchrony is ”visual events that occur at the same time will be perceived as belonging together.” Although these laws and principals are generally accepted, they do not apply to every case and are considered more rules of thumb.[2] Feature Integration Theory “Binding is the process by which features such as color, form, motion, and location are combined to create our perception of a coherent object.”[2] Attention plays a critical role in binding, according to feature integration theory, because before attention is focused on the object color, motion, form, and location are separate. Focusing attention on the object binds the features into one coherent object. Perceptual Segregation Perceptual segregation is when an object or group of objects is separated from the rest of the scene. Figure—ground segregation is how humans perceive many scenes. The figure is the object that stands out from the background, and the contour separating the figure from the ground appears to belong to the figure. The ground is the rest of the scene and is seen as uniform material. Also, humans respond more to verticals and horizontals called the “obliqUe effect”[2]. Theory of Unconscious Inference and Likelihood Principl_e There are many ideas about how intelligence affects our perceptions. The theory of unconscious inference says that some perceptions are made by assumptions we unconsciously make about an environment. The likelihood principle states that we perceive the most likely object from the image that we receive. There are two different types of perceptual processes: bottom-up processing and top-down processing. Bottom-up processing is solely based on incoming information. Top-down processing is based on knowledge that the observer already has and the incoming information. Many perceptions are influenced by the knowledge that a person has, sometimes even without them knowing it. Perceptual recognition has been described as an association between an incoming stimulus event (e.g., reading a medication order, label, or package) and a recognized "template" that is stored in long-term memory (e.g., recalling previous similar drug orders, labels, or packages).[4] 10 Categories of Study Related to Visual Perception and Attention Perception can be studied at different levels of analysis: psychophysical and physiological. The psychophysical level of analysis involves presentation of a stimuli followed by a request for a response from the observer, and is the relationship between stimulus and perception. The physiological level of analysis involves presenting a stimuli and measuring the neurological response in the brain, and is the relationship between stimulus and physiology. Some studies use both psychophysical and physiological levels of analysis to develop a more complete understanding of perception and response to a given stimuli. Methodological approaches are considered when researchers want to characterize the psychophysical level of interaction between subject and stimulus include recognition and visual search. Recognition is a simple method in which an observer is asked what a stimulus is and the observer responds; the dependent variable in these types of studies are the number of correct responses. Visual search methodologies involve requesting the observer find a certain stimulus in as little time as possible; the dependent variable in these types of studies is primarily the time to find the stimulus of interest. To measure the ability of textual formatting to garner attention in labels, we employed a psychophysical technique that is commonly used in the field of visual psychology. This technique is referred to as change detection, or a flicker 11 task. “Change detection is presenting one picture, then a blank field, followed by the same picture but with an item missing, followed by a blank field and so on.”[2, 5] (See Figure 5) Iterative Loop of Standard and Test Images (Test image with a Single Change) Standard Image Grey blank- Test Image Appears for 240 milliseconds Appears for Appears for 240 milliseconds 80 milliseconds Figure 5 - Timing of Change Detection Software During the change detection task, participants are instructed to find the change in the scene as quickly as possible. As such, the dependent variable is the time to detect a change. Change blindness is the difficulty of detecting changes in a scene [5-6]. People think that they do not experience change blindness, because they think abrupt changes rarely occur in real life, this is called change blindness blindness [2]. [Many experiments show that the disruption of attention is the reason we miss the changes.[5, 7] 12 Studies in visual cognition suggest that observers “never form a complete, detailed representation of their surroundings,” and that in the absence of localized motion signals, attention is guided based on interest. As such, “perception is mediated through a narrow attentional bottleneck with attention attracted to various parts of a scene based on high-level interest.” [8] Given this reliance on attention, the time required to detect a change (when using change detection) is a good indication of the time when attention first selected that object. Faster detection of the change implies early attention to the changed property, which can also provide insights about the viewers’ “attentional scan-paths” [8-9]. As such, change detection has been used to measure the locus of attention (a combination of perception and cognition) much as eye tracking measures the fixation of the eye (i.e. perception) [8-9]. Research indicates that change blindness, or the inability to detect a change, occurs for a variety of reasons. This includes when two disparate scenes are separated by a blank screen (the flicker change detection technique) [8, 10-13], eye blinks [14], saccades [15-17], camera cuts [18] or occluding events[19]. It can also be produced when multiple "mud-splashes" hit the screen simultaneous with the change [20] , or when changes occur extremely slowly [21]. All of these techniques have one thing in common; the change fails to produce a sudden and unique visual transient, which would otherwise capture attention to the location of the change. 13 In other words, in our field of vision, overall scenes appear to be in focus, but change blindness experiments show othenivise. If we are not cued to look at a specific area we miss changes, and that shows that only a small part of our environment is actually encoded in detail.[2, 6] A recent study was conducted using change detection on food labels. Researchers tested varied information on food labels including: fat content, best before date, recycling information, and organic information. The research team conducted an additional study to test two graphic presentations: textual and graphic symbol. The study showed the fastest detection of the change was for the organic information and the slowest was for the fat content. These detection times were contradictory to a self report survey, wherethe participants were asked which information on the label was the most important to them. An additional study indicated that the information on the label determined the time to detect rather than the graphic presentation of either textual or graphic symbol. The study concluded that the results are positive and suggest that incorporating change detection as a measure of attention blindness in future studies, in addition to seif-report instruments, might be a fruitful approach[22]. Given the documented prevalence of medication error and noncompliance, and the criticality of the information contained on the labeling that accompanies medication, we chose to conduct research using change detection to the labeling of look-alike, sound-alike drugs. 14 Look-Alike, Sound-Alike Drug Names and Medication Error Many drug names can look or sound like other drug names, which leads to confusion and potentially harmful medication errors.[23-26] Among other types of adverse events, the subject of medication error has received significant national attention since publication of two reports from the Institutes of Medicine (IOM), which focused on medication errors in the US. [27-28] It’s estimated that one in four medication errors involve products that sound or look alike.[23-25] ISMP [Institute for Safe Medication Practices] receives 1200-1500 reports each year involving serious complications resulting from the use of drugs. Approximately 25% of these (300-400) are related to drug name confusion and another 25% related to labeling and packaging issues.[29] Reports submitted to the United States Pharrnacopeia Medication Errors Reporting (MER) and MEDMARX programs underscore how similarity in a products drug name, label, or packaging can lead to errors. From January 2000 to March 2004, there were 31,932 reports submitted to MEDMARX that listed one or more causes of error were related to look-alike or sound-alike drug names, packaging, and/or labeling and approximately 2.6% of these reports were categorized as harmful to the patient[4]. As a result of statistics such as these, the FDA is turning more attention to reducing errors from lookalike and sound-alike drug names. Some action items 15 were developed as a result of the Institute of Medicine report. But the problem exists, and it will probably get worse as we keep adding more drug products into the crowded market of drug names[25]. All drug names have a chemical name that chemists use, but these names have limited usability for practicing healthcare professionals. Many drugs have code designations and/or trivial names that are used during research and development, which may cause confusion once the drug has established generic and brand names. Each drug is assigned a generic (non-proprietary) name, which is essential for communication about that product regardless of other names that are used for the drug. Phannacopeias have also proposed pharmacy equivalent names for complicated products, such as combination products. These names, while not official, are sometimes useful to help simplify drug names and thereby reduce confusion for pharmacists. Finally, brand names, also called proprietary or trade names, are trademarked names developed when the drug is marketed [23, 29-30]. The similarity of all these names between all drugs increases the risk of drug name confusion errors.[24] Sound alike and look alike drug names can lead healthcare professionals to unintended interchanges of drugs that can result in patient injury or death. The existing medication system is flawed because its safety depends on human perfection. Simplicity, standardization, differentiation, lack of duplication, and unambiguous communication are human factors 16 concepts that are relevant to the process, but these principals have often been ignored in drug naming and labeling of packaging[29]. A significant contributor to the problem is that many drugs are from similar families and are used in similar manners. Consider, for example, ephedrine and epinephrine. Not only do these drug names look similar, but their use as vasoconstrictors makes storage near each other likely. Both products also may be packaged alike in 1 mL ampoules or vials. [31] The inherent problems with this have been realized in the form of serious adverse events. One such instance was illustrated in a labor and delivery unit, where, “a healthy young woman became hypotensive after epidural anesthesia was administered. A nurse immediately called the obstetrics resident to inform him of the patient's condition. The resident, known to be "difficult" at times, became angry and snapped at the nurse as he ordered ephedrine 10 mg to be given slow IV push. When processing the order, the nurse, who was anxious because of the physician's behavior, made a mental slip and thought of "epinephrine." With only a few ampuls [ampoules] of epinephrine 1 mg on the unit, she decided to borrow more from the nursery. She found a 30 mL vial of epinephrine 131,000 (1 mg/mL), per withdrew 10 mL, and returned to administer that amount to the patient. Almost immediately, the patient developed tachycardia, severe hypertension, and pulmonary edema. Fortunately, an anesthesia staff member was present and recognized the problem immediately. The patient was treated successfully and the baby was delivered safely.” [31] “An eerily similar scenario played out recently at a different hospital where yet another patient was hypotensive from epidural 17 anesthesia. A nurse called pharmacy to report that her automated dispensing cabinet didn't have enough epinephrine to administer a 5 mg IV dose. A pharmacist immediately reviewed a copy of the order in which the physician had clearly prescribed ephedrine 5 mg IV. The reporter noted that, had there been enough epinephrine in unit stock, a 5 mg dose might have been given. We've [Institute of Safe Medication Practices] also received reports where diluted ephedrine was administered in error instead of epinephrine. In one case, a patient received an irrigation solution during an orthopedic procedure where ephedrine, not epinephrine, was added to a 3 liter container. In yet another hospital, ephedrine was used to compound an epinephrine infusion.”[31] Another such example concerns chlorpropamide and chlorpromazine. “One patient died and another developed serious symptoms after receiving 750 mg of chlorpropamide (Diabinese), instead of 75 mg of chlorpromazine (T horazine).”[25] TALL Man lettering as a Solution In 2001, the F DA/Office of Generic Drugs began a "Name Differentiation Project". This program was designed to promote the use of TALL Man lettering, the use of small and large letters, in drug names to visually differentiate between drugs with very similar names[32]. The use of TALL Man lettering is gaining wide acceptance[33—34]. It has been recommended that TALL Man lettering be used to differentiate a standard set of look-alike drug name pairs on pharmacy-generated labels, drug selection screens, shelf labels, medication administration records, and order sets. [31-36] 18 The Joint Commission, National Association of Boards of Pharmacy, and other safety-conscious organizations have promoted the use of TALL Man lettering as a means of reducing confusion between similar drug names[34]. Research regarding the effectiveness of TALL Man lettering formats in preventing medication error is limited in nature, but what does exist tends to suggest that it is a positive step for healthcare. Studies are varied in approach, using methods that range from surveys to eye tracking, but fairly consistently agree that TALL Man has the potential to positively impact the problem. Studies of the Efficacy of TALL Man Lettering The Institute of Safe Medication Practices conducted a survey on their website to gather information about the use of TALL Man lettering. Out of 451 survey responses, the vast majority of respondents (87%) agreed that the use of TALL Man lettering helps reduce drug selection errors, and two-thirds (64%) reported that TALL Man lettering has actually prevented them from dispensing or administering the wrong medication. Half to three-quarters of respondents who have used TALL Man letters felt that this strategy was effective in reducing the risk of errors, depending on where it was used. Between a quarter and a third of respondents were undecided about the effectiveness of TALL Man lettering, but very few reported that TALL Man lettering was an ineffective strategy in reducing the risk of errors[34]. 19 The Standards Development subcommittee of the Safe Medication Use (SMU) Expert Committee of the United States Pharrnacopeia also collected data on the effectiveness of TALL Man lettering. Responses totaled 900 from the state association constituents. Of these, 840 (93%) were actively involved in patient medication activities. 777 (92%) of these 840 were aware of TALL Man lettering, and 547 (70%) of the 777 say TALL Man Lettering is being used in their organizations[37]. Psychophysical studies have results congruent to the survey work presented above, finding that highlighting sections of words using TALL Man lettering can make similar drug names easier to distinguish, resulting in fewer errors among products with look-alike names[38-39]. One such study employed eyetracking on a computer screen using an array of products as stimulus material. In the eyetracking experiment, one drug from each pair of lookalike sound alike drugs was presented as a target to search for (e.g., chlorpromazine), and the other drug was present in the array as a distractor (e.g., chlorpropamide). It is critical to note that the target was never present in the array. The array consisted of the distractor plus 19 other drug products. It was found that participants were less likely to incorrectly indicate that a target drug was present in an array when the name contained TALL Man lettering. The eye movement data directly 20 corresponded with the error data ,that is, when participants made more errors, they also made more fixations and spent longer fixating on the distractor drug name in the array[39]. In a different experiment, which consisted of two phases, participants were given a “same/different” judgment task in which they were presented with a pair of lookalike sound alike drug names on a computer screen and had to indicate, as quickly and as accurately as possible, whether the two names were the same drug name presented twice or if they were two different drug names[38]. Unlike the eye tracking study and the surveys previously discussed, results of the experiment indicate that name pairs containing TALL Man letters are not easier to distinguish from each other than other name pairs that are presented in lowercase. In a second phase of the same experiment, participants, who were different from those tested in the first phase, performed the same task except for they knew that TALL Man lettering was used in an attempt to make similar names less confusable with each other[38]. Results suggest that if participants are aware of the purpose of TALL Man lettering, they can more easily tell the difference between drug name pairs with TALL Man lettering than between lowercase drug name pairs. Taking results from the first experiment into account, this would suggest that TALL Man lettering can be effective if participants are aware of its purpose[38]. 21 The “same/different” study also tested bolding, coloring, underlining, enlarging and italicizing the different portions of the drug names. Printing sections of the names in color or any other form of differentiation did not aid recognition, and there was no significant benefit of a TALL Man lettering and color combination over using TALL Man lettering alone[38]. 22 CHAPTER 2 - METHODS The objective of this study is to add to the body of work that qualifies TALL Man lettering as method that should be used to differentiate lookalike sound alike drum names, and to evaluate the change detection “flicker task” method so it can possibly be used in future labeling studies. Methods employed in the recruitment, testing and subsequent data storage and handling were approved under IRB 09-623. To participate, subjects had to be at least 18 years of age or older and were screened for a history of seizure. Additionally, they could not be legally blind. Two populations were targeted for this study: those who were (or had been) employed as healthcare professionals and those who were not. The healthcare population was recruited from three locations: (1) in Westlake, Ohio, an area in close proximity to the Cleveland Clinic, (2) at the National Center for Patient Safety in Ann Arbor, MI, and (3) through the College of Nursing at Michigan State University (East Lansing, MI). The nonhealthcare population represented a convenience sample consisting primarily of students and University employees. All participants were recruited with email fliers and by word of mouth (see Appendix A for flier). 23 Researchers provided participants with a verbal explanation of all test procedures. All participants were provided with a consent form (approved under IRB 09-623 - see Appendix A) which they were asked to review and sign. This consent form mentions that participants may withdraw from the study at any time without penalty. Details regarding contact information for both experimental questions and treatment issues are provided at the end of the consent form. Each participant was then assigned a participant number. The participants were asked if they require corrective eye wear for reading. If they indicated this to be the case, they were asked to wear it. An Adobe flash program created by the research team was used for all testing. During a trial, a “standard image” continuously. alternates with a “test image,” with a brief, gray screen separating the presentation of said images. Each of the two alternating images are displayed for 240 milliseconds, and the gray blank between the images is displayed for 80 milliseconds (see figure 6); these timings are based on the work of Rensink et al[7]. 24 Iterative Loop of Standard and Test Images (Test Image with a bSingle Change) Standard Image Greyb Ian Test Image 16 packages Appears for Appears for 240 milliseconds Appears for 240 milliseconds 80 milliseconds Figure 6 - Timing of Change Detection Software Prior to collection of recorded data, each participant was tested using three pairs of images that allowed them to become familiar with the technique and become comfortable with how the study worked (See figure 7). Participants were seated comfortably in front of a laptop computer, a Sony Vaio laptop with an Intel Centrino processor 1.86GHz, 1 GB of RAM, and a 15” Widescreen with X-Brite technology. This was followed by the instruction, "This is a demonstration of the change detection software that will be used throughout testing. You will see two images flash over one another with a single change between the two. We are trying to see how long it takes people to notice the change in these images. Please hit the space bar on the computer as soon as you notice the changefl 25 vincristine f9 vincristine 113 Injection Injection 1mg/1mL 5mg/1mL Figure 7 - Demo Pair; Standard Image and Test Image After a participant indicated that they had found the change, by pressing the space bar, they were asked to identify the change by clicking on the area of the image that was changing using the mouse cursor. If they identified the change correctly, the dependent variable, time to detect, was recorded by the software and they continued testing by viewing a new pair of images. If they failed to click on the area of the image that was changing, the software returned to that pair again until they indicated the change correctly. In the event this happened, the dependent variable, time to detect, was the sum of all attempts collected for a given pair. If the participant took more than 2 minutes to identify the change, in either a single attempt or multiple attempts, the researcher ended the attempt, indicated the change, recorded two dependent variables (1)time as 120 seconds; (2) the binary variable, fail to detect, and the participant moved on to the next image pair. 26 The stimulus materials (2 images in a pair and 16 labels per image) were created using Adobe Illustrator CS4; all images were 800 pixels wide by 600 pixels tall. Each individual label within each image measured 175 pixels by 108 pixels spaced evenly. Image pairs were presented as a four x four label configuration (see Figure 9 for the standard image of one test pair). Within each label the brand name, route of administration, and dosage were created with Arial font in type size 12, and the RX statement was also Arial font in type size 9 (See Figure 10). Images were converted to a jpg file format so that they would be recognized by the change detection software. Based on information obtained from the website of the Institute for Safe Medication Practices (ISMP) eight pairs of look-alike, sound-alike names were selected; in other words a total of 16 drug names were tested. Additionally all of the eight word pairs selected have been approved for use by the FDA in TALL Man formats (see Table 1). Drug names were chosen so that the dissimilar portion of the names appeared in varied locations of the word (beginning, middle, and end). Each word pair was presented in two levels of graphic presentation (TALL Man and traditional-See Figure 8 for the two graphic presentation levels of the same word pair), for a total of 16 pairs (8 word pairs by 2 levels of graphic presentation). 27 8 Word Pairs by Graphic Presentation TALL Man Orientation acetoHEXAMIDE acetaZOLAMlDE chlorproMAZlNE chlorproPAMlDE clomiPHENE clomiPRAMlNE DAUNOrubicin DOXOrubicin dimenhyDRlNATE diphenhydrAMlNE medroxyPROGESTERone methleESTOSTERone predniSONE prednisoLONE vinBLAStine vinCRlStine Traditional Orientation acetohexamide acetazolamide chlorpromazine chlorpropamide clomiphene clomipramine daunorubicin doxorubicin dimenhydrinate diphenhydramine medroxyprogesterone methyltestosterone prednisone ’ prednisolone vinblastine vincristine Table 1 - Look alike sound-alike drug name pairs 28 acetazolamide 1 D acetohexamide 1 B Injection Injection 250 mg 250 mg acetoHEXAMlDE f9 acetaZOLAMlDE *9 Injection Injection 500 mg 500 mg Figure 8 - Two graphic Presentations of Word pair - 1A flashes against 18 - 2A flashes against 23 In other words, a single image contained 16 labels that flashed against another image with 16 labels (See Figure 9). A single change was present as the image pair “flickered”. Changes were classified as “critical” if they involved the drug names in either of the graphical formats (TALL Man or traditional); for example, chlorpropamide and chlorpromazine (flashing one against the other in the lower case graphic presentation). Another critical change occurred when the 29 lookalike sound alike names were presented in the TALL Man presentation (chlorproPAMlDE would flash against chlorproMAZlNE). The dependent test variables were was the time it took participants to notice the change and the binary variable, detected, yes/no. 3O 25 xx mE mm .90 0* mz.$. - l__.-__J Acuity is IS —} -—> Collected T959" Group 1 ———-———— -— Group 2 l . . _ Survey 4 critical pairs 4 non cntIcal paIrs ‘ regarding ‘ 4 non critical pairs L———J medication L———~l usage Group 4 ——————_ Group 3 4 critical pairs Order of test pairs and dummy pairs within set are randomized, as is position of change Subject views a total of 32 changes; 16 are of interest to the research team Figure 11 - Layout of Experimental Design At the first break, the participants demographic information was collected. This included: their gender, age, profession, education level, requirements for eyewear and self declared ethnicity. After the demographic information was collected, the participant was asked to view 8 more image pairs using the change detection tool. As another break, visual acuity was tested using a Dow Corning Ophthalmics Near Point Visual Acuity Card. Participants were instructed to hold 33 the test card approximately 16" from their eyes and read the lowest line on the card that they could aloud. In the event that the participant could not correctly report all letters on a given line, they were asked to read the line directly above it. This continued until they did not miss any letters. Their visual acuity was recorded (20/20, 20/30, etc.) accordingly. At this time, a refreshment break was offered. Researchers provided juice boxes and colas as well as a small snack. Following the test of visual acuity and a snack, 8 more image pairs were tested. Following the test of the third group of image pairs, a survey was conducted asking the participant some questions involving their behavior using medications (See Appendix B). Following the break, the final group of 8 image pairs was tested. Statistical Methods This experiment is a split plot design. The first split is health care professionals versus nonhealth care professionals. The second split is 8 word pairs* 2 graphic presentations (T ALL Man versus traditional) for a total of 16 critical pairs. It is a complete block design; to control for any effect of positioning extensive randomization was conducted. The labels in the images are split into quadrants (1-4); top left comer, top right comer, bottom left corner, and bottom right corner (See Figure 12). Then the quadrants are split into positions (1-4). The first level of randomization occurs in the positions within a given quadrant. 34 The second level of randomization was the quadrants of each image pair within each participant. The third randomization was the order of the all critical and noncritical image pairs for each participant (See Appendix C for more information). Each of the four groups is balanced to contain 2 TALL Man graphic presentations, 2 traditional graphic presentations, and 4 noncritical image pairs (See Table 2). Because previous work regarding the use of change detection and labels indicated both a run effect and a positional effect, the balanced randomization was necessary. [40] Quadrant 1 Quadrant 2 muons maproPAM DE W Om sol o... I ‘9 raccoon 1. 5 ML 1m mg 5 ml- ' osmon 1 . R. ow Q. 0* R- 0'0. drawn-um ommyommra prom: om 1° 0.. 1» 0.....m +- 50 mg 50 mg 5 mglml. W 91 M RI 0'!" mm My" ESTOSTERona “onus“ chlorpromazine Inpcbon f. ""l°'13‘l"3"l {0 Injection Oral # ImgllmL 150ml 10rng/10rnl. MW RI Cm J 3- 09, RI 0* mam MW terms WM MMWINE . W [mm m .1. 0" 4. Oral 1 25° "‘9 150 ml. 50 mo 25 "‘9 17' . _ _ ‘2. Or» Quadrant 3 Quadrant 4 Figure 12 - Quadrants and Positions for Randomization- Quadrant and Position designations added for reader clarification and were not present on test materials 35 Example of a Single Group Type of Change predniSONE — prednisoLONE Critical- TALL Man daunorubicin - doxorubicin Critical- Traditional Graphic Noncritical Graphic Noncritical dimenhyDRlNATE — diphenhydrAMlNE Critical- TALL Man prednisone — prednisolone Critical- Traditional Root of Administration Noncritical RX Statement Noncritical Table 2 - Example of a Single Group After the healthcare professionals completed the change detection testing, they were provided questions regarding their experiences (or lack thereof) with medication error, TALL Man lettering and asked to recommend solutions. 36 CHAPTER 3 - RESULTS AND DISCUSSION Demographic Statistics A total of 80 participants were tested in this research study; of these, 40 were, or had been, employed as healthcare providers. Within the healthcare providers, there were 16 who reported their profession as nursing and 24 listed other types of healthcare professions. Table 3 presents frequency counts by gender and professional status for the test population. Twenty seven of the participants were male and 53 were female. Of the males participants, 4 were healthcare providers and 23 were non healthcare providers. Within the male healthcare providers there were 2 nurses and 2 other types of healthcare professions. Of the female participants there were 35 healthcare providers and 18 non healthcare providers. Within the female healthcare providers there were 14 nurses and 21 other types of healthcare professions. Non Healthcare Healthcare (# of Totals healthcare providers that were nurses) Male 23 4 (2) 27 Female 17 36 (14) 53 Total 40 40 (16) 80 Table 3 - Number of participants; Malelfemale, healthcarelnon healthcare, nurses 37 Table 4 presents a breakdown of the participants by age and professional status, and Table 5 provides age frequency data by gender. Non Healthcare Healthcare (# of Totals healthcare providers who were nurses) 1 8-24 22 4 (0) 26 25-34 10 10(4) 20 35-49 2 1 1 (4) 13 50+ 6 1 5 (8) 21 Total 40 40 (16) 80 Table 4 - Number of participants; Age Groups, healthcarelnon healthcare 1 8-24 25-34 35-49 50+ Total Male 1 5 7 3 2 27 Female 1 1 1 3 10 19 53 Total 26 20 1 3 21 80 Table 5 - Male/Female in each Age Group Of the 80 participants, 64 people indicated their first language to be English. Sixty-six participants self-declared their racial status as white, eleven as Asian, two as black, and one as Hispanic. Positioning of Word Pairs As referenced in the Methods section, each participant viewed 32 pairs consisting of both critical and non-critical images, and position and order of 38 presentation were randomized. Readers interested in the frequency of quadrant (14), graphic presentation (I' ALL Man and traditional), group, and position within quadrant should refer to Appendix C. Dependent Variable- Time to Detect The dependent variable, “time to detect,” recorded in seconds, was modeled using a general linear mixed model fitted with the MIXED procedure of SAS software (Version 9.2 SAS, Inc., Cary, NC). The model included the fixed effects: profession (students, nurses, and “other healthcare provider”), graphic presentation (TALL Man and traditional), their two-way interaction, as well as the ordered group of presentation to the participant. Also included in the model were fixed positional effects, provided as quadrant by position combinations (see Appendix C for a complete description of the randomization methodology), and the interactions between treatment and quadrant and treatment and group. Random effects used in the model included: participant nested within profession, participant by treatment (to account for technical replications), participant by ordered group and participant by position by quadrant. Run order within group, gender, ethnicity and language were evaluated as fixed effects but were excluded from the final model based on P-value > 0.20. Age was excluded from the final model due to overparameterizationlmulticollinearity with the fixed effect factor profession. In order to appropriately meet model assumptions, the dependent variable (time) was log transformed. 39 Table 6 indicates the results of the statistical analysis for the fixed effects and Table 7 indicates the results of the statistical analysis for the covariance parameter estimates. Effects that are bolded represent those that indicated statistical significance at a=0.05. Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatment 1 923 40.40 <0.0001 Profession 2 76.9 3.00 0.0557 Treatment x Profession 2 905 3.73 0.0243 Group 3 239 8.24 <0.0001 Position x Quadrant 12 735 5.86 <0.0001 Tallman x Quadrant 3 1084 1.98 0.1147 Tallman x Group 3 933 1.65 0.1759 Table 6 - Type 3 Tests of Fixed Effects Covariance Parameter Estimates Cov Pan'n Estimate Standard Error 2 Value Pr > Z Subject x Profession 0.04868 0.1726 2.82 0.0024 Subject x Treatment x Profession 0 - - - Subject x Group 0.1334 0.01956 0.68 0.2476 Subject x Position x Quadrant 0.09018 0.03215 2.80 0.0025 Word Pair 0.01710 0.1229 1.39 0.0821 Residual 0.6303 0.0477 15.46 <0.0001 Table 7 - Covariance Parameter Estimates 40 LiLL Man Vs Traditional by Professional Status As indicated in the Tables (6 and 7), several interaction terms showed evidence of significance. Because the effect of graphic presentation (TALL Man vs traditional) is of primary interest to this study, understanding the significant interaction between treatment and profession is important (P=0.0243). Figure 13 and Table 8 provide further insights into this interaction. Time to change detection was decreased for all professions when the change was presented in a TALL Man presentation as compared to the traditional text. However, the magnitude of this difference and its significance was greatest for nurses (P<0.0001) compared to the students or other healthcare professionals. 41 7O Seconds TM+N TM+ TM+ Tra+ Tra+ Tra+N NHCP HCP NHCP HCP Figure 13 — Graphic presentation by profession of time to detect a change with standard error bars and significance; Different letters indicate statistical significance at a=0.05; TM = TALL Man, Tra = Traditional, N = Nurse, NHCP = Nonhealthcare Provider, HCP = Healthcare Provider Time in Seconds Lower Upper Tallman+ Nurse 26.864 21.7011 33.2552 Tallman+ NonhealthcarePro 27.0048 23.1026 31 .566 Tallman+ HealthcarePro 31.2473 25.7719 37.886 Traditional+ NonhealthcarePro 33.5817 28.6799 39.321 1 Traditional+ HealthcarePro 41.911 34.5132 50.8945 Traditional+ Nurse 46.8211 37.7709 58.0397 Table 8 - Time in Seconds and 95% Confidence Intervals 42 Group Across all subjects each group 1-4 had 320 critical image pairs, including TALL Man and traditional lettering. (See Figure 11 - Methods for description of Group and Appendix C for a description of randomization). A main effect of ordered group was evident on time to detect the change (P<0.0001). Overall, detection of change for the first group took longer than for the subsequent groups (2, 3 or 4). No differences were apparent between groups 2, 3 and 4 (See figure 14 and table 9). This suggests that participants became faster as they acclimated to the testing, and is consistent with previously published work. [40] Seconds Group 1 Group 2 Grow 3 Group 4 I Figure 14 — Groups of time to detect a change with standard error bars and significance; Different letters indicate statistical significance at a=0.05 43 Time in Seconds Lower Upper Group! 41.5901 35.657 48.5104 Group 2 32.9792 28.257 38.4907 Group 3 30.8455 26.4423 35.982 Group 4 30.8869 26.469 36.0421 Table 9 - Time in Seconds and 95% Confidence Intervals Quadrant and Position A significant positional effect of change was detected on time to detect the change on the computer screen (P < 0.0001). This positional effect is characterized by the position and quadrant combinations, and included a total of 16 positions (see Appendix C for a discussion of positional randomization and its relationship to quadrant). See table 10 for time to detect in seconds and figure 15 and 16 for visual representation. According to these figures it could be hypothesized, the participants started their search in the top left comer‘(bar position 1-S1 in Figure 16) of the computer screen, and then they began many different types of search patterns. 44 Time in Seconds Lower Upper Quadrant 4 Position 3 52.1089 1 1.0216 13.978 Quadrant 3 Position 3 48.2404 13.256 18.2787 Quadrant 4 Position 4 45.5659 9.1355 1 1.4262 Quadrant 4 Position 1 44.3698 10.7312 14.1546 Quadrant 4 Position 2 40.7908 8.6137 10.9195 Quadrant 3 Position 4 39.5111 7.8105 9.735 Quadrant 3 Position 2 37.8503 10.7565 1 5.0267 Quadrant 2 Position 3 35.6234 9.3685 12.71 15 Quadrant 2 Position 4 35.3615 7.1972 9.0363 Quadrant 1 Position 4 34.4737 8.4982 11.2785 Quadrant 3 Position 1 33.3975 6.7805 8.5077 Quadrant 1 Position 3 31.8691 7.3949 9.6292 Quadrant 2 Position 2 30.6114 5.5667 6.804 Quadrant 2 Position 1 28.7356 7.7128 10.5426 Quadrant 1 Position 2 19.1047 3.7605 4.6822 Quadrant 1 Position 1 12.1089 2.3723 2.9503 45 Table 10 — Time in Seconds and 95% Confidence Intervals mucooow Figure 15 - Quadrant by Position Time and 95% Confidence Intervals 46 Figure 16 - Comparison of 16 positions with dependant variable time to detect; 1-81 is top left comer, 1-S4 is top right comer, 81-84 is the bottom Word Pair Comparisons There was marginal evidence for a difference between word pairs when the dependent variable was time to change detection. Pairs are ranked in Table 11 on the predicted time to change detection. 47 Rank TALL Man Orientation 7 acetoHEXAMlDE acetaZOLAMlDE 8 chlorproMAZlNE chlorproPAMlDE 3 clomiPHENE clomiPRAMlNE 2 DAUNOrubicin DOXOrubicin 4 dimenhyDRlNATE diphenhydrAMlNE 1 medroxyPROGESTERone methleESTOSTERone 6 predniSONE prednisoLONE 5 vinBLAStine vinCRlStine Table 11 - Word pair time to detect, Ranked 1-8 shortest to longest Binary Response — Successful Detection A second analysis was performed in which the binary response variable, “detection of a change,” (yes/no) was modeled using a generalized linear mixed model fitted with the GLIMMIX procedure of SAS software (version 9.2 SAS Institute, Inc., Cary, NC). The statistical model included the fixed effects of profession (students, nurses and other healthcare professionals), graphic presentation (TALL Man and traditional), their two-way interaction, and the ordered group of presentation to the participant (1-4). Also included in the model were fixed positional effects, as per quadrant by position combinations. The following random effects were considered: participant nested within profession, participant by graphic presentation (to account for technical replication), participant by ordered group, and participant by position by quadrant. Run order within group, gender, ethnicity and language were evaluated as fixed effects in the model but were excluded from the final model based on the P value > 0.20. 48 Age was excluded from the final model due to overparameterization/muIticollinearity with the fixed effect factor, profession. Analysis results are presented in Table 12 for the model’s fixed effects. Bolded effects indicate significance at a level of a =0.05. Type 3 Tests of Fixed Effects for the Dependent Variable Successful Detection (yes/no) Effect Num DF Den DF F Value Pr > F Treatment 1 77 23.61 <0.0001 Profession 2 77 7.22 0.0013 Treatment x Profession 2 77 2.10 0.1289 Group 3 237 9.76 <0.0001 Position x Quadrant 15 709 1.65 0.0554 Table 12 - Type 3 Tests of Fixed Effects for the Dependent Variable Successful Detection (yes/no) _‘[_A_LL Man vs Traditional The analysis indicated a significant effect of graphic presentation on the probability of detecting a change (P<0.0001), whereby change presented in a TALL Man format was more likely to be detected than changes presented in a traditional format (95.1: 1.4% and 85.9i3.3%, respectively). See Figure 17. Percentages of Iikeliness to detect are in Table 13. 49 l l 95% - ~ 90% ~ Percent 80% 75% - . TALL Man Traditional Figure 17 - Percentage of Iikeliness to Detect Graphic Presentation Profession Regardless of graphic presentation, the probability of change detection differed by Profession (P = 0.0013), whereby students were more likely to detect change than either nurses or other health care professionals (95811.2, 87.214.0% and 89.2:32, respectively). No difference was apparent between nurses and other health care professionals (See figure 18). 5O 100% A Non Healthcare Pro Healthcare Pro Nurse 1 95% ~ 3 l .. 90% - C I § 0 | Q 85% ‘ l 80% ' 75% A V Figure 18 — Percentage of Iikeliness to Detect between Professions with standard error bars and significance; Different letters indicate statistical significance at a=0.05 Group A main effect of group (see Table 13) was evident when the dependent variable was the probability of a successful detection in under two minutes (P<0.0001). Detection of the change was more likely if it occurred in groups 2-4 as compared with group 1. There was no evidence of difference when groups 2-4 were compared (See Figure 19). This further supports the idea that participants had a period of acclimation as they adjusted to the test. 51 Percent \l ”3 a~ Figure 19 - Percentage of Iikeliness to Detect between Groups with standard error bars and significance; Different letters indicate statistical significance at a=0.05 Quadrant by Position A marginal positional effect of change was detected on the probability of change detection (P = 0.0554). Specifically, a change in quadrant 1 was more likely to be detected than a change in quadrant 2 (P<0.05). See Figure 20. 52 1 ‘ ‘ ‘ch41‘ Yt.‘ I . -,. t. k '- .. 100% Ememn. 50% N £99 ’5 g (1' 0 so <2°9