Embed Size (px)
Citation preview
Communicating risk information: The influence of graphical display
format on quantitative information perception—Accuracy,
comprehension and preferences
Melanie Price a,b,*, Rachel Cameron b, Phyllis Butow a,b
a Medical Psychology Research Unit, University of Sydney, Australiab School of Psychology, University of Sydney, Australia
Received 18 December 2006; received in revised form 1 August 2007; accepted 12 August 2007
www.elsevier.com/locate/pateducou
Patient Education and Counseling 69 (2007) 121–128
Abstract
Objective: Statistical health risk information has proved notoriously confusing and difficult to understand. While past research indicates that
presenting risk information in a frequency format is superior to relative risk and probability formats, the optimal characteristics of frequency
formats are still unclear. The aim of this study is to determine the features of 1000 person frequency diagrams (pictographs) which result in the
greatest speed and accuracy of graphical perception.
Methods: Participants estimated the difference in chance of survival when taking or not taking Drug A, on a pictograph format, varying by mode
(one-graph/two-graph), direction (vertical/horizontal), and shading (shaded/unshaded), and their preferences for the different formats. Their
understanding of different components of the 1000 person diagram was assessed. Responses were timed and scored for accuracy.
Results: Horizontal pictographs were perceived faster and more accurately than vertical formats. Two-graph pictographs were perceived faster
than one-graph formats. Shading reduced response time in two-graph formats, but increased response times in one-graph formats. Shaded and one-
graph pictographs were preferred.
Conclusions: As shading and one-graph formats were preferred, further clarification as to why shading negatively impacts on response times in the
one-graph format is warranted.
Practice Implications: Horizontal pictographs are optimal.
# 2007 Elsevier Ireland Ltd. All rights reserved.
Keywords: Risk communication; Risk perception; Medical decision-making; Graphical format; Quantitative information
1. Introduction
One of the major challenges in promoting informed consent
and shared decision-making relates to the effective commu-
nication of risk and benefit information to patients about
different treatment options [1–5]. Decision aids are increas-
ingly being used to assist with this process. Decision aids are
designed to facilitate shared decision making, and one aspect of
this process is providing quantitative information about the
risks and benefits associated with treatment choices [6]. Despite
this, there is little empirical evidence about how best to
* Corresponding author at: Medical Psychology Research Unit, School of
Psychology, Brennan MacCallum Building (A18), the University of Sydney,
NSW 2006, Australia. Tel.: +61 2 9351 3916; fax: +61 2 9036 5292.
E-mail address: [email protected] (M. Price).
0738-3991/$ – see front matter # 2007 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.pec.2007.08.006
represent quantitative data to improve risk communication
[7,8]. While part of the challenge in communicating risk
information may be due to the inherent difficulties in
understanding probabilistic information [9,10], recent research
suggests that it may be the format of the quantitative
information which may be a source of confusion [11]. For
example, research into cognitive biases generally uses single
event or conditional probabilities (e.g. 0.03 or 3%), relative risk
(e.g. an increase of 20%), or unrelated base rates (e.g. 9/100
versus 1/11), all of which have been found to result in confusion
[12,13]. Of utmost concern are data that suggest that different
treatment decisions are made when treatment outcomes are
described in terms of absolute risk rather than relative risk
[14,15]. Frequency statements (e.g. three women in 100
compared with seven women in the same group of 100), where
the same class of observations are referred to, have been found
to promote insight and facilitate statistical inferences [13].
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128122
Graphical formats are increasingly being used to convey risk
information [5,9,16,17] and have been shown to assist in the
perception, understanding and interpretation of quantitative
information over textual or written formats [18,19]. Despite
this, few systematic studies have examined changes in the
perception or interpretation of risk according to variations in
graphical format [20,21].
Feldman-Stewart et al. [22] have reported a series of studies
comparing the accuracy and speed of perception of quantitative
information, in choice and estimate tasks, across different
presentation formats. For tasks involving a simple choice
(bigger versus smaller), systematic ovals (pictographs with
icons systematically highlighted), vertical bars, horizontal bars
and numbers were equally well perceived, while pie charts and
random ovals (pictographs with icons randomly highlighted)
resulted in slower and less accurate performances. For
estimating differences between risk magnitude (a more detailed
task), numbers led to the most accurate estimates, followed by
systematic ovals. The context of the estimate task (i.e. treatment
context or neutral context) did not affect accuracy.
Fagerlin et al. [9] and Burkell [23] have suggested that
pictographs (such as 100 or 1000 person dot/icon diagrams) are
the most advantageous format for presenting risk information,
as they are likely to be perceived and interpreted more
accurately than other visual formats, such as bar and pie graphs.
Pictographs show the reference class to which the person
belongs, thereby removing the main source of statistical
confusion. Visually presenting cure rates using pictographs, in
contrast to verbal presentation of cure rates, has been found to
significantly reduce the impact of anecdotal information on
hypothetical treatment decisions [9]. Risk information pre-
sented as icons have also been reported as more helpful in
decision-making than risk information presented as numbers or
vertical bars [24]. Improved understanding with presentation of
data in pictographs over numbers alone holds in both older and
younger patients [25,26]. Note that a major benefit of the
pictographs is that they do represent frequency (which is more
easily understood than probability).
A recent qualitative study investigating preferences for the
presentation of risk information between different 1000 dot
pictograph formats, reported a preference for dot diagrams
highlighted as separately coloured blocks rather than indivi-
dually highlighted dots [27]. Horizontal presentations of risk
data were preferred over vertical pictograph presentation [27].
To date, there has been no research systematically examining
the effect of varying pictograph format on the accuracy,
comprehension, and format preferences of quantitative risk
information perception.
Given that the visual communication of health risk
information is still in its infancy [19], the most appropriate
place to begin examining the effects of visual formats and
testing their relative impact on performance, is at one of the
most elementary stages of processing, namely perception
[18,28,29]. Cleveland and McGill’s [28] hierarchy of elemen-
tary perceptual tasks provides some guidance as to the expected
features within a graph that influence speed and accuracy of
performance [18]. From the more accurately processed to less
accurately processed, these perceptual features are: position
along a common scale; positions along non-aligned scales;
length; slope; angle; area; volume; and colour hue, saturation
and density [29]. Graphs with features at the more accurately
processed end of the hierarchy will facilitate a better
organisation of quantitative information and increase the
chances of correct perception of relevant patterns [18,28].
Pictographs fall within the first, most accurately processed end
of the hierarchy.
In the health context, most often the goal of risk
communication involves risk magnitude (how large or small)
and relative risk (the difference between two options).
Carswell’s [30] taxonomy of basic graphical tasks, includes
two tasks often required in risk communication, namely ‘local
comparison’ (comparison between two risks) and ‘point
reading’ (precise risk information). Both fall within Cleveland
and McGill’s most accurately processed tasks [30]. However,
greater computational effort may be required even in the low
level, focused task of ‘local comparison’ (making a comparison
between two graphs) compared with ‘point reading’ (reading a
value from a single graph) [31].
Ideally, a graphical display of this information should
include the relevant information in a form that minimises the
computation effort required to understand it [31]. The addition
of a perceptual feature to a pictograph, such as length (or block
shading as preferred by participants in McCaffery et al. [27])
may add extraneous detail and visual complexity to the graph
without adding any extra information [31]. Alternatively, this
may aid perception by highlighting the area of relevance
[18,28]. According to Tufte [32], this would reduce the data-ink
ratio (the ratio of ink devoted to data compared to the total
amount of ink used for all the other features of a graph), and
thereby potentially reduce accuracy and efficiency of percep-
tion [33]. While the reported preference for shading in
pictographs by McCaffery et al. [27] is relevant, they may
not facilitate performance [16].
With respect to the vertical–horizontal differential of
graphical display, perception research also provides some
guidance. Western text is traditionally presented left to right
along horizontal lines and read more quickly than vertically
aligned text [34]. Seo and Lee [35] found that the Korean
language, written vertically, is also read more quickly hori-
zontally than vertically. This appears to be due to larger eye gaze
amplitudes for horizontal reading, the smaller number of rapid
eye movements, and the longer period of time the eye remains
stable and gathers the information when reading horizontal text
[35]. Whether the same pattern of significantly faster perception
of risk information presented horizontally rather than vertically
also occurs within pictographs is worth examining.
It is possible that framing may influence understanding and
perception of risk. Studies of the effect of verbal framing on
interpretation and decision-making have generally focused on
positive (e.g. 75% chance of survival) versus negative framing
(25% chance of dying). While there is evidence to support the
greater effectiveness of positive framing in persuading people
to take risky or more invasive treatment options, some studies
suggest a negative frame may be more effective in promoting
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128 123
screening uptake [5,14]. The use of neutral framing overcomes
the possibility of framing bias [5].
Drawing on theories of health risk communication and
graphical and text perception, this study was designed to
investigate the impact of systematically varying a number of
features of dot pictographs on perception. There is recent
evidence suggesting that partially shaded figures should be
avoided in frequency pictographs, due to a ‘rounding up’ effect
[23]. Given that frequency risk information is being developed
and used in a wide variety of contexts in addition to treatment
outcomes, such as treatment side effects health screening and
prevention, 1000 person dot diagrams formed the basis of the
experiment. This is particularly relevant in the context of health
screening and prevention, where for example, 15 new breast
cancers may be diagnosed for every 1000 women screened.
Using a 100 person pictograph in this context would require
partial shading.
Three aspects with the potential to affect speed and accuracy
of perception of pictographs were examined:
(a) t
wo simple graphs versus one compound graph: complexgraphs (combining several data points) may be more
efficient and require less visual scanning [35], or may add
processing and computational effort [31], increasing the
likelihood of judgement errors and the time needed to
perform the task [36];
(b) s
hading versus no shading: the addition of length or blockshading may highlight the area of relevance and assist in the
extraction of the quantitative information, or may make it
harder to count the data [31]; and
(c) h
orizontal versus vertical presentation of data: horizontaldata may be perceived more quickly than vertical [35].
With patient preferences increasingly considered in the
healthcare setting, preferences with respect to the different
pictograph formats are also of interest.
Therefore, the aims of the study were to:
(1) s
ystematically evaluate the differential effects of a varietyof pictograph formats displaying hypothetical treatment
Fig. 1. Example of one complex graph, horizontal data display
risks and benefits on the accuracy and speed of perception
of such quantitative information in an estimate task;
(2) a
ssess the level of understanding of the visually presentedrisk information; and
(3) e
xamine participant preferences for pictographic formats.2. Methods
2.1. Participants
A convenience sample of 76 first year psychology students
were recruited via an online research sign up program and
participated in the study for course credit. Eligibility criteria
included normal or corrected-to-normal vision, normal colour
vision and fluency in English. Approval was obtained from the
University of Sydney Human Research Ethics Committee.
2.2. Design and procedure
The study was conducted on computer displaying a series of
pictographs representing a hypothetical treatment risk context.
The quantitative information presented was the chance of
survival and the chance of dying, with and without treatment.
Treatment was a hypothetical drug named ‘Drug A.’ Survival
and mortality data only were presented (rather than also listing
side effects) for simplicity and neutrality of framing.
Participants were instructed to work quickly and accurately.
2.2.1. Estimate Task
The experiment was a three way within-subject design
investigating: mode of graph presentation (two simple graphs
versus one compound graph); direction (vertical versus
horizontal); and shading (shaded versus unshaded). Examples
of pictographs are displayed in Figs. 1 and 2.
Twelve trials of each of the eight possible combinations (96
trials) of graphical displays were presented in two blocks. The
predominant outcome presented (survival or death) were
counterbalanced for each graphical combination. The trials
were divided into two blocks of 48 by mode, with Block A
consisting of the 48 one compound graph formats and Block B
, shaded pictograph (predominant outcome death).
Fig. 2. Example of two simple graphs, vertical data display, unshaded (predominant outcome survival).
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128124
consisting of the 48 two simple graph formats. This was done to
minimise the degree of image change presented from trial to
trial. This enabled the text describing each graph and the legend
related to each graph to be consistently placed for all the one-
graph formats and for all two graph formats. The blocks were
counterbalanced to prevent order effects, with 47.4% of
participants beginning with Block A and 52.6% beginning with
Block B. The order of trials within each block was randomised
for each participant.
2.2.1.1. Stimulus material. The text accompanying Block A
(one compound graph) pictographs read: ‘Of 1000 people in
their 50 s, over 10 years.’ In Block B (two simple graph)
pictographs the text over the first pictograph read ‘Of 1000
people in their 50 s, over 10 years, who DON’T take Drug A’
and over the second graph read’ Of 1000 people in their 50 s,
over 10 years, who DO take Drug A.’ The legend explaining the
meaning of each colour was located on the upper right side of
each card. Yellow represented ‘Chance of survival whether or
not Drug A is taken,’ blue represented ‘Chance of dying
whether or not Drug A is taken,’ and orange (only in the
compound graphs) represented ‘Chance of survival due to Drug
A.’ The question asked in relation to each pictograph was:
‘What is the difference in the chance of survival between those
who don’t take Drug A versus those who do?’ Each block
included an example and three practice trials with feedback.
Response time for each format was calculated on median
response time across the six variations of each pictograph
format. Accuracy was scored as the absolute size of the error in
the difference estimate.
2.2.2. Comprehension Task
Participants were presented with five multiple choice
questions designed to test their understanding of the meaning
of the blue, yellow and orange areas depicted in the pictographs
(see Appendix A). A comprehension score was calculated by
summing the number of correct responses to the multiple
choice questions.
2.2.3. Preferences task
Participants were presented with two of each of the eight
pictograph combinations and were simply asked to rate each
format on a five point Likert scale from 1 ‘do not like at all’ to 5
‘like very much.’
3. Results
The sample comprised 55 women and 21 men, ranging in
age from 18 to 54 years (M = 19.5 years, SD = 4.9 years). Data
were analysed using SPSS 12.0.1 for Windows [SPSS, Inc.].
There were no differences on any of the dependent variables
due to gender or according to the order of completing Blocks A
and B. Thus, all data were collapsed across gender and order of
completion.
3.1. Response time on estimate task
Table 1 summarises the data for response time as a function
of pictograph format. Analysis of variance revealed a
significant difference in response times for direction, with
quicker reaction times for horizontal pictographs than vertical
pictographs (F(1,75) = 32.27, p < 0.001). While there were no
significant differences within mode ( p > 0.05) or shading
( p > 0.05), there was a significant interaction between shading
and mode. In the one-graph condition, shading was associated
with slower response times (mean (standard deviation) shaded
reaction time (RT) = 10.72 (2.63) in seconds versus unshaded
RT = 10.29 (2.54)), whereas in the two-graph condition,
Table 1
Mean (standard deviation) response time in seconds, accuracy (absolute size of
error) and preferences rating as a function of pictograph mode, direction, and
shading (N = 76)
Format Response timea Accuracyb Preferencesc
Horizontal 10.06 (0.25)* 0.16 (0.02)* 3.08 (0.07)
Vertical 10.83 (0.29)* 0.21 (0.03)* 2.98 (0.07)
Two-graph 10.37 (0.29) 0.18 (0.03) 2.55 (0.06)*
One-graph 10.52 (0.29) 0.19 (0.02) 3.51 (0.07)*
Shaded 10.42 (0.26) 0.18 (0.02) 3.67 (0.08)*
Unshaded 10.46 (0.27) 0.19 (0.02) 2.40 (0.09)*
a Mean (standard deviation) response time in seconds.b Mean (standard deviation) of absolute size of error—lower scores equate to
higher accuracy.c Mean (standard deviation) score between 1 and 5—higher scores equate to
greater preference.* p < 0.05.
Table 2
Number of absolute error score as a function of pictograph mode, direction and
shading
Format Absolute error frequencies
0 1 2 3 4 5+
One-graph 3138 380 87 24 6 6
Two-graph 3169 339 73 30 9 7
Vertical 3097* 405 81 29 12 8
Horizontal 3210* 314 79 25 3 5
Shaded 3151 351 92 23 9 10
Unshaded 3156 368 68 31 6 3
* p < 0.05.
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128 125
shading was associated with faster responses times (shaded
RT = 10.10 (2.50) versus unshaded RT = 10.63 (2.62)),
(F(1,75) = 42.50, p < 0.001).
3.2. Accuracy on estimate task
Over 86% of responses were correct and 99.2% of errors
were between 0 and 3 (see Table 2). Analysis of variance
revealed a significant difference in accuracy for direction, with
greater accuracy for the horizontal formats than vertical formats
(F(1,75) = 5.0, p < 0.05). There were no significant differences
within mode or shading, nor their interactions.
3.3. Accuracy on comprehension questions
The mean comprehension score was 4.1 (SD = 1.0) on a
scale of one to five (range 2–5). However, despite the
Table 3
Mean (standard deviation) response time in seconds and accuracy (absolute size
of error) on estimate task as a function of comprehension task score (N = 76)
Comprehension task score Response timea Accuracyb
Score of 5 10.90 (0.28)a 0.10 (0.03)*
Score of <5 10.11 (0.25) 0.25 (0.03)*
a Mean (standard deviation) response time in seconds.b Mean (standard deviation) of absolute size of error—lower scores equate to
higher accuracy.* p < 0.05.
conceptual content being the same, questions involving two
graphs were significantly less likely to be answered correctly
than questions involving one graph (63.8% correct versus
93.0% respectively, t(75) = 5.4, p < 0.001). There were no
significant differences detected in comprehension within
direction or shading (Table 3).
3.4. Preferences
Shaded pictographs were preferred to unshaded pictographs
(F(1,75) = 144.8, p < 0.001) and one-graph presentations
were preferred to two-graph presentations (F(1,75) = 106.9,
p < 0.001) (see Table 1). There were no differences in
preference ratings for direction (F(1,75) = 1.5, p > 0.05).
There were no significant interactions.
4. Discussion and conclusions
4.1. Discussion
This study was conducted to systematically investigate
the effects of different 1000 person pictograph formats on
the speed and accuracy of perception, participant pre-
ferences and conceptual understanding of treatment risk
scenarios. As predicted, based on previous research findings
of horizontal aligned text being read more quickly than
vertically aligned text [34,35], data presented along the
horizontal axis of the pictograph were perceived faster than
vertical presentation. The current results are the first to
demonstrate that risk information presented in horizontal
format is also perceived more accurately than that presented
in vertical format. This link between accuracy and response
time has been previously reported in the graphical perception
literature, indicating that there was no speed-accuracy trade-
off [18,29].
Some findings were contrary to expectations. Firstly, the
lack of differences in speed and accuracy of risk perception
according to mode (one compound graph versus two simple
graphs) was somewhat surprising. However, both point
reading (compound graph condition) and local comparisons
(two simple graph conditions) are considered to be low level
and focused perceptual tasks [30], and it is possible that any
impact of differences in complexity on response time and
accuracy is minimal. Secondly, also contrary to expectation,
was the finding that overall, the addition of shading had no
impact on the speed and accuracy of risk perception [32].
However, there was a significant interaction between shading
and mode, with shading having a detrimental affect on
response time in the one compound graph condition and a
significantly beneficial impact on response times in the two
simple graph condition. Of interest, there were no differences
in accuracy detected, indicating the trade-off was restricted
to perception time only. This result suggests that an increase
in the data-ink ratio with the addition of shading (an
additional perceptual feature of length), reduces the
computational effort required in the local comparison task
(two simple graphs), while interfering or increasing the effort
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128126
required in the point comparison task (one compound graph)
[31].
With respect to preferences, shaded graphs were preferred to
unshaded graphs, providing empirical support for the findings
of McCaffery et al. [27]. A reason for this may be that the
shading seems to unify the mass of dots into a coherent whole
and provides a definable boundary to the 1000 person dot
matrix, potentially a more aesthetically pleasing result. One
compound graph formats were preferred to two simple graphs,
most likely due to the relative coherence and simplicity of
presenting data together in one graph. The compound graph
format also increased saliency of the difference in risk, by
highlighting it in a different colour.
In contrast to McCaffery et al. [27], there was no evidence to
suggest a preference for horizontal pictographs over vertical
pictographs. However, McCaffery et al. [27] ascertained these
preferences using a compound graph format only, and it is
possible that the large preferences participants in the current
study reported for compound graphs over simple graphs and
shaded compared to unshaded graphical presentations over-
shadowed the direction of directional preferences.
This study also explored the level of conceptual under-
standing of treatment risk scenarios, as presented by the
diagrams. Overall, the participants’ comprehension of the
information presented was good. The high level of under-
standing is consonant with previous research results, where
frequency formats (of which a pictograph is a visual
analogue) were usually correctly comprehended by between
50% and 80% of participants, compared with 20–40%
understanding single event and conditional probabilities
[11,13]. Comprehension was greater in the one compound
graph formats than the two-simple graph formats. The
compound graph condition allows the risk information to be
presented in a simple, condensed format. The legend of the
compound graph format specifies which section of the
pictograph relates to the ‘chance of survival whether or not
Drug A is taken,’ the ‘chance of dying whether or not Drug A
is taken,’ and most importantly, the ‘chance of survival due to
Drug A,’ the focus of the estimate task. The legend of the two
simple graph formats, due to its layout, designates a ‘chance
of survival’ section and ‘chance of dying’ section. Therefore,
it is understandable that the answers to the questions were not
directly provided in the key (namely ‘What represents the best
description of the blue area of diagram B?’ and ‘What
represents the best description of the yellow area of diagram
A?’), but required computation in the two-graph format, and
therefore would be more difficult.
Some limitations to this study should be noted, namely the
small sample size and student sample, and it would be useful
to replicate the study in a larger clinical sample. However,
Feldman-Stewart et al. [22] found similarities in risk
perception between students and clinical samples. In this
study only 1000 dot pictographs were presented and
therefore it cannot be assumed that the results are
generalisable to results from using 100 dot pictographs.
Further studies are required to compare and contrast the
accuracy and speed of perception of quantitative risk
information between 100 and 1000 dot pictograph formats.
Perhaps more problematic is the skewness of the accuracy
task data. The majority of participants responded to all items
without error. While this may be a feature of the relatively
straight forward perceptual task, repeating the study in a
clinical sample is required, as they are likely to be slower and
less accurate overall [22].
4.2. Conclusions
This study is the first to systematically investigate the effect
of pictograph format on perception and to provide some support
for the application of general perceptual theories to the
perception of pictograph formats. This is clearly demonstrated
in the main effect result for direction, with the horizontal
presentation of risk data being perceived more quickly and
accurately than vertical format. The use of general perception
theories may guide future investigation into the impact of task
complexity in understanding the significant interaction found
between shading and mode (one-graph versus two-graphs) on
perceptual performance. Future research should move towards
assessing the impact of incorporating more complex data, likely
to be used in medical decision-making, such as the inclusion of
side effects data. Additionally, the effects of other pictograph
formats on perception and understanding should be explored in
clinical populations, in tandem with the influence of numeracy
or quantitative ability.
4.3. Practice implications
These results demonstrate differences in the understanding
of risk perception according to the way quantitative informa-
tion is displayed in 1000 person pictographs. These results
provide useful information for those developing health
promotion materials, patient information materials and
decision aids. Taken together, the results of this study indicate
that horizontal pictographs are perceived both more accurately
and efficiently, and should be the format of choice. The one
compound graph format is preferred to the two simple graph
format, is easier to comprehend, and there is no trade-off in
terms of accuracy of perception between the risk data
presented in one-graph compared to two-graph formats. There
is a clear preference for including shading as an additional
feature. While, shading had no detrimental impact on accuracy
of risk perception, it did result in slower response times in the
compound graph formats (but not the shaded two-graph
formats). This suggests that the addition of shading to the one
graph formats has a trade-off in risk perception in terms of
response time, but not accuracy.
Acknowledgements
Melanie Price is supported by the School of Psychology
Postdoctoral Research Fellowship, University of Sydney.
Phyllis Butow is supported by an NHMRC of Australia
Principal Research Fellowship.
Appendix A. Comprehension task questions
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128 127
References
[1] Ruland CM. Improving patient safety through informatics tools for shared
decision making and risk communication. Int J Med Inform 2004;73:551–
7.
[2] Cantillon P, Irish B, Sales D. Using computers for assessment in medicine.
Br Med J 2004;329:606–9.
[3] O’Connor AM, Legare F, Stacey D. Risk communication in practice: the
contribution of decision aids. Br Med J 2003;327:736–40.
[4] Edwards A. Flexible rather than standardised approaches to communicat-
ing risks in health care. Qual Saf Health Care 2004;13:169–70.
[5] Edwards A, Elwyn G, Covey J, Matthews E, Pill R. Presenting risk
information—a review of the effects of ‘‘framing’’ and other manipula-
tions on patient outcomes. J Health Commun 2001;6:61–82.
[6] O’Connor AM, Rostom A, Fiset V, Tetroe J, Entwistle V, Llewellyn-
Thomas H, Holmes-Rovner M, Barry M, Jones J. Decision aids for
patients facing health treatment or screening decisions: systematic review.
Br Med J 1999;319:731–4.
[7] Edwards A, Thomas R, Williams R, Ellner AL, Brown P, Elwyn G.
Presenting risk information to people with diabetes: evaluating effects and
preferences for different formats by a web-based randomised controlled
trial. Patient Educ Couns 2006;63:336–49.
[8] Waters EA, Weinstein ND, Colditz GA, Emmons K. Formats for improv-
ing risk communication in medical tradeoff decisions. J Health Commun
2006;11:167–82.
[9] Fagerlin A, Wang C, Ubel PA. Reducing the influence of anecdotal
reasoning on people’s health care decisions: is a picture worth a thousand
statistics? Med Decis Making 2005;25:398–405.
M. Price et al. / Patient Education and Counseling 69 (2007) 121–128128
[10] Fagerlin A, Rovner D, Stableford S, Jentoft C, Wei JT, Holmes-Rovner M.
Patient education materials about the treatment of early-stage prostate
cancer: a critical review. Ann Intern Med 2004;140:721–8.
[11] Gigerenzer G, Edwards A. Simple tools for understanding risks: from
innumeracy to insight. Br Med J 2003;327:741–4.
[12] Yamagishi K. When a 12.86% mortality is more dangerous than 24. 14%:
implications for risk communication. Appl Cognitive Psych 1997;11:495–
506.
[13] Hoffrage U, Lindsey S, Hertwig R, Gigerenzer G. Communicating
statistical information. Science 2000;290:2261–2.
[14] Moxey A, O’Connell D, McGettigan P, Henry D. Describing treatment
effects to patients. J Gen Intern Med 2003;18:948–59.
[15] Wills CE, Holmes-Rovner M. Patient comprehension of information for
shared treatment decision making: state of the art and future directions.
Patient Educ Couns 2003;50:285–90.
[16] Feldman-Stewart D, Brundage MD. Challenges for designing
and implementing decision aids. Patient Educ Couns 2004;54:
265–73.
[17] Epstein RM, Alper BS, Quill TE. Communicating evidence for
participatory decision making. J Amer Med Assoc 2004;291:
2359–66.
[18] Cleveland WS, McGill R. Graphical perception: theory, experimentation
and application to the development of graphical methods. J Am Stat Assoc
1984;79:531–54.
[19] Lipkus IM, Hollands JG. The visual communication of risk. J Natl Cancer
Inst Monogr 1999;25:149–63.
[20] Schapira MM, Nattinger AB, McHorney CA. Frequency or probability? A
qualitative study of risk communication formats used in health care. Med
Decis Making 2001;21:459–67.
[21] Schapira MM, Nattinger AB, McAuliffe TL. The influence of graphic
format on breast cancer risk communication. J Health Commun
2006;11:569–82.
[22] Feldman-Stewart D, Kocovski N, McConnell BA, Brundage MD, Mack-
illop WJ. Perception of quantitative information for treatment decisions.
Med Decis Making 2000;20:228–38.
[23] Burkell J. What are the chances? Evaluating risk and benefit information
in consumer health materials. J Med Libr Assoc 2004;92:200–8.
[24] Timmermans D, Molewijk B, Stiggelbout A, Kievit J. Different formats
for communicating surgical risks to patients and the effect on choice of
treatment. Patient Educ Couns 2004;54:255–63.
[25] Fuller R, Dudley N, Blacktop J. Risk communication and older people-
understanding of probability and risk information by medical inpatients
aged 75 years and older. Age Ageing 2001;30:473–6.
[26] Fuller R, Dudley N, Blacktop J. How informed is consent? Understanding
of pictorial and verbal probability information by medical inpatients.
Postgrad Med J 2002;78:543–4.
[27] McCaffery KJ, Trevena L, Barratt A, Clarke J, Armstrong B, O’Connor A.
Examining the acceptability of a decision aid developed for HRT among
consumers from multiethnic backgrounds and health practitioners. In:
International Shared Decision Making Conference; 2005.
[28] Cleveland WS, McGill R, Graphical Perception. Graphical methods for
analyzing scientific data. Science 1985;229:828–33.
[29] Hollands JG, Spence I. The discrimination of graphical elements. Appl
Cognitive Psych 2001;15:413–31.
[30] Carswell CM. Choosing specifiers: an evaluation of the basic tasks model
of graphical perception. Hum Factors 1992;34:535–54.
[31] Sparrow JA. Graphical displays in information systems: some data proper-
ties influencing the effectiveness of alternative forms. Behav Inform
Technol 1989;8:43–56.
[32] Tufte ER. The visual display of quantitative information, 2nd ed.,
Cheshire: Graphics Press; 2001.
[33] Gillan DJ, Richman EH. Minimalism and the syntax of graphs. Hum
Factors 1994;36:619–44.
[34] Laarni J, Simola J, Kojo I, Nasanen R. Reading vertical text from a
computer screen. Behav Inform Technol 2004;23:75–82.
[35] Seo H, Lee C. Head-free reading of horizontally and vertically arranged
texts. Vision Res 2002;42:1325–37.
[36] Lipkus IM, Crawford Y, Fenn K, Biradavolu M, Binder RA, Marcus A,
Mason M. Testing different formats for communicating colorectal cancer
risk. J Health Commun 1999;4:311–24.
