better health graphs volume 1...12 graphs, and an identical questionnaire asking 39 questions...
TRANSCRIPT
Better health graphs
A report of an experimental study of interventionsfor improving graph comprehension
Volume 1
Copyright © NSW Department of Health, July 2006
This work is copyright. It may be reproduced in whole or in part, subject to
the inclusion of an acknowledgment of the source and no commercial usage
or sale.
State Health Publication No: HSP 060048
ISBN 0 7347 3922 2
suggested citation:
Centre for Epidemiology and Research and Hunter Valley Research
Foundation, Better Health Graphs (Volume 1): A report of an experimental
study of interventions for improving graph comprehension. Sydney: NSW
Department of Health, 2006.
produced by:
Centre for Epidemiology and Research
Population Health Division
NSW Department of Health
Locked Mail Bag 961
North Sydney NSW 2059 Australia
Tel: 61 2 9391 9408
Fax: 61 2 9391 9232
further copies of this publication can be obtained
from the NSW Department of Health website at:
www.health.nsw.gov.au
NSW Health Better health graphs – Volume 1 1
Contents
Acknowledgments..........................................3
Executive summary........................................5
Introduction.....................................................7
Methods ..........................................................8
Results...........................................................10
Discussion.....................................................18
Recommendations .......................................21
Conclusion ....................................................23
References ....................................................24
Appendix 1.The control booklet of graphs
Appendix 2. The intervention booklet of graphs
Appendix 3. The questionnaire
Acknowledgments
NSW Health Better health graphs – Volume 1 3
This project was jointly funded by the Commonwealth
Department of Health and Ageing and the Centre for
Epidemiology and Research, NSW Department of Health.
The project was conducted under the National
Population Health Information Development Plan
(Australian Institute of Health and Welfare, 1999).
The Hunter Valley Research Foundation conducted
the project under contract to the NSW Department
of Health. These final reports represent the collaborative
work of the Hunter Valley Research Foundation and
staff from the Centre for Epidemiology and Research,
in particular:
The Hunter Valley Research Foundation
■ Mr Andrew Searles
■ Ms Robin Mcdonald
Centre for Epidemiology and Research, NSW Department of Health
■ Mr David Muscatello
The Centre for Epidemiology and Research would like to
acknowledge the valuable assistance of the additional
members of the project Steering Committee:
■ Dr Tim Churches, Centre for Epidemiology and
Research
■ Dr Paul Jelfs, Australian Institute of Health and
Welfare
■ Dr Louisa Jorm, Centre for Epidemiology and
Research
■ Ms Jill Kaldor, Centre for Epidemiology and Research
Acknowledgment of graphsreproduced for the surveyThe following documents were used to provide example
graphs used as control graphs in the study. Permission to
reproduce the graphs is gratefully acknowledged.
Illustration AAustralian Institute of Health and Welfare (AIHW) 2000,
Australasian Association of Cancer Registries, Cancer in
Australia 1997, Incidence and mortality data for 1997
and selected data for 1998 and 1999, AIHW Cat. no.
CAN 10, Canberra, reproduced with permission from
the Australian Institute of Health and Welfare.
Illustrations B and CDepartment of Human Services 1999, Victorian burden
of disease study: Morbidity, Public Health Division,
Department of Human Services, Melbourne,
reproduced with permission from the Victorian
Department of Human Services.
Illustration DDepartment of Human Services 2000, Victorian burden
of disease study: Mortality, Public Health Division,
Department of Human Services, Melbourne,
reproduced with permission from the Victorian
Department of Human Services.
Illustration EQueensland Health 2001, Health indicators for
Queensland: Central Zone 2001, Public Health Services,
Queensland Health, Brisbane, reproduced with permission
from the State of Queensland (Queensland Health).
Illustration FNSW Health 2000, The health of the people of NSW
– Report of the Chief Health Officer 2000, NSW Health
Department, Sydney, reproduced with permission from
the NSW Department of Health.
Acknowledgments
4 Better health graphs – Volume 1 NSW Health
Illustration GCondon JR, Warman G, Arnold L (editors) 2001,
The health and welfare of Territorians, Epidemiology
Branch, Territory Health Services, Darwin 2001,
reproduced with permission from Northern Territory
Department of Health and Community Services.
Illustration HRidolfo B, Sereafino S, Somerford P and Codde J,
Health measures for the population of Western
Australia: Trends and comparisons, Health Department
of Western Australia, Perth 2000, reproduced
with permission from the Health Department
of Western Australia.
Illustration INSW Health 2000, The health of the people of NSW –
Report of the Chief Health Officer 2000, NSW Health
Department, Sydney, reproduced with permission from
the NSW Department of Health.
Illustration JKee C, Johanson G, White U, McConnell J 1998,
Health indicators in the ACT, Epidemiology Unit, ACT
Dept of Health and Community Care: Health series
No. 13, ACT Government Printer, ACT, reproduced with
permission from the ACT Department of Health and
Community Care.
Illustration KAustralian Bureau of Statistics (ABS) and the Australian
Institute of Health and Welfare (AIHW) 2001, The health
and welfare of Australia’s Aboriginal and Torres Strait
Islander peoples, ABS Cat. no. 4704.0, AIHW Cat. no.
IHW 6, Canberra 2001 (www.abs.gov.au), reproduced
with permission from the Australian Bureau of Statistics
and the Australian Institute of Health and Welfare.
Illustration LCoats MS, Tracey EA, Cancer in NSW: Incidence and
mortality 1999 featuring 30 years of cancer registration,
Cancer Council NSW, Sydney 2001, reproduced with
permission from the NSW Department of Health.
Executive summary
NSW Health Better health graphs – Volume 1 5
IntroductionThis project aimed to recommend ways to improve the
graphical communication of population health statistics
to a broad audience.
It was conceived to explore the hypothesis that much
of the statistical information presented in graphical
form in official population health publications is poorly
understood by people who are not trained in public
health, epidemiology or statistics.
The Centre for Epidemiology and Research (CER)
of the New South Wales (NSW) Department of Health,
Australia, was the lead agency in the project. It was
developed under the National Publication Health
Information Development Plan,1 and was co-funded
by the Australian Government Department of Health
and Ageing and CER’s Program for Enhanced Population
Health Infostructure (PEPHI). CER contracted the project
to the Hunter Valley Research Foundation (HVRF) –
a not-for-profit research institution based in Newcastle,
New South Wales, Australia. A working group that
consisted of representatives from HVRF, CER and the
Australian Institute of Health and Welfare supported
the project.
The project had two parts: a literature review and
an experimental study. The literature review examined
available evidence regarding graph readability.
It is available as Volume 2 of this report at
www.health.nsw.gov.au.
MethodsThe experimental study is reported here. It was a
double-blind, randomised, controlled trial that tested
a variety of changes to the design of existing graphs.
The population studied included staff members of
the NSW public sector health system, regardless of
employment type. Respondents were randomly assigned
to receive either a ‘control’ or ‘intervention’ booklet of
12 graphs, and an identical questionnaire asking 39
questions relating to the interpretation of the graphs.
The ‘control’ graphs were replicas of graphs used
in Australian population health publications.
The ‘intervention’ graphs included one or two
changes to the control graphs that were hypothesised
to improve comprehension of the graph. Questions were
targeted to specific changes, where possible. The success
of the intervention was measured as a prevalence ratio
of the proportion of correct answers in the two groups.
ResultsThe overall response rate was 67%. Demographic
characteristics were similar between the control and
intervention groups, although the intervention group
were more likely to rank themselves as more frequent
graph users and as having good visual ability.
For the control graphs, the proportion of subjects
responding correctly to the 39 interpretation questions
ranged from 13% to 97%. Questions requiring an
understanding of confidence intervals (32%) and age
standardisation (37%) had poor comprehension rates.
(Table 2). There were seven tasks with comprehension
rates of at least 90%.
In terms of the effect of the interventions, the tasks
which benefited most from an intervention were:
changing a pie chart to a bar graph and point reading
the magnitude of a single category (prevalence ratio 3.6,
95% CI 2.8-4.6); changing the y axis of a graph so that
the upward direction represented an increase instead
of a decrease in the plotted quantity and judging the
direction of a trend (2.9, 95% CI 2.1-9.9); including
a footnote to explain an acronym and performing
a task that requires knowledge of the meaning of
the acronym (2.5, 95% CI 1.6-3.8); and making the
axis range of two adjacent graphs match and comparing
the size of a difference between the two series shown
on each graph (2.0, 95% CI 1.7-2.4). Only one
intervention had a clear negative impact.
Success at comprehending the control graphs
was generally lower in subjects without university
qualifications, although an exception was the pie
chart, where twice as many non university-educated as
university-educated control subjects correctly estimated
the magnitude of a category within a pie chart. For
subjects without a university education, the generally
lower success for the control charts was complemented
by a generally greater difference in the prevalence of
Executive summary
6 Better health graphs – Volume 1 NSW Health
correct answers between the intervention and control
groups, although there were no statistically significant
differences in prevalence ratios between the two
education groups.
DiscussionOur findings are of benefit from two perspectives.
First, we were able to quantify the proportion of readers
who could extract some typical statistical interpretations
from a sample of graphs used in Australian official
health publications. Second, we were able to measure
the benefits associated with particular interventions
in a broad sample of readers.
The most dramatic result of the study related to a graph
showing that Aboriginal people in a region of Australia
had an increased risk of mortality compared with the
general Australian population. A combination of
interventions that included a simple title and explanatory
words, rather than numbers, on the vertical axis more
than halved the proportion of subjects who did not
grasp that Aboriginal people had a higher mortality risk.
Less than 60% of subjects could answer a question
that required an understanding that disease incidence
refers to the rate of new cases of disease in a period
of time. Using a non-technical label for incidence had
a statistically significant benefit.
Two statistical techniques and concepts occur frequently
in population health graphs: age standardisation and
confidence intervals. Respondents found it difficult to
understand these concepts. Simple explanatory
footnotes offered improvements of up to 2.5-fold, but
there remained a large proportion who were unable to
make the required interpretations.
We tried two interventions aimed at reducing the
volume of information to be interpreted. Reducing the
number of layers in a stacked layer graph did not offer a
benefit. Removing an independent (categorisation)
variable from a vertical bar graph raised the
comprehension rate by 20% for one task.
A line graph and a grouped vertical bar graph of
multiple disease trends by year performed equally well
for point-reading tasks, but the line graph produced a
marginal improvement in trend judgement in subjects
without a university education. Among university-
educated subjects, a ‘population pyramid’ represented
as a horizontal format line graph improved broad
comparison of the shape of the population distribution
by sex.
Bar graphs out-performed dot graphs, particularly
among those without university qualifications.
A stacked layer graph worked well for some tasks
requiring interpretation of trend and broad comparisons,
but worked poorly for a task requiring the estimation
of a difference between the absolute rate at two
points along a layer.
For simple quantitative tasks such as identifying
minimum and maximum categories or making
comparisons where the differences were distinct, a pie
chart performed as well as a bar chart. It performed
poorly for point readings of the displayed quantity, but
this deficiency could potentially be overcome by labelling
the relevant quantity on each pie segment.
For tasks comparing the relative magnitude of quantities
between two adjacent graphs, a matching scale range
on each graph greatly improved comprehension.
We found strong evidence for ensuring that higher
values of the quantity presented on the graph be in the
upward direction, even if this means the numerical labels
are decreasing in the upward direction.
Recommendations■ Use plain, non-technical language in the graph title
and graph components
■ Use the minimum number of sub-categories
(independent variables) necessary
■ Use conventional line or bar graphs where possible
■ Recognise that the interpretation of confidence
intervals and age standardisation requires technical
knowledge
■ Assist readers to interpret ratios
■ Ensure that quantities (not labels) increase from the
bottom to the top or left to right
■ If using pie charts, label the quantity represented by
each segment on or near the segment
■ If presenting graphs in pairs, ensure the axes have
the same ranges
■ Use line graphs to represent trend information rather
than bar graphs
NSW Health Better health graphs – Volume 1 7
Introduction
This project aimed to recommend ways to improve the
graphical communication of population health statistics
to a broad audience.
It was conceived to explore the hypothesis that much of
the statistical information presented in graphical form in
official population health publications is poorly
understood by people who are not trained in public
health, epidemiology or statistics.
The Centre for Epidemiology and Research (CER) of the
New South Wales (NSW) Department of Health,
Australia, was the lead agency in the project. It was
developed under the National Publication Health
Information Development Plan,1 and was co-funded by
the Australian Government Department of Health and
Ageing and CER’s Program for Enhanced Population
Health Infostructure (PEPHI).
The Centre for Epidemiology and Research contracted
the project to the Hunter Valley Research Foundation
(HVRF) – a not-for-profit research institution based in
Newcastle, New South Wales, Australia. A working
group that consisted of representatives from HVRF,
CER and the Australian Institute of Health and
Welfare supported the project.
The project had two parts: a literature review and
an experimental study. The literature review examined
available evidence regarding graph readability. It is
available as a Volume 2 of this report at
www.health.nsw.gov.au.
This report summarises the findings of the experimental
study and presents its major recommendations for
improving the graphical presentation of population
health statistics.
8 Better health graphs – Volume 1 NSW Health
Methods
Study designThe study was designed as a double-blind, randomised,
controlled trial, with data collected through a self-
completed questionnaire. Subjects were randomly
assigned to receive either a ‘control’ or an ‘intervention’
booklet of graphs. Both groups received an identical
questionnaire that explored subjects’ understanding
of the meaning of the graphs.
Study subjects were blinded to their control or
intervention status. Study personnel and researchers
were blinded to the status of respondents until after
data analysis occurred. Each respondent group was
assigned an arbitrary group identifier that did not reveal
their status, even while analysis of the results was being
undertaken. Data entry personnel were blinded to the
respondent status, as they were not shown the graph
booklet that was returned with the questionnaire.
The status of each group was only revealed after
analysis was complete.
Control and intervention graphsThe ‘control’ booklet (Appendix 1) contained
12 examples of graphs that were reproduced from
their original publication. Graphs were chosen to
represent the kind of information commonly presented
in Australian national and State population health
publications. They covered a range of different graph
styles and numeric measures, including population size,
disease incidence rates, disease prevalence, incidence
rate ratios, and risk of developing disease. Statistical
concepts, such as age standardisation and confidence
intervals, were presented in some graphs.
Graphs for the ‘intervention’ booklet (Appendix 2)
presented the same statistical information as those in
the control booklet, but were subject to one or more
changes. The changes were chosen in an effort to
improve comprehension of the statistical information
depicted by the graph. They were selected on the basis
of findings from the literature review for which evidence
was limited, after considering the nature of graphs used
in population health publications. To limit the number of
graphs and thus respondent workload, more than one
change was made to some graphs. In some cases,
these changes were collectively intended to improve
understanding, while in others, they were chosen
to be as independent as possible.
The details of each intervention are described in Table 2,
along with reduced scale images of both versions of the
12 graphs studied.
QuestionnaireThe questionnaire (Appendix 3) contained several
questions relating to each of the graphs, 39 questions
in total. The questions were designed to assess how well
subjects understood the information presented in the
graphs, and to specifically assess the impact of each
of the changes made for the intervention booklet.
The questionnaire also collected demographic details,
as follows: education level, preferred language, age
group, and sex. Respondents were also asked how
frequently they used graphs, their work title, and
to rate their visual ability to read the graphs presented.
Study sampleThe study population included all employees of the
NSW public sector health system, regardless of the
nature of their work. This population was chosen
for the following reasons:
■ it was anticipated that there would be a poor
response rate from the general public
■ there was a readily available sampling frame
of public sector health employees
■ public sector health employees are an important
audience for population health statistics
The sampling frame included those employees whose
contact details were listed on one of five telephone
directory databases, that listed employees of the main
NSW Department of Health administration, an urban
regional Area Health Service (AHS), a mixed urban/rural
AHS, and two rural AHSs. Six hundred and fifty subjects
were randomly selected from the combined directories.
The directories were not restricted by occupation and
NSW Health Better health graphs – Volume 1 9
Methods
included medical, allied health, managerial, clerical,
policy, maintenance, and other occupations. Those
people who no longer worked at the position indicated
in the database were excluded.
The 650 subjects were then divided randomly into two
groups of 325; the intervention and control groups. Each
subject was posted a package containing a cover letter
signed by the NSW Chief Health Officer inviting their
participation, a questionnaire booklet, a control or
intervention graph booklet, and a reply-paid envelope.
Up to six follow-up reminder calls were made to non-
responders. These calls also allowed ineligible subjects to
be identified. Ineligible subjects were those who no
longer worked for the health service or who were
unknown at the available contact address.
AnalysisUnanswered questions were treated as incorrectly
answered. A prevalence of correct answers to an
interpretation task, that is, a ‘comprehension rate’,
was calculated in the control and intervention groups.
The effect of the interventions on each task was
assessed by calculating the prevalence ratio of the
comprehension rate in the intervention and control
groups with a 95% confidence interval (CI).
Analysis was conducted using SPSS version 10.
10 Better health graphs – Volume 1 NSW Health
Results
Response rate and study sampleOf the 650 subjects selected, 543 were eligible, and of
these, 366 returned completed questionnaires, giving an
overall response rate of 67% (intervention group 67%,
control group 66%).
Sex, age, preferred language, education and work
position were similarly distributed between the control
and intervention arms of the study. Intervention subjects
were somewhat more likely to rate themselves as
frequent graph users than control subjects and were
more likely to rate themselves as having good visual
ability (Table 1).
Comprehension of the unaltered (control) graphsOf the 39 interpretation tasks for the 12 graphs, the
proportion of subjects responding correctly ranged from
13% for a task requiring specific knowledge of an
acronym, to 97% for a task identifying the largest
category in a pie chart. Other tasks with a poor
comprehension rate included judging the direction of a
trend in a line graph in which the y axis represented an
increasing quantity in the downward direction (21%
answered correctly), and estimating a point reading of a
quantity from a pie chart (26%). Questions requiring an
understanding of confidence intervals (32%) and age
standardisation (37%) also had poor comprehension
rates (Table 2).
There were seven tasks with comprehension rates of at
least 90%. These included: choosing the largest (97%)
and smallest (91%) categories and comparing the
magnitude of two categories (95%) from a pie chart;
determining the largest category from a dot graph
(94%); choosing the category with the lowest
proportion at a single point along the x axis in
a grouped vertical bar graph (94%); and broad
judgements of the collective relative magnitude by
sex and rurality of bars on a vertical bar graph with
bars subdivided first by sex and then by rurality
(93% for sex and 90% for rurality) (Table 2).
Effect of interventionsIn terms of the prevalence ratio of correct answers
between the control and intervention groups, the tasks
which benefited most from an intervention were:
changing a pie chart to a bar graph and point reading
the magnitude of a single category (prevalence ratio 3.6,
95% CI 2.8-4.6); changing the y axis of a graph so that
the upward direction represented an increase instead of
decrease in the plotted quantity and judging the
direction of a trend (2.9, 95% CI 2.1-9.9); including
a footnote to explain an acronym and perform a task
that requires knowledge of the meaning of the acronym
(2.5, 95% CI 1.6-3.8); and making the y axis range
of two adjacent graphs match and comparing the size
of a difference between the two series shown on
each graph (2.0, 95% CI 1.7-2.4) (Table 2).
The only intervention that had a clear negative impact
was a combination of reducing the number of layers on
a stacked layer graph and inserting a footnote explaining
the meaning of a layer’s thickness. For a task of judging
the direction of trend in one layer, the prevalence ratio
was 0.8 (95% CI 0.7-0.9) (Table 2).
Influence of educationSuccess at comprehending the control graphs
was generally lower in subjects without university
qualifications. The largest differences were for the
following tasks: judging the statistical significance of
the difference between two categories using confidence
intervals (16% of non-university educated controls
versus 40% of university-educated controls);
understanding the influence of age standardisation
on graph interpretation (23% versus 44%); and judging
the relative magnitude of risk between two series on
a graph when the upward direction on the y axis
represents reducing risk (32% versus 58%).
An exception was the pie chart, where twice as many
non university-educated as university-educated control
subjects correctly estimated the magnitude of a category
within a pie chart (40% versus 19%).
NSW Health Better health graphs – Volume 1 11
For subjects without a university education, the generally
lower success for the control charts was complemented
by a generally greater impact of the interventions,
although there were no statistically significant
differences in prevalence ratios between the two
education groups. The greatest differences were for the
interventions applied to the dot graph with confidence
intervals (“hi-lo-close” graph), which was changed to a
horizontal bar graph with confidence intervals and a
footnote was included for interpreting the confidence
intervals. The prevalence ratio for correctly interpreting
the statistical significance of the difference between two
categories on the graph was 2.5 (95% CI 1.3-4.9) for
subjects without a university education compared with
1.6 (95% CI 1.2-2.0) for subjects with a university
education. For interpreting whether a category was
higher or lower than a reference line representing the
average of all categories on this graph, the prevalence
ratio was 2.3 (95% CI 1.6-3.3) for those without a
university education and 1.4 (95% CI 1.2-1.7) for
those with a university education.
Among university educated subjects, there was
a marginal reduction in the comprehension rate
for one task using a graph with a dual intervention.
The interventions were: changing a horizontal divided
bar graph with two bars for each sex to a side-by-side
divided bar graph with the sides representing each sex;
and including a footnote explaining acronyms used in
the graph. The task involved comparing the relative
magnitude of the two segments within a single bar in
both the control and intervention graph (prevalence ratio
0.9, 95% CI 0.8-1.0). The latter task did not require an
understanding of the acronyms, but the bar segments
represented the quantities labelled by the acronyms,
so the extra reading introduced by the footnote may
have added complexity or confusion for some readers.
Results
Table 1. Sample characteristics
*Work position: Clinical=doctors, nurses, allied health dealing with patients; non-clinical public health/policy=health-related but not dealing directly with patients;other=non-health admin, computing, clerical, maintenance etc.
Category totals may not add to 100% because of missing responses
Characteristic CategoryNumber(N=176) Per cent
Number(N=187) Per cent p-value
Sex Male 53 30.1% 47 25.1% 0.26
Under 34 years 37 21.0% 41 21.9%35-54 years 109 61.9% 106 56.7%55 years and over 27 15.3% 36 19.3%
Preferred Language
English 171 97.2% 183 97.9% 0.53
Education University qualification 116 65.9% 124 66.3% 0.83
Clinical 61 34.7% 76 40.6%Public health/policy 36 20.5% 35 18.7%Other 72 40.9% 70 37.4%
Often 55 31.3% 44 23.5%Occasionally or never 118 67.0% 141 75.4% 0.09
Good 122 69.3% 110 58.8%Average or poor 48 27.3% 74 39.6% 0.02
Self-rated visual ability
Frequency ofgraph use
0.54Age
Work position*
Intervention group Control group
0.53
Tabl
e 2.
Com
paris
on o
f the
pro
port
ion
of c
orre
ct a
nsw
ers
betw
een
the
inte
rven
tion
("In
t.") a
nd c
ontr
ol ("
Con
.") g
roup
s fo
r eac
h in
terv
entio
n te
sted
, and
by
high
est
leve
l of e
duca
tion
atta
ined
.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
Und
erst
and
the
mea
ning
of a
poi
nt
read
ing
of a
n in
cide
nce
rate
80.7
%57
.2%
1.4
(1.2
-1.6
)76
.8%
45.6
%1.
7(1
.2-2
.3)
81.9
%62
.9%
1.3
(1.1
-1.5
)
Line
gra
ph o
f age
-sta
ndar
dise
d in
cide
nce
and
deat
h ra
tes (
verti
cal a
xis)
fo
r all
canc
ers,
by se
x an
d ye
ar
(hor
izon
tal a
xis)
.
1. P
lain
serie
s lab
els:
cha
nged
"I
ncid
ence
..."
to "
New
cas
es
(inci
denc
e)...
" an
d "M
orta
lity.
.." to
"D
eath
s..."
2. F
ootn
ote
expl
aini
ng h
ow to
inte
rpre
t ag
e st
anda
rdis
ed ra
tes.
Und
erst
and
the
influ
ence
of a
ge
stan
dard
isat
ion
on c
ompa
rison
s be
twee
n in
cide
nce
rate
s for
mal
es
and
fem
ales
.
58.0
%36
.9%
1.6
(1.3
-2.0
)42
.9%
22.8
%1.
9(1
.1-3
.3)
65.5
%44
.4%
1.5
(1.2
-1.9
)
For a
sing
le d
isor
der,
estim
ate
the
diff
eren
ce b
etw
een
inci
denc
e ra
tes
betw
een
two
age
poin
ts.
57.4
%57
.8%
1.0
(0.8
-1.2
)51
.8%
47.4
%1.
1(0
.8-1
.6)
60.3
%63
.7%
0.9
(0.8
-1.2
)
Com
pare
an
inci
denc
e ra
te re
adin
g fo
r a d
isor
der b
etw
een
two
sexe
s.85
.2%
88.2
%1.
0(0
.9-1
.1)
83.9
%82
.5%
1.0
(0.9
-1.2
)87
.1%
90.3
%1.
0(0
.9-1
.1)
Und
erst
and
the
shap
e of
the
inci
denc
e ra
te tr
end
by a
ge fo
r one
di
sord
er a
nd o
ne se
x.
69.9
%84
.0%
0.8
(0.7
-0.9
)58
.9%
80.7
%0.
7(0
.6-0
.9)
75.0
%86
.3%
0.9
(0.8
- 1.
0)
Und
erst
and
that
the
topm
ost s
erie
s re
pres
ents
the
over
all r
ate
and
com
pare
the
over
all r
ate
for t
he
sam
e ag
e ra
nge
betw
een
the
two
grap
hs.
89.2
%85
.6%
1.0
(1.0
-1.1
)89
.3%
87.7
%1.
0(0
.9-1
.2)
90.5
%83
.9%
1.1
(1.0
-1.2
)
1. R
educ
ed th
e nu
mbe
r of c
ateg
orie
s of
men
tal d
isor
ders
from
five
to th
ree,
w
ith th
e re
mov
ed c
ateg
orie
s com
bine
d in
to th
e 'o
ther
' cat
egor
y.
2. F
ootn
ote
expl
aini
ng w
hat t
he
thic
knes
s of a
laye
r rep
rese
nts.
A p
air o
f sta
cked
line
gra
phs (
area
or
laye
r gra
phs)
of i
ncid
ent r
ates
of
disa
bilit
y-ad
just
ed li
fe-y
ears
(DA
LYs)
(v
ertic
al a
xis)
for s
elec
ted
men
tal
diso
rder
s, by
age
(hor
izon
tal a
xis)
. Ea
ch g
raph
in th
e pa
ir re
pres
ente
d m
ales
and
fem
ales
, res
pect
ivel
y.
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Sou
rce:
Can
cer i
n A
ustra
lia 1
997,
AIH
W &
AA
CR
200
0.
Illu
stra
tion
A: T
rend
s in
age
-sta
ndar
dise
d in
cide
nce
and
mor
talit
y ra
tes
for
all c
ance
rs(e
xclu
din g
non
-mel
anoc
ytic
ski
n ca
ncer
s), A
ustr
alia
, 198
3-19
98
0
100
200
300
400
500
600 19
8319
8419
8519
8619
8719
8819
8919
9019
9119
9219
9319
9419
9519
9619
9719
98
Inci
denc
e - m
ales
Mor
talit
y - f
emal
es
Mor
talit
y - m
ales
Inci
denc
e - f
emal
es
New cases and deaths per 100,000 population
Not
e: a
ge-s
tand
ardi
sed
rate
s al
low
com
paris
ons
over
yea
rs a
nd b
etw
een
mal
es a
nd fe
mal
es.
Diff
eren
t age
-sta
ndar
dise
d ra
tes
are
not
due
to d
iffer
ence
s in
the
rela
tive
prop
ortio
ns o
f old
er o
ryo
unge
r peo
ple
in e
ach
year
or
sex.
Sou
rce:
Can
cer i
n A
ustra
lia 1
997.
AIH
W &
AA
CR
200
0.
Illu
stra
tion
A: T
rend
s in
age
-sta
ndar
dise
d in
cide
nce
and
deat
h ra
tes
for
all c
ance
rs(e
xclu
ding
non
-mel
anoc
ytic
ski
n ca
ncer
s), A
ustr
alia
, 198
3-19
98
0
100
200
300
400
500
600 19
8319
8419
8519
8619
8719
8819
8919
9019
9119
9219
9319
9419
9519
9619
9719
98
New
cas
es (i
ncid
ence
) - m
ales
Dea
ths
- fe
mal
es
Dea
ths
- m
ales
New
cas
es (
inci
denc
e) -
fem
ales
New cases and deaths per 100,000 populatio
Illu
stra
tion
B: I
ncid
ent D
ALY
Rat
es p
er 1
,000
Pop
ulat
ion
by M
enta
l Dis
orde
r, A
ge a
nd S
ex, V
icto
ria
1996
01020304050
010
2030
4050
6070
80A
ge
DALYs per 1,000 population
Oth
erS
ubst
ance
use
dis
orde
rsS
chiz
ophr
enia
Anx
iety
dis
orde
rsD
epre
ssio
n
Mal
es
01020304050
010
2030
4050
6070
80A
ge
DALYs per 1,000 population
Oth
er
Sub
stan
ce u
se d
isor
ders
Sch
izop
hren
ia
Anx
iety
dis
orde
rs
Dep
ress
ion
Fem
ales
Illu
stra
tion
B: I
ncid
ent D
AL
Y R
ates
per
1,0
00 P
opul
atio
n by
Men
tal D
isor
der,
Age
and
Sex
, Vic
tori
a 19
96
Not
e: T
he th
ickn
ess
of th
e sh
aded
laye
r =
DA
LYs
per
1,00
0 po
pula
tion
for
that
dis
orde
r
01020304050
010
2030
4050
6070
80A
ge
DALYs per 1,000 population
Oth
er
Anx
iety
dis
orde
rs
Dep
ress
ion
Mal
es
01020304050
010
2030
4050
6070
80A
ge
DALYs per 1,000 population
Oth
er
Anx
iety
dis
orde
rs
Dep
ress
ion
Fem
ales
Tabl
e 2
(Con
tinue
d). C
ompa
rison
of t
he p
ropo
rtio
n of
cor
rect
ans
wer
s be
twee
n th
e in
terv
entio
n ("
Int."
) and
con
trol
("C
on."
) gro
ups
for e
ach
inte
rven
tion
test
ed, a
nd
by h
ighe
st le
vel o
f edu
catio
n at
tain
ed.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Com
pare
the
mag
nitu
de o
f YLL
an
d Y
LD fo
r a si
ngle
dis
ease
ca
tego
ry a
nd se
x.
65.9
%74
.9%
0.9
(0.8
-1.0
)69
.6%
71.9
%1.
0(0
.8-1
.2)
64.7
%77
.4%
0.8
(0.7
-1.0
)
Kno
w th
at Y
LD re
pres
ents
"d
isab
ility
bur
den"
and
sele
ct th
e di
seas
e w
ith th
e hi
ghes
t dis
abili
ty
burd
en fo
r a si
ngle
sex.
32.4
%12
.8%
2.5
(1.6
-3.8
)33
.9%
10.5
%3.
2(1
.4-7
.5)
31.9
%14
.5%
2.2
(1.3
-3.6
)
With
in a
sing
le d
isea
se c
ateg
ory,
co
mpa
re th
e m
agni
tude
of Y
LLs
betw
een
sexe
s.
85.8
%88
.8%
1.0
(0.9
-1.1
)83
.9%
89.5
%0.
9(0
.8-1
.1)
87.9
%88
.7%
1.0
(0.9
-1.1
)
Sele
ct th
e di
seas
e w
ith th
e hi
ghes
t D
ALY
val
ue fo
r a si
ngle
sex.
83.0
%67
.9%
1.2
(1.1
-1.4
)80
.4%
61.4
%1.
3(1
.0-1
.7)
85.3
%71
.8%
1.2
(1.0
-1.4
)
Judg
e w
hich
sex
had
the
grea
ter
prop
ortio
n fo
r a si
ngle
inju
ry
cate
gory
.
93.8
%89
.3%
1.1
(1.0
-1.1
)92
.9%
78.9
%1.
2(1
.0-1
.4)
94.8
%95
.2%
1.0
(0.9
-1.1
)
A d
ot g
raph
of t
he p
ropo
rtion
of
hosp
ital s
epar
atio
ns (h
oriz
onta
l axi
s)
by c
ause
s of i
njur
y an
d po
ison
ing
(ver
tical
axi
s) in
a ti
me
perio
d, b
y se
x.
Sex
was
repr
esen
ted
by a
shad
ed o
r no
n-sh
aded
dot
, and
eac
h do
t was
co
nnec
ted
to th
e ve
rtica
l axi
s by
a da
shed
line
.
Cha
nged
the
grap
h ty
pe to
a h
oriz
onta
l ba
r gra
ph w
ith se
x re
pres
ente
d by
di
ffer
ently
shad
ed b
ars.
The
bars
for
each
sex
appe
ared
adj
acen
tly fo
r eac
h in
jury
cat
egor
y al
ong
the
verti
cal a
xis.
Judg
e w
hich
inju
ry c
ateg
ory
had
the
grea
test
pro
porti
on o
f hos
pita
l se
para
tions
with
in a
sing
le se
x.
96.0
%94
.1%
1.0
(1.0
-1.1
)94
.6%
89.5
%1.
1(1
.0-1
.2)
97.4
%97
.6%
1.0
(1.0
-1.0
)
A h
oriz
onta
l div
ided
bar
gra
ph o
f di
seas
e bu
rden
in d
isab
ility
adj
uste
d lif
e ye
ars (
DA
LY) (
horiz
onta
l axi
s), f
or
thre
e ca
tego
ries o
f res
pira
tory
dis
ease
(v
ertic
al a
xis)
and
sex
for a
sing
le y
ear.
DA
LY b
ars w
ere
divi
ded
into
yea
rs o
f lif
e lo
st (Y
LL) a
nd y
ears
live
d w
ith a
di
sabi
lity
(YLD
), be
caus
e D
ALY
=
YLL
+YLD
. Sex
was
repr
esen
ted
as
adja
cent
bar
s with
in e
ach
cate
gory
. D
iffer
ent s
hadi
ng w
as u
sed
for Y
LLs,
YLD
s and
sex.
1. C
hang
ed th
e gr
aph
type
to a
side
-by-
side
div
ided
bar
gra
ph, e
ach
side
re
pres
entin
g a
sing
le se
x. T
he sa
me
shad
ing
was
use
d fo
r bot
h m
ales
and
fe
mal
es.
2. F
ootn
ote
expl
aini
ng a
cron
yms Y
LL,
YLD
and
DA
LY, a
nd st
atin
g th
at
DA
LYs a
re th
e su
m o
f YLL
and
YLD
.
Illu
stra
tion
C: T
he B
urde
n of
Chr
onic
Res
pira
tory
Dis
ease
by
Con
ditio
nan
d Se
x, V
icto
ria
1996
16,0
0012
,000
8,00
04,
000
04,
000
8,00
012
,000
16,0
00
Oth
er
Ast
hma
CO
PD
DA
LYs
Mal
esFe
mal
es
YLL
YLD
YLL
=
Yea
rs o
f Life
Los
t: su
mm
aris
es th
e to
tal y
ears
of l
ife lo
st
fr
om a
ll pe
ople
that
die
pre
mat
urel
y of
the
dise
ase.
Y
LD =
Y
ears
Liv
ed w
ith D
isab
ility
: sum
mar
ises
the
tota
l yea
rs o
f hea
lthy
life
lost
due
to d
isab
ility
in p
eopl
e liv
ing
with
the
dise
ase.
D
ALY
= D
isab
ility
Adj
uste
d Li
fe Y
ears
: tot
al b
urde
n =
the
sum
of Y
LL a
nd
Y
LD: l
ost y
ears
due
to b
oth
deat
h an
d di
sabi
lity.
HO
SP
ITA
L S
EP
AR
ATI
ON
S, C
ause
of I
njur
y or
Poi
soni
ng(a
)---
1998
-99
05
1015
2025
30
Oth
er
Inte
ntio
nal s
elf h
arm
Uns
peci
fied
acci
dent
al e
xpos
ures
Com
plic
atio
ns o
f med
ical
& s
urgi
cal c
are
Tra
nspo
rt a
ccid
ents
Acc
iden
tal f
alls
Exp
osur
e to
inan
imat
e m
echa
nica
l for
ces
(b)
Ass
ault
Mal
es id
entif
ied
as In
dige
nous
Fem
ales
iden
tifie
d as
Indi
geno
us
(a) D
ata
are
from
pub
lic a
nd m
ost p
rivat
e ho
spita
ls.
Cau
se o
f inj
ury
is b
ased
on
the
first
repo
rted
exte
rnal
cau
se w
here
the
prin
cipa
l dia
gnos
is w
as 'i
njur
y, p
oiso
ning
and
cer
tain
oth
er c
onse
quen
ces
of
exte
rnal
cau
ses'
.(b
) Inc
lude
s in
jurie
s du
e to
acc
iden
tal c
onta
ct w
ith m
achi
nery
or o
ther
obj
ects
, acc
iden
tal d
isch
arge
fro
m fi
rear
ms,
exp
losi
ons,
& e
xpos
ure
to n
oise
.
Sou
rce:
AIH
W N
atio
nal H
ospi
tal M
orbi
dity
Dat
abas
e.% o
f inj
ury
or p
oiso
ning
sep
arat
ions
Tabl
e 2
(Con
tinue
d). C
ompa
rison
of t
he p
ropo
rtio
n of
cor
rect
ans
wer
s be
twee
n th
e in
terv
entio
n ("
Int."
) and
con
trol
("C
on."
) gro
ups
for e
ach
inte
rven
tion
test
ed, a
nd
by h
ighe
st le
vel o
f edu
catio
n at
tain
ed.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Rea
d th
e to
tal r
ate
of Y
LLs f
or a
si
ngle
geo
grap
hic
cate
gory
and
sex.
93.8
%80
.2%
1.2
(1.1
-1.3
)89
.3%
71.9
%1.
2(1
.0-1
.5)
96.6
%83
.9%
1.2
(1.1
-1.3
)
Bro
ad ju
dgem
ent o
f the
rela
tive
mag
nitu
de o
f ove
rall
YLL
rate
s be
twee
n m
etro
polit
an a
nd ru
ral
geog
raph
ic c
ateg
orie
s, re
gard
less
of
sex.
94.9
%90
.4%
1.1
(1.0
-1.1
)94
.6%
84.2
%1.
1(1
.0-1
.3)
95.7
%94
.4%
1.0
(1.0
-1.1
)
Bro
ad ju
dgem
ent o
f the
rela
tive
mag
nitu
de o
f ove
rall
YLL
rate
s be
twee
n se
xes,
rega
rdle
ss o
f ge
ogra
phic
cat
egor
y.
92.6
%92
.5%
1.0
(0.9
-1.1
)89
.3%
84.2
%1.
1(0
.9-1
.2)
94.8
%96
.0%
1.0
(1.0
-1.1
)
Bro
ad c
ompa
rison
bet
wee
n m
ales
an
d fe
mal
es o
f the
ove
rall
popu
latio
n co
unt a
cros
s a ra
nge
of
age
grou
ps, f
or o
ne g
eogr
aphi
c ar
ea.
90.3
%78
.1%
1.2
(1.1
-1.3
)85
.7%
77.2
%1.
1(0
.9-1
.3)
93.1
%78
.2%
1.2
(1.1
-1.3
)
Bro
ad c
ompa
rison
of t
he to
tal
popu
latio
n si
ze o
f the
two
geog
raph
ic re
gion
s, re
gard
less
of
age
or se
x.
78.4
%41
.2%
1.9
(1.6
-2.3
)73
.2%
29.8
%2.
5(1
.6-3
.8)
81.9
%46
.8%
1.8
(1.4
-2.2
)
Bro
ad c
ompa
rison
of t
he
popu
latio
n si
ze o
f you
nger
and
ol
der s
egm
ents
of t
he p
opul
atio
n fo
r bot
h ge
ogra
phic
are
as.
89.2
%85
.6%
1.0
(1.0
-1.1
)83
.9%
80.7
%1.
0(0
.9-1
.2)
92.2
%87
.9%
1.1
(1.0
-1.1
)
Rem
oved
one
inde
pend
ent v
aria
ble,
the
dise
ase
grou
ping
s, re
sulti
ng in
un
divi
ded
bars
. Thi
s als
o re
sulte
d in
no
lege
nd a
nd a
shor
ter t
itle
as o
nly
over
all t
otal
s wer
e no
w re
pres
ente
d by
ea
ch b
ar.
Ver
tical
div
ided
bar
gra
ph o
f the
m
orta
lity
burd
en in
yea
rs o
f life
lost
(Y
LL) r
ates
(ver
tical
axi
s) b
y th
ree
geog
raph
ic c
ateg
orie
s and
sex
(hor
izon
tal a
xis)
, with
eac
h ba
r div
ided
in
to fo
ur m
ajor
dis
ease
gro
ups.
The
geog
raph
ic c
ateg
orie
s wer
e pr
esen
ted
in tw
o gr
oups
by
sex
on th
e ho
rizon
tal
axis
.
A p
air o
f pyr
amid
-sty
le si
de-b
y-si
de
bar g
raph
s ('p
opul
atio
n py
ram
ids')
sh
owin
g po
pula
tion
coun
ts (h
oriz
onta
l ax
is) b
y ag
e gr
oup
(ver
tical
axi
s) a
nd
sex
for t
wo
geog
raph
ic a
reas
. The
ge
ogra
phic
are
a on
the
leftm
ost g
raph
w
as a
zon
e w
ithin
the
othe
r geo
grap
hic
area
.
For e
ach
geog
raph
ic a
rea,
the
popu
latio
n co
unts
for e
ach
sex
wer
e re
pres
ente
d as
two
serie
s on
a ho
rizon
tal f
orm
at li
ne g
raph
.
Illu
stra
tion
D: R
ates
of Y
LLs
by R
ural
itySt
atus
, Sex
and
Maj
or C
ause
s of
Dea
th
020406080
Met
roRu
ral
town
sO
ther
rura
lM
etro
Rura
lto
wns
Othe
rru
ral
Rate of YLLs per 1,000 populatio
Card
iova
scul
arCa
ncer
In
jurie
sO
ther
caus
es
Mal
esFe
mal
es
Illu
stra
tion
D: R
ates
of Y
LLs
by R
ural
ity, S
tatu
s an
d Se
x
020406080
Met
roRu
ral
town
sOt
her
rura
lM
etro
Rura
lto
wns
Oth
erru
ral
Rate of YLLs per 1,000 populatio
Mal
esFe
mal
es
Illus
trat
ion
E: E
stim
ated
resi
dent
pop
ulat
ion
by a
ge, s
ex a
nd H
ealth
Ser
vice
Dis
trict
, 199
9an
d di
ffere
nce
in a
ge s
truct
ure
betw
een
Heal
th S
ervi
ce D
istri
ct p
opul
atio
n an
d Q
ueen
slan
d po
pula
tion
Cent
ral z
one
Mal
eFe
mal
eQ
ueen
slan
d
6000
040
000
2000
00
2000
040
000
6000
0
0-4
5-9
10-1
4
15-1
9
20-2
4
25-2
9
30-3
4
35-3
9
40-4
4
45-4
9
50-5
4
55-5
9
60-6
4
65-6
9
70-7
4
75-7
9
80-8
4
85+
Age group
Num
ber o
f per
sons
2000
0010
0000
010
0000
2000
00
0-4
5-9
10-1
4
15-1
9
20-2
4
25-2
9
30-3
4
35-3
9
40-4
4
45-4
9
50-5
4
55-5
9
60-6
4
65-6
9
70-7
4
75-7
9
80-8
4
85+
Age group
Num
ber o
f per
sons
Illus
trat
ion
E: E
stim
ated
resi
dent
pop
ulat
ion
by a
ge, s
ex a
nd H
ealth
Ser
vice
Dis
tric
t, 19
99an
d di
ffere
nce
in a
ge s
truc
ture
bet
wee
n H
ealth
Ser
vice
Dis
tric
t pop
ulat
ion
and
Que
ensl
and
popu
latio
n
Cen
tral
zon
e
Que
ensl
and
0
2000
0
4000
0
6000
0
0-4
5-9
10-1
415
-19
20-2
425
-29
30-3
435
-39
40-4
445
-49
50-5
455
-59
60-6
465
-69
70-7
475
-79
80-8
485
+
Age
gro
up
Number of persons
Mal
eFe
mal
e
0
1000
00
2000
00
0-4
5-9
10-1
415
-19
20-2
425
-29
30-3
435
-39
40-4
445
-49
50-5
455
-59
60-6
465
-69
70-7
475
-79
80-8
485
+
Age
grou
p
Number of persons
Mal
eFe
mal
e
Tabl
e 2
(Con
tinue
d). C
ompa
rison
of t
he p
ropo
rtio
n of
cor
rect
ans
wer
s be
twee
n th
e in
terv
entio
n ("
Int."
) and
con
trol
("C
on."
) gro
ups
for e
ach
inte
rven
tion
test
ed, a
nd
by h
ighe
st le
vel o
f edu
catio
n at
tain
ed.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Inte
rpre
t the
stat
istic
al si
gnifi
canc
e of
the
diff
eren
ce in
pro
porti
ons
repr
esen
ted
by tw
o ba
rs w
ith
clea
rly o
verla
ppin
g co
nfid
ence
ra
nges
.
54.5
%31
.6%
1.7
(1.4
-2.2
)39
.3%
15.8
%2.
5(1
.3-4
.9)
62.9
%40
.3%
1.6
(1.2
-2.0
)
Judg
e w
heth
er th
e pr
opor
tion
amon
g m
othe
rs b
orn
in o
ne c
ount
ry
was
hig
her o
r low
er th
an th
at o
f the
ot
her c
ount
ry, w
here
the
diff
eren
ce
is d
istin
ct.
91.5
%84
.5%
1.1
(1.0
-1.2
)92
.9%
71.9
%1.
3(1
.1-1
.5)
91.4
%90
.3%
1.0
(0.9
-1.1
)
Judg
e w
heth
er th
e pr
opor
tion
amon
g m
othe
rs b
orn
in o
ne c
ount
ry
was
hig
her o
r low
er th
an a
ll m
othe
rs o
vera
ll.
79.5
%50
.3%
1.6
(1.4
-1.9
)80
.4%
35.1
%2.
3(1
.6-3
.3)
80.2
%58
.1%
1.4
(1.2
-1.7
)
Bro
ad ju
dgem
ent o
f whe
ther
A
borig
inal
peo
ple
over
all h
ad a
hi
gher
risk
of d
eath
than
mos
t A
ustra
lians
.
82.4
%58
.8%
1.4
(1.2
-1.6
)69
.6%
38.6
%1.
8(1
.3-2
.6)
90.5
%69
.4%
1.3
(1.2
-1.5
)
For a
spec
ific
age
grou
p an
d se
x,
read
the
poin
t est
imat
e of
the
ratio
of
the
the
two
popu
latio
n gr
oups
' ra
tes.
83.0
%55
.6%
1.5
(1.3
-1.7
)69
.6%
36.8
%1.
9(1
.3-2
.8)
91.4
%65
.3%
1.4
(1.2
-1.6
)
Und
erst
and
the
mea
ning
of a
dea
th
rate
ratio
for a
spec
ific
age
grou
p an
d se
x.
84.7
%59
.9%
1.4
(1.2
-1.6
)71
.4%
42.1
%1.
7(1
.2-2
.4)
92.2
%69
.4%
1.3
(1.2
-1.5
)
A v
ertic
al b
ar g
raph
com
parin
g th
e ra
tio o
f dea
th ra
tes (
verti
cal a
xis)
be
twee
n no
n-A
borig
inal
peo
ple
of a
ge
ogra
phic
regi
on w
ith th
ose
of th
e co
untry
's ov
eral
l pop
ulat
ion
over
a
perio
d of
yea
rs, b
y ag
e (h
oriz
onta
l ax
is) a
nd se
x. T
he tw
o se
xes w
ere
disp
laye
d as
adj
acen
t bar
s for
eac
h ag
e gr
oup.
1. P
lain
gra
ph ti
tle st
atin
g th
e pr
imar
y qu
estio
n th
e gr
aph
answ
ered
: "…
how
m
any
times
mor
e lik
ely
to d
ie w
as a
n A
borig
inal
per
son
com
pare
d w
ith a
ll A
ustra
lians
for e
ach
sex
and
age
grou
p?",
inst
ead
of "
…A
borig
inal
: A
ustra
lian
deat
h ra
te ra
tios…
".
2. C
hang
ed th
e ve
rtica
l axi
s lab
els
indi
catin
g ra
tios o
f 1, 5
and
10
to
"Equ
ally
as l
ikel
y", "
Five
tim
es a
s lik
ely"
and
"Te
n tim
es a
s lik
ely"
re
spec
tivel
y.
A d
ot g
raph
with
95%
con
fiden
ce
inte
rval
s ("h
i-lo-
clos
e gr
aph"
) co
mpa
ring
the
prop
ortio
n of
birt
hs th
at
wer
e pr
emat
ure
(hor
izon
tal a
xis)
by
the
mot
her's
cou
ntry
of b
irth
(ver
tical
ax
is).
A v
ertic
al re
fere
nce
line
indi
cate
d th
e ov
eral
l pro
porti
on fo
r all
mot
hers
.
1. C
hang
ed th
e gr
aph
type
to a
ho
rizon
tal b
ar g
raph
.
2. F
ootn
ote
givi
ng a
pra
ctic
al
expl
anat
ion
of c
onfid
ence
inte
rval
s and
th
eir i
nter
pret
atio
n.
Pre
mat
ure
birt
hs b
y co
untr
y of
birt
h of
mot
her,
NS
W 1
994
to 1
998
C
ount
ry o
f birt
h
Not
e:
C
onfid
ence
inte
rval
s in
dica
te s
tatis
tical
unc
erta
inty
abo
ut e
ach
valu
e on
the
grap
h. L
onge
r in
terv
als
m
ean
mor
e un
cert
aint
y. W
hen
two
inte
rval
s ov
erla
p th
en th
ere
is m
ore
unce
rtai
nty
that
the
two
grou
ps a
re r
eally
diff
eren
t.
Birt
hs w
here
ges
tatio
nal a
ge w
as le
ss th
at 3
7 w
eeks
wer
e cl
assi
fied
as p
rem
atur
e bi
rths.
Inf
ants
of a
t lea
st 4
00 g
ram
s
birt
h w
eigh
t or
at le
ast 2
0 w
eeks
ges
tatio
n w
ere
incl
uded
.S
ourc
e:
N
SW
Mid
wiv
es D
ata
Col
lect
ion
(HO
IST
). E
pide
mio
logy
and
Sur
veill
ance
Bra
nch,
NS
W H
ealth
Dep
artm
ent
02
46
810
12
All
Uni
ted
Sta
tes
Egy
ptP
olan
dM
alta
Mal
aysi
aS
outh
Afri
caFiji
Net
herla
nds
Indi
aG
erm
any
Hon
g K
ong
Gre
ece
Phi
lippi
nes
Leba
non
Vie
tnam
Chi
naF
orm
er Y
ugo.
Italy
New
Zea
land
Uni
ted
Kin
gdom
Aus
tral
ia
Per
cen
t
Illu
stra
tio
n G
: N
ort
her
n T
erri
tory
(N
T)
Ab
ori
gin
al:
Au
stra
lian
dea
th r
ate
rati
os
1991
to
199
5
Not
e:
Rat
io o
f NT
Abo
rigin
al to
Aus
tral
ian
deat
h ra
tes
for
all c
ause
s
by f
ive-
year
age
gro
ups
Sou
rce:
Dem
psey
& C
ondo
n 19
99
12345678910
0-4
5-9
10-1
415
-19
20-2
425
-29
30-3
435
-39
40-4
445
-49
50-5
455
-59
60-6
465
-69
70-7
475
+
Age
gro
ups
(yea
rs)
Rate ratio
Mal
es
Fem
ales
Illus
tratio
n G
: Bet
wee
n 19
91 a
nd 1
995,
how
man
y tim
es m
ore
likel
y to
die
was
a No
rther
n Te
rrito
ry (N
T) A
borig
inal
per
son
com
pare
d w
ith a
ll Au
stra
lians
for e
ach
sex
and
age
grou
p?
12345678910
0-4
5-9
10-1
415
-19
20-2
425
-29
30-3
435
-39
40-4
445
-49
50-5
455
-59
60-6
465
-69
70-7
475
+
Age
grou
ps (y
ears
)
Mal
es
Fem
ales
Equa
lly a
s lik
ely
Ten
times
as
like
ly
Five
tim
es
as li
kely
Sour
ce:
Dem
psey
& C
ondo
n 19
99
Tabl
e 2
(Con
tinue
d). C
ompa
rison
of t
he p
ropo
rtio
n of
cor
rect
ans
wer
s be
twee
n th
e in
terv
entio
n ("
Int."
) and
con
trol
("C
on."
) gro
ups
for e
ach
inte
rven
tion
test
ed, a
nd
by h
ighe
st le
vel o
f edu
catio
n at
tain
ed.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Judg
e th
e re
lativ
e m
agni
tude
of r
isk
betw
een
the
sexe
s at o
ne p
oint
al
ong
the
horiz
onta
l axi
s.
79.5
%48
.7%
1.6
(1.4
-1.9
)66
.1%
31.6
%2.
1(1
.4-3
.2)
87.1
%58
.1%
1.5
(1.3
-1.8
)
For a
sing
le se
x, ju
dge
the
dire
ctio
n of
the
trend
ove
r tim
e in
term
s of
incr
easi
ng o
r dec
reas
ing
risk.
60.2
%20
.9%
2.9
(2.1
-9.9
)62
.5%
19.3
%3.
2(1
.8-5
.7)
58.6
%21
.8%
2.7
(1.9
-3.9
)
Rea
d th
e po
int e
stim
ate
of th
e ris
k fo
r a si
ngle
sex
in a
sing
le y
ear.
90.9
%85
.6%
1.1
(1.0
-1.1
)78
.6%
77.2
%1.
0(1
.0-1
.4)
97.4
%91
.1%
1.1
(0.9
-1.1
)
Bro
ad ju
dgm
ent o
f whe
ther
bot
h vi
ruse
s had
the
sam
e re
lativ
e di
ffer
ence
in p
reva
lenc
e be
twee
n th
e tw
o in
ject
ing
hist
ory
grou
ps in
a
sing
le y
ear.
This
requ
ired
a co
mpa
rison
bot
h w
ithin
and
be
twee
n th
e tw
o gr
aphs
.
90.9
%45
.5%
2.0
(1.7
-2.4
)89
.3%
35.1
%2.
5(1
.8-3
.7)
93.1
%51
.6%
1.8
(1.5
- 2.2
)
Judg
e w
hich
inje
ctin
g hi
stor
y gr
oup
had
a lo
wer
pre
vale
nce
of
HC
V in
fect
ion
over
the
year
s sh
own
on th
e gr
aph.
80.7
%75
.9%
1.1
(1.0
-1.2
)78
.6%
66.7
%1.
2(1
.0-1
.5)
81.9
%79
.8%
1.0
(0.9
-1.2
)
Bro
ad ju
dgm
ent o
f the
whi
ch v
irus
had
a gr
eate
r pre
vale
nce
of
infe
ctio
n in
a si
ngle
yea
r, re
gard
less
of i
njec
ting
hist
ory.
Thi
s re
quire
d a
com
paris
on b
etw
een
the
two
grap
hs.
92.0
%63
.6%
1.5
(1.3
-1.6
)87
.5%
47.4
%1.
9(1
.4-2
.5)
94.8
%73
.4%
1.3
(1.2
-1.5
)
Rea
d a
poin
t est
imat
e of
HC
V
infe
ctio
n pr
eval
ence
for a
sing
le
year
and
inje
ctin
g hi
stor
y ca
tego
ry.
71.0
%73
.3%
1.0
(0.9
-1.1
)64
.3%
63.2
%1.
0(0
.8-1
.3)
74.1
%78
.2%
1.0
(0.8
-1.1
)
Line
gra
ph o
f the
life
time
risk
of
expe
rienc
ing
lung
can
cer (
verti
cal a
xis)
by
yea
r (ho
rizon
tal a
xis)
and
sex.
Ris
k la
bels
on
the
verti
cal a
xis w
ere
expr
esse
d as
a n
umbe
r, x,
whe
re th
e nu
mbe
r rep
rese
nted
a o
ne in
x ri
sk.
Num
bers
incr
ease
d fr
om th
e bo
ttom
to
the
top
of th
e ax
is, s
uch
that
hig
her
valu
es m
eant
low
er ri
sk.
The
verti
cal a
xis w
as re
vers
ed so
that
la
bels
wen
t fro
m a
larg
e nu
mbe
r at t
he
botto
m to
a sm
alle
r num
ber a
t the
top.
Th
is m
eant
that
gra
ph v
alue
s tow
ards
th
e to
p of
the
grap
h re
pres
ente
d hi
gher
ris
k.
A p
air o
f gra
phs s
how
ing
the
prev
alen
ce (v
ertic
al a
xis)
of h
avin
g an
tibod
ies t
o hu
man
imm
unod
efic
ienc
y vi
rus (
HIV
) and
hep
atiti
s C v
irus
(HC
V) a
mon
g cl
ient
s of n
eedl
e an
d sy
ringe
pro
gram
s, by
yea
r (ho
rizon
tal
axis
) and
two
cate
gorie
s of i
njec
ting
hist
ory.
The
two
grap
hs re
pres
ente
d H
IV a
nd H
CV
resp
ectiv
ely.
1. M
ade
the
verti
cal a
xis o
n bo
th
grap
hs c
over
the
sam
e ra
nge,
0-1
00%
. B
ecau
se th
e pr
eval
ence
of H
IV
infe
ctio
ns w
as v
ery
low
, thi
s ver
tical
ly
com
pres
sed
the
serie
s in
the
left
grap
h.
2. P
lain
er g
raph
title
: "Pr
eval
ence
of
hum
an im
mun
odef
icie
ncy
viru
s (H
IV)
and
hepa
titis
C v
irus (
HC
V)
infe
ctio
n…"
inst
ead
of "
Ant
ibod
ies t
o hu
man
imm
unod
efic
ienc
y vi
rus (
HIV
) an
d he
patit
is C
viru
s (H
CV
)…".
Illus
trat
ion
H:L
ifetim
e ris
kfo
r lun
g ca
ncer
to a
ge 7
4 ye
ars
5152535455565
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
Mal
es -
Wes
tern
Aus
tralia
Fem
ales
- W
este
rn A
ustra
liaO
ne in
Illus
tratio
n H:
Life
time
risk
for l
ung
canc
er to
age
74
year
s
5 15 25 35 45 55 6519
8319
8419
8519
8619
8719
8819
8919
9019
9119
9219
9319
9419
9519
96
Mal
es -
Wes
tern
Aus
tralia
Fem
ales
- W
este
rn A
ustra
liaO
ne in
Illus
tratio
n I:
Antib
odie
s to
hum
an im
mun
odef
icien
cy v
irus (
HIV)
and
hep
atiti
s C
viru
s (HC
V)
by i
njec
ting
hist
ory,
clien
ts o
f nee
dle a
nd s
yrin
ge p
rogr
ams,
NSW
199
5 to
1998
HIV
HCV
012345 1995
1996
1997
1998
Year
Injec
ting <
3 ye
ars
Injec
ting 3
+ ye
ars
Per c
ent a
ntibo
dy p
ositiv
e
020406080100 19
9519
9619
9719
98
Year
Injec
ting <
3 ye
ars
Injec
ting 3
+ ye
ars
Per c
ent a
ntibo
dy p
ositiv
e
Illustr
ation
I: Pr
evale
nce o
f hum
an im
muno
defic
iency
viru
s (HI
V) an
d hep
atitis
C vi
rus (
HCV)
infec
tion
by i
njecti
ng hi
story
, clie
nts o
f nee
dle an
d syr
inge p
rogr
ams,
NSW
1995
to 19
98
HIV
HCV
020406080100 19
9519
9619
9719
98
Year
Injec
ting <
3 yea
rsInj
ectin
g 3+
years
Per c
ent a
ntibo
dy po
sitive
020406080100 19
9519
9619
9719
98
Year
Injec
ting <
3 yea
rsInj
ectin
g 3+ y
ears
Per c
ent a
ntibo
dy po
sitive
Tabl
e 2
(Con
tinue
d). C
ompa
rison
of t
he p
ropo
rtio
n of
cor
rect
ans
wer
s be
twee
n th
e in
terv
entio
n ("
Int."
) and
con
trol
("C
on."
) gro
ups
for e
ach
inte
rven
tion
test
ed, a
nd
by h
ighe
st le
vel o
f edu
catio
n at
tain
ed.
Gra
ph d
escr
iptio
nIn
terv
entio
n(s)
Inte
rpre
tatio
n ta
skIn
t. %
(N
=176
)C
on. %
(N
=187
)In
t. %
(N
=56)
Con
. %
(N=5
7)In
t. %
(N
=116
)C
on. %
(N
=124
)
All
resp
onde
nts
Non
uni
vers
ity-q
ualif
ied
Rat
io(9
5% C
.I.)
Uni
vers
ity-q
ualif
ied Rat
io(9
5% C
.I.)
Rat
io(9
5% C
.I.)
Rea
d th
e po
int e
stim
ate
of th
e pr
opor
tion
of d
eath
s cau
sed
by a
di
seas
e ca
tego
ry in
one
yea
r.
83.0
%82
.9%
1.0
(0.9
-1.1
)78
.6%
73.7
%1.
1(0
.9-1
.3)
86.2
%88
.7%
1.0
(0.9
-1.1
)
Judg
e w
hich
dis
ease
cat
egor
y ha
d th
e lo
wes
t pro
porti
on o
f dea
ths i
n a
sing
le y
ear.
96.6
%94
.1%
1.0
(1.0
-1.1
)96
.4%
87.7
%1.
1(1
.0-1
.2)
97.4
%97
.6%
1.0
(1.0
-1.0
)
Judg
e w
hich
dis
ease
cat
egor
y ha
d th
e m
ost i
ncre
asin
g tre
nd in
the
prop
ortio
n of
dea
ths o
ver t
he
perio
d of
the
grap
h.
83.5
%76
.5%
1.1
(1.0
-1.2
)75
.0%
56.1
%1.
3(1
.0-1
.8)
89.7
%85
.5%
1.1
(1.0
-1.2
)
Iden
tify
the
canc
er c
ateg
ory
acco
untin
g fo
r the
larg
est
prop
ortio
n of
can
cers
in a
sing
le
sex.
97.7
%96
.8%
1.0
(1.0
-1.1
)96
.4%
93.0
%1.
0(1
.0-1
.1)
99.1
%10
0.0%
1.0
(1.0
-1.0
)
Iden
tify
the
larg
er o
f tw
o ca
tego
ries
for a
sing
le se
x.96
.6%
95.2
%1.
0(1
.0-1
.1)
94.6
%93
.0%
1.0
(1.0
-1.1
)98
.3%
97.6
%1.
0(1
.0-1
.1)
Iden
tify
the
sex
havi
ng th
e gr
eate
r co
ntrib
utio
n of
the
sam
e ca
ncer
ca
tego
ry. T
his r
equi
red
com
paris
on
acro
ss th
e tw
o gr
aphs
.
95.5
%63
.6%
1.5
(1.3
-1.7
)92
.9%
80.7
%1.
2(1
.0-1
.3)
97.4
%56
.5%
1.7
(1.5
-2.0
)
Iden
tify
the
canc
er c
ateg
ory
acco
untin
g fo
r the
smal
lest
pr
opor
tion
of c
ance
rs in
a si
ngle
se
x.
96.6
%90
.9%
1.1
(1.0
-1.1
)94
.6%
91.2
%1.
0(0
.9-1
.2)
98.3
%91
.9%
1.1
(1.0
-1.1
)
Estim
ate
the
poin
t rea
ding
of t
he
prop
ortio
n of
can
cers
for a
sing
le
cate
gory
for a
sing
le se
x.
92.0
%25
.7%
3.6
(2.8
-4.6
)91
.1%
40.4
%2.
3(1
.6-3
.1)
93.1
%19
.4%
4.8
(3.4
-6.9
)
A p
air o
f pie
cha
rts sh
owin
g th
e pr
opor
tiona
l con
tribu
tion
of in
divi
dual
ca
ncer
cat
egor
ies t
o al
l can
cers
in
child
ren.
Eac
h gr
aph
repr
esen
ts a
si
ngle
sex.
Cha
nged
the
grap
h ty
pe to
a h
oriz
onta
l ba
r gra
ph w
ith c
ance
r cat
egor
y on
the
verti
cal a
xis a
nd p
ropo
rtion
as a
pe
rcen
tage
on
the
horiz
onta
l axi
s.
A v
ertic
al b
ar g
raph
show
ing
the
prop
ortio
n of
dea
ths (
verti
cal a
xis)
co
ntrib
uted
by
each
of f
ive
maj
or
dise
ase
cate
gorie
s, by
yea
r (ho
rizon
tal
axis
). Th
e ba
rs fo
r the
dis
ease
ca
tego
ries w
ere
grou
ped
toge
ther
at
each
yea
r alo
ng th
e ho
rizon
tal a
xis.
Cha
nged
the
grap
h ty
pe to
a li
ne g
raph
w
ith e
ach
serie
s rep
rese
ntin
g th
e tre
nd
in e
ach
dise
ase
cate
gory
ove
r tim
e.
Illus
trat
ion
J: P
rinc
ipal
cau
ses o
f dea
th, A
CT,
199
1-96
Sou
rce:
Cau
ses
of d
eath
Aus
tralia
199
1-96
. ABS
Cat
alog
ue N
o. 3
303.
0
05101520253035
1991
1992
1993
1994
1995
1996
Proportion of deaths
Can
cer
Hea
rt di
seas
e
Cer
ebro
vasc
ular
Acc
iden
ts, p
oiso
ning
s an
dvi
olen
ceR
espi
rato
ry
Illus
trat
ion
J: P
rinc
ipal
caus
es o
f dea
th, A
CT,
199
1-96
Sour
ce: C
ause
s of
dea
th A
ustra
lia 1
991-
96. A
BS C
atal
ogue
No.
330
3.0
05101520253035
1991
1992
1993
1994
1995
1996
Proportion of deaths
Can
cer
Hea
rt di
seas
e
Cer
ebro
vasc
ular
Acci
dent
s, p
oiso
ning
san
d vio
lenc
e
Res
pira
tory
Illus
trat
ion
L: C
hild
hood
can
cers
(0
to 1
4 ye
ars)
Fem
ales
Leuk
aem
ia
Cen
tral n
ervo
us
syst
emN
euro
blas
tom
a
Wilm
s' tu
mou
r
Lym
phom
as
Sof
t tis
sue
sarc
oma
Bon
e tu
mou
rs
Mel
anom
a
Ret
inob
last
oma
Oth
erM
ales
Leuk
aem
ia
Cen
tral
ner
vous
sy
stem
Lym
phom
as
Neu
robl
asto
ma
Sof
t tis
sue
sarc
oma
Bon
e tu
mou
rs
Ret
inob
last
oma
Wilm
s' tu
mou
r
Mel
anom
a
Oth
er
Illus
trat
ion
L: C
hild
hood
can
cers
(0 to
14
year
s)
Fem
ales
010
2030
4050
Leuk
aem
ia
Cen
tral n
ervo
us s
yste
m
Neu
robl
asto
ma
Wilm
s' tu
mou
r
Lym
phom
as
Soft
tissu
e sa
rcom
a
Bon
e tu
mou
rs
Mel
anom
a
Ret
inob
last
oma
Oth
er
Prop
ortio
n (%
)
Mal
es
010
2030
4050
Leuk
aem
ia
Cen
tral n
ervo
us s
yste
m
Lym
phom
as
Neu
robl
asto
ma
Soft
tissu
e sa
rcom
a
Bone
tum
ours
Ret
inob
last
oma
Wilm
s' tu
mou
r
Mel
anom
a
Oth
er
Prop
ortio
n (%
)
18 Better health graphs – Volume 1 NSW Health
Discussion
We believe this is the first randomised, controlled
trial assessing interventions aimed at increasing readers’
ability to understand statistical information about
population health. In fact, the evidence-base for
graph comprehension and related cognitive processes
in general is largely limited to studies conducted in
laboratory settings with small groups of subjects, usually
university students. We are aware of only one other
study that randomly selected subjects from a defined
population, and it had a response rate of 50%.2
Further, we found only a limited number of randomised,
controlled study designs in the graph literature.2,3,4
Our findings are of benefit from two perspectives.
First, we were able to quantify the proportion of readers
who could extract some typical statistical interpretations
from a sample of graphs used in Australian official
population health publications. Depending on the graph
and the specific interpretation sought, the proportion
of readers able to correctly interpret the graphs ranged
from as few as 13% to as many as 97%. Second, we
were able to quantify the impact on comprehension
levels achieved through the simple changes we applied
to the graphs. This resulted in a maximum three to
four-fold increase in the proportion of readers who
correctly extracted specific information from the graphs.
Titles and labelsWhile recommendations have been made about graph
titles or captions and labels,5,6,7,8,9 there is little evidence
relating to techniques for making their content easily
understood.
The most dramatic result of the study related to
a vertical bar graph showing that Aboriginal people
in a region of Australia had an increased risk of mortality
at every age compared with the general population;
in some age groups the increase was almost ten-fold.
More than 40% of control subjects (60% of those
without university qualifications) were unable to
determine from the graph the simple fact that
Aboriginal people had a higher risk of death.
A combination of interventions that included
a simple title expressing the question that was
answered by the graph and the addition of words on
the vertical axis that directly related to the title, more
than halved the proportion of subjects who did not
grasp this fact.
People working in public health and epidemiology
regard the concept of disease incidence as quite
commonplace. However, we found that less than
60% of all subjects, and less than half of non university-
qualified subjects, could answer a question that required
an understanding that disease incidence refers to the
rate of new cases of disease in a period of time. Simply
changing the label on the incidence rate series from
‘Incidence…’ to ‘New cases (incidence)...’ had
a statistically significant benefit in both university
and non-university qualified subjects.
FootnotesTo our knowledge, there is no literature on whether
graph readers understand statistical concepts used
in graphs, despite some recommendations being
available.7,9 Two statistical techniques and concepts
occur frequently in population health graphs:
age standardisation and confidence intervals.
We hypothesised that interpretive tasks requiring
an understanding of these concepts would be difficult
for people without specialist knowledge. This was borne
out, with the effect of age standardisation being
understood by only 23% and 44% of non university-
qualified and university qualified subjects respectively.
For a task requiring the interpretation of overlapping
confidence limits, the proportions were 16% and 40%
respectively. We further hypothesised that a footnote
providing a simple, practical explanation of the concepts
and their interpretation, could improve the level of
understanding, and this was also borne out, with
improvements of up to 2.5-fold in one of the tasks
among non-university qualified subjects.
A footnote explaining acronyms that would not be
known to a general audience increased the correct
response to an interpretation task by between two
and three-fold depending on level of education.
NSW Health Better health graphs – Volume 1 19
However, not all footnotes are successful. The
explanatory footnote that we added to a stacked layer
graph (which differs from other graph types because
values for the component categories cannot be read
directly from the axis) had no benefit for any of the
interpretative tasks we investigated and in fact had a
detrimental effect on one task among non-university
educated subjects. We speculate that this particular
footnote confused rather than assisted many readers.
Volume of informationReducing unnecessary information in graphs should
improve reader performance,10,11,12 but by how much?
We tried two interventions aimed at reducing the
volume of information to be interpreted.
First, we reduced the number of categories for which
results were presented in the stacked layer graph.
This did not offer a benefit for the interpretations
we investigated.
Second, we completely removed an independent
(categorisation) variable from a vertical bar graph
that originally presented results for a quantity against
three independent variables within the one graph.
Without the intervention, the graph was reasonably
well understood with the lowest proportion of correct
answers being 72% among non university-qualified
subjects for a task requiring the estimated total quantity
represented by one of the bars. Despite this, the
intervention raised comprehension by 20% even
among university-educated subjects.
Graph typesWe investigated the relative value of line and bar
graphs for displaying information that is plotted against
a categorical x axis that represents a numeric quantity,
such as year or age. A line graph and a grouped bar
graph of multiple disease trends by year performed
equally well for point-reading tasks, but the line graph
produced a marginal improvement in trend judgement
in subjects without a university education. This is as
expected; bar graphs encourage discrete rather than
trend-based comparisons,13 although bar graphs
have been found to be versatile.14,15
The ‘population pyramid’ is a popular choice for
representing the age distribution by sex of a population.
It is in fact a vertically oriented side-by-side bar graph.
It can however, also be represented as a horizontal
format line graph with two series, each series showing
the population size by age for each sex. Among,
surprisingly, university-educated subjects only, the
line graph improved broad comparison of the shape
of the population distribution by sex.
Dot graphs have been proposed as an improvement
on bar graphs.16 We found that a bar graph with 95%
confidence intervals clearly out-performed dot graphs
with 95% confidence intervals (sometimes called ‘hi-lo-
close” graphs), particularly among those without
university qualifications. For another type of dot graph
that had each dot connected by a dashed line to the x
axis, but no confidence intervals, a horizontal bar graph
performed equally well, and even showed a marginal
improvement for those without a university education.
Given that bar graphs are probably more familiar to
general readers and given their ready availability in
common statistical software products, we would
recommend the use of bar graphs over
dot-based graphs.
The stacked layer graph worked well for some tasks
requiring interpretation of trend and broad comparisons,
as expected,7 but worked poorly for a task requiring the
estimation of a difference between the absolute rate at
two points along a layer. This highlights the unsuitability
of these graphs for communicating absolute levels of a
quantity because point estimates for a single category
cannot be read directly from the axis.
Pie charts are often derided because their non-linear
format inhibits precise estimation of statistical
quantities.17,18 However, they do provide a visual
representation of how each category contributes to the
whole.7 This is not easily achieved with other graph
styles. The difficulty of estimating specific quantities or
judging subtle differences from pie charts was borne out
in this study. For simple quantitative tasks such as
identifying minimum and maximum categories or
making comparisons where the differences were distinct,
the pie chart performed as well as a bar chart. If an
important aim is to visually represent how each category
contributes to the whole, then a useful recommendation
would be to use pie charts but ensure the actual
quantities are labelled on each segment of the pie chart.
Discussion
20 Better health graphs – Volume 1 NSW Health
Scales and axesSeveral of our graphs explored the consequences
of using differing scales in adjacent graphs. Many
respondents, particularly those without university
qualifications, appeared to answer questions based on
visual relativities rather than from studying the labels on
the axes. For tasks comparing the relative magnitude of
quantities between the two graphs, a matching scale
range on each graph greatly improved comprehension.
If comparisons between adjacent graphs are important
then the same axis range should be used to avoid
confusion. This is consistent with Kosslyn’s
recommendation,7 and should serve as a qualification
of Cleveland’s recommendation that data should fill the
graph space.6 If such comparisons are not important,
then the two graphs should be presented with a clear
visual separation.
We found strong evidence for ensuring that higher
values of the quantity presented on the graph be in the
upward direction, even if this means the numerical labels
are decreasing in the upward direction. This situation
can arise when the risk of experiencing a disease is
expressed as ‘one in x’, and x is the quantity graphed,
because, for example, a one in 20 risk is larger than a
one in 50 risk. Although this finding may be culturally-
specific, it would be reasonable to assume that for a
horizontally oriented graph, the left to right direction
should represent increasing values.
Limitations of the studySeveral issues need to be borne in mind when
considering the findings of our study.
Despite the randomised design, there were differences
between the control and intervention groups in terms
of self-rated visual ability and frequency of graph use.
Intervention subjects were somewhat more likely to rate
themselves as frequent graph users than control subjects
and were more likely to rate themselves as having good
visual ability. However, the observed differences may
reflect the fact that many of the intervention graphs
were more easily understood than the control graphs.
These questions were asked at the end of the
questionnaire, and intervention subjects may have
felt more comfortable rating themselves more highly
on these characteristics.
The results we obtained would be an overestimate of
levels of comprehension that would be achieved in the
general population. People working in public health and
policy-related areas represented approximately one-fifth
of respondents. These employees would be most likely
to require information on population health statistics
for their work. Many other people in the health system
would have a professional understanding of health and
medicine. Two-thirds of respondents in our study had
university qualifications, compared with approximately
one-fifth of persons aged 25-64 in Australia.19
The graphs we used were taken out of the context of
their original report and we recognise that much of the
explanatory information required to understand the
graph may have been contained in the surrounding text.
Nevertheless, if readers unfamiliar with the subject are
required to hunt for explanatory information, they may
weary of obtaining knowledge about population health.
Publishers of scientific journals often require graphs to
be able to ‘stand alone’, and we support this objective,
but would add that for documents that are intended for
a public audience, the graphs should stand-alone for a
broad sector of the reading population.
Finally, because in some cases we made more than
one change to the intervention graph, we could not
completely isolate the impacts of each of the changes
made. However, we aimed to minimise this difficulty
by making the questions as specific as possible to the
anticipated effects of each of the changes we made.
This approach balanced respondent burden with the
need to test the effects of a number of changes.
Discussion
NSW Health Better health graphs – Volume 1 21
Recommendations
Use plain, non-technicallanguage in the graph title and graph components Techniques that could be considered include:
■ Express the graph title as a simple question
that is answered by the graph.
■ Express technical terms in non-technical terms
followed by the technical term in parentheses.
■ Replace numeric axis labels with descriptive text
that explicitly states the meaning of the quantities
they represent.
■ Explain complex concepts in a simply worded
footnote.
■ Don’t use acronyms unless their meaning
is clearly labelled in close proximity to the graph.
Use the minimum number of sub-categories (independentvariables) necessaryThe graph examples we examined used a variety of
techniques to delineate the quantities expressed for
different population groups. The techniques included
plotting the equivalent graphs as a pair for males and
females, grouping graph bars along the x axis according
to sex or disease category, and/or dividing bars into
segments according to some sub-categorisation.
While these techniques increase the volume of
information that can be communicated, they also
increase the visual complexity of the information.
For example, dividing bars into segments, mean that
the quantity expressed by the length of the segment
cannot be read directly from the y axis, and because
the reference position of the segments varies from one
bar to the next, comparison of length is hindered.
If the difference between two quantities is more
important than the quantities themselves, consider
plotting a graph of the differences, rather than the
two individual quantities.
Use conventional line orbar graphs where possibleOften simpler graph styles communicate as well as,
or better than, more complex or less common designs.
Graphs that can be simplified include ‘population
pyramids’, dot graphs with confidence limits
(‘hi-lo-close’) graphs and dot graphs which
connect the dots to the x axis.
Recognise that theinterpretation of confidenceintervals and age standardisationrequires technical knowledgeConsider methods of simplifying the interpretation
of these concepts. Simple footnotes help dramatically,
but not completely.
Assist readers to interpret ratiosUsing plain, non-technical titles and labels as described
above, assist readers to recognise that a ratio represents
the number of times bigger the numerator quantity is
than the denominator quantity. This can apply to rate
ratios or relative risks, for example.
Ensure that quantities (not labels) increase from thebottom to the top or left to rightThis is a problem when graphing disease risk expressed
as ‘One in x’, for example, where a higher value of x
means a lower risk. The graph should be drawn so that
risk increases from the bottom to top or left to right,
depending on the orientation of the graph. This means
the numeric labels on the risk axis will increase in
the opposite direction to risk, but this will ensure that
readers will interpret relative changes within the graph
in the correct direction.
22 Better health graphs – Volume 1 NSW Health
If using pie charts, label thequantity represented by eachsegment on or near the segmentPie charts do have limitations for comparing relative
magnitudes, but are useful for conveying part-to-whole
relationships. Although not tested in our study, it is likely
that the limitations can be overcome by labelling the
quantities represented by each segment on the graph
itself. If the part-to-whole relationship is not important,
a bar graph will serve just as well.
If presenting graphs in pairs, ensure the axes have the same rangesGraphs that are presented in pairs or groups imply
that they have a relationship. Visual comparisons will
take precedence over the details of axis labels, so ensure
that visual impressions are meaningful by using the same
axis ranges.
Use line graphs to represent trendinformation rather than bar graphs For some readers, a line graph performs better than a
bar graph for assessing trends, without affecting other
tasks. This applies to graphs where the x axis gives the
opportunity to assess trends over time, age or some
other continuous or ordinal variable.
Recommendations
NSW Health Better health graphs – Volume 1 23
Conclusion
Our study provided new evidence to support a range of
recommendations about how to improve the design of
population health graphs. These provide a clear
opportunity to improve delivery of public health
messages through graphs to a wider sector of the
population. Fortunately, this can be achieved through
greater simplicity rather than greater complexity.
However, it is clear that, regardless of graph design,
concepts such as age standardisation and confidence
intervals were not understood by the majority of
subjects, regardless of their level of education. This is a
vexed problem, because these concepts are crucial to
accurate interpretation of statistical information in
population health and epidemiology. There remains,
therefore, an opportunity for inventive thought on
delivering the messages implied by these manipulations
without increasing the complexity of the graph.
24 Better health graphs – Volume 1 NSW Health
References
1 Australian Institute of Health and Welfare and
the National Public Health Information Working
Group 1999, National Public Health Information
Development Plan, Canberra: Australian Institute
of Health and Welfare.
2 Henry GT 1993, Using graphical displays for
evaluation, Evaluation Review; 17(1): 60-78.
3 Meyer J, Shinar D 1992, Estimating Correlations from
Scatterplots, Human Factors; 34(3): 335-349.
4 Lee ML, MacLachlan J 1986, The Effects of
3D Imagery on Managerial Data Interpretation,
MIS Quarterly; September: 257-268.
5 Schmidt CF 1983, Statistical Graphs Design Principles
and Practices, New York: John Wiley and Sons.
6 Cleveland WS 1994, The Elements of Graphing Data,
Murray Hill NJ: AT and T Bell Laboratories.
7 Kosslyn SM 1994, Elements of Graph Design,
New York: WH Freeman and Company.
8 Gillan DJ 1994, A Componential Model of
Human Interaction with Graphs: I Linear Regression
Modelling, Human Factors; 36(3): 419-440.
9 Gillan DJ, Wickens CD, Hollands JC, Carswell CM
1998, Guidelines for Presenting Quantitative Data
in HFES Publications, Human Factors: 40(1): 28-41.
10 Schutz HG 1961, An Evaluation of Methods for
Presentation of Graphic Multiple Trends, Human
Factors; 3: 108-119.
11 Casali JG, Gaylin KB 1988, Selected Graph
Design Variables in Four Interpretation Tasks:
A Microcomputer-Based Pilot Study, Behaviour
and Information Technology; 7(1): 31-49.
12 Kosslyn SM 1989, Understanding Charts and Graphs,
Applied Cognitive Psychology; 3: 185-226.
13 Zacs J, Tversky, B 1999, Bars and lines: A study of
graphic communication, Memory and Cognition;
27(6): 1073-1079.
14 Shah P, Mayer RE, Hegarty M 1999, Graphs as Aids
to Knowledge Construction: Signaling Techniques for
Guiding the Process of Graph Comprehension,
Journal of Educational Psychology; 91(4): 690-702.
15 Carswell CM, Ramzy C 1997, Graphing Small Data
Sets: Should We Bother? Behaviour and Information
Technology; 16: 61-71.
16 Cleveland WS, McGill R 1984, Graphical perception:
theory, experimentation and application to the
development of graphical methods, Journal of
American Statistical Association; 79: 531-554.
17 Tufte ER 1983, The Visual Display of Quantitative
Information, Cheshire CT: Graphics Press, 1983.
18 Cleveland WS, McGill R 1985, Graphical Perception
and Graphical Methods for Analyzing Scientific Data,
Science; 229: 828-833.
19 Australian Bureau of Statistics 2005,
Australian Social Trends 2005 (Catalogue 4102.0).
Canberra: Australian Bureau of Statistics.
NSW Health Better health graphs – Volume 1 25
Appendix 1.The control booklet of graphs
- Page C 1 -
Booklet of graphs
Introduction
The NSW Department of Health has asked the Hunter Valley Research Foundation (HVRF) to conduct a study to determine guidelines for designing informative and useful graphs. Graphs are an important tool for communicating health related information. Your participation in completing the accompanying questionnaire will be greatly appreciated.
Instructions
This booklet contains examples of different health graphs. You do not need to know the topic of the graph. In fact we ask you to answer all questions from the information in each graph, not from any knowledge you may have on the subject.
Please write all answers in the questionnaire booklet supplied.
Any questions? Should you have any questions regarding this research, feel free to contact Andrew Searles at the HVRF on (02) 4969 4566 (extension 525). Alternatively, call David Muscatello of NSW Health on (02) 9391 9408.
Bet
ter
heal
th g
raph
s pr
ojec
t
- Page C 2 -
Illustration A
Illustration B
Source: Cancer in Australia 1997, AIHW & AACR 2000.
Illustration A: Trends in age-standardised incidence and mortality rates for all cancers(excluding non-melanocytic skin cancers), Australia, 1983-1998
0
100
200
300
400
500
600
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Incidence - males
Mortality - females
Mortality - males
Incidence - females
New
cas
es a
nd
dea
ths
per
100
,000
po
pu
lati
on
Illustration B: Incident DALY Rates per 1,000 Population by Mental Disorder, Age and Sex, Victoria 1996
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80Age
DA
LYs
per
1,00
0 po
pula
tion Other
Substance use disorders
Schizophrenia
Anxiety disorders
Depression
Males
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80Age
DA
LYs
per
1,00
0 po
pula
tion
Other
Substance use disorders
Schizophrenia
Anxiety disorders
Depression
Females
- Page C 3 -
Illustration C
Illustration D
Illustration C: The Burden of ChronicRespiratory Disease by Conditionand Sex, Victoria 1996
0 4,000 8,000 12,000 16,000DALYs
YLL
YLD
YLD
YLLFemales
Males
COPD
Asthma
Other
Illustration D: Rates of YLLs by RuralityStatus, Sex and Major Causes of Death
0
20
40
60
80
Metro Ruraltowns
Otherrural
Metro Ruraltowns
Otherrural
Rat
e of
YLL
s pe
r 1,0
00 p
opul
atio
n
Cardiovascular Cancer
Injuries Other causes
Males Females
- Page C 4 -
Illustration E
Illustration E: Estimated resident population by age, sex and Health Service District, 1999and difference in age structure between Health Service District population and Queensland population
Central zone Male Female Queensland
60000 40000 20000 0 20000 40000 60000
0-4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85+
Age
gro
up
Number of persons200000 100000 0 100000 200000
0-4
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85+
Age
gro
up
Number of persons
- Page C 5 -
Illustration F
Illustration G Illustration G: Northern Territory (NT) Aboriginal: Australian death rate ratios 1991 to 1995
Note: Ratio of NT Aboriginal to Australian death rates for all causes
by five-year age groups
Source: Dempsey & Condon 1999
1
2
3
4
5
6
7
8
9
10
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+
Age groups (years)
Rat
e ra
tio
Males
Females
- Page C 6 -
Illustration H
Illustration I
Illustration I: Antibodies to human immunodeficiency virus (HIV) and hepatitis C virus (HCV) by injecting history, clients of needle and syringe programs, NSW 1995 to 1998
HIV HCV
0
1
2
3
4
5
1995 1996 1997 1998
Year
Injecting <3 years
Injecting 3+ years
Per cent antibody positive
0
20
40
60
80
100
1995 1996 1997 1998
Year
Injecting <3 years
Injecting 3+ years
Per cent antibody positive
Illustration H: Lifetime risk for lung cancer to age 74 years
5
15
25
35
45
55
65
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
Males - Western Australia Females - Western AustraliaOne in
- Page C 7 -
Illustration J
Illustration K
Illustration J: Principal causes of death, ACT, 1991-96
Source: Causes of death Australia 1991-96 . ABS Catalogue No. 3303.0
0
5
10
15
20
25
30
35
1991 1992 1993 1994 1995 1996
Pro
po
rtio
n o
f d
eath
s
Cancer
Heart disease
Cerebrovascular
Accidents, poisonings andviolenceRespiratory
- Page C 8 -
Illustration L
Illustration L: Childhood cancers (0 to 14 years)
Females
Leukaemia
Central nervous system
Neuroblastoma
Wilms' tumour
Lymphomas
Soft tissue sarcoma
Bone tumours
Melanoma
Retinoblastoma
Other
Males
Leukaemia
Central nervous system
Lymphomas
Neuroblastoma
Soft tissue sarcoma
Bone tumours
Retinoblastoma
Wilms' tumour
Melanoma
Other
26 Better health graphs – Volume 1 NSW Health
Appendix 2. The intervention booklet of graphs
Page T 1
Booklet of graphs
Introduction
The NSW Department of Health has asked the Hunter Valley Research Foundation (HVRF) to conduct a study to determine guidelines for designing informative and useful graphs. Graphs are an important tool for communicating health related information. Your participation in completing the accompanying questionnaire will be greatly appreciated.
Instructions
This booklet contains examples of different health graphs. You do not need to know the topic of the graph. In fact we ask you to answer all questions from the information in each graph, not from any knowledge you may have on the subject.
Please write all answers in the questionnaire booklet supplied.
Any questions? Should you have any questions regarding this research, feel free to contact Andrew Searles at the HVRF on (02) 4969 4566 (extension 525). Alternatively, call David Muscatello of NSW Health on (02) 9391 9408.
Bet
ter
heal
th g
raph
s pr
ojec
t
Page T 2
Illustration A
Illustration B
Note: age-standardised rates allow comparisons over years and between males and females.Different age-standardised rates are not due to differences in the relative proportions of older oryounger people in each year or sex.Source: Cancer in Australia 1997. AIHW & AACR 2000.
Illustration A: Trends in age-standardised incidence and death rates for all cancers(excluding non-melanocytic skin cancers), Australia, 1983-1998
0
100
200
300
400
500
600
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
New cases (incidence) - males
Deaths - females
Deaths - males
New cases (incidence) - females
New
cas
es a
nd
dea
ths
per
100
,000
po
pu
lati
on
Illustration B: Incident DALY Rates per 1,000 Population by Mental Disorder, Age and Sex, Victoria 1996
Note: The thickness of the shaded layer = DALYs per 1,000 population for that disorder
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80Age
DA
LYs
per
1,00
0 po
pula
tion Other
Anxiety disorders
Depression
Males
0
10
20
30
40
50
0 10 20 30 40 50 60 70 80Age
DA
LYs
per
1,00
0 po
pula
tion
Other
Anxiety disorders
Depression
Females
Page T 3
Illustration C
Illustration D
Illustration D: Rates of YLLs by Rurality, Status and Sex
0
20
40
60
80
Metro Ruraltowns
Otherrural
Metro Ruraltowns
Otherrural
Rat
e of
YLL
s pe
r 1,0
00 p
opul
atio
n Males Females
Illustration C: The Burden of ChronicRespiratory Disease by Conditionand Sex, Victoria 1996
16,000 12,000 8,000 4,000 0 4,000 8,000 12,000 16,000
Other
Asthma
COPD
DALYs
MalesFemales
YLL YLD
YLL = Years of Life Lost: summarises the total years of life lost from all people that die prematurely of the disease.
YLD = Years Lived with Disability: summarises the total years of healthy life lost due to disability in people living with the disease.
DALY = Disability Adjusted Life Years: total burden = the sum of YLL and YLD: lost years due to both death and disability.
Page T 4
Illustration E
Illustration E: Estimated resident population by age, sex and Health Service District, 1999and difference in age structure between Health Service District population and Queensland population
Central zone
Queensland
0
20000
40000
60000
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+
Age group
Num
ber
of p
erso
ns
Male
Female
0
100000
200000
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+
Age group
Num
ber
of p
erso
ns
Male
Female
Page T 5
Illustration F
Illustration G
Illustration G: Between 1991 and 1995, how many times more likely to die wasa Northern Territory (NT) Aboriginal person compared with all Australians for each sex and age group?
1
2
3
4
5
6
7
8
9
10
0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+
Age groups (years)
Males
Females
Equally aslikely
Ten timesas likely
Five timesas likely
Source: Dempsey & Condon 1999
Premature births by country of birth of mother,NSW 1994 to 1998
Country of birth
Note: Confidence intervals indicate statistical uncertainty about each value on the graph. Longer intervals mean more uncertainty. When two intervals overlap then there is more uncertainty that the two groups are really different. Births where gestational age was less that 37 weeks were classified as premature births. Infants of at least 400 grams birth weight or at least 20 weeks gestation were included.Source: NSW Midwives Data Collection (HOIST). Epidemiology and Surveillance Branch, NSW Health Department
0 2 4 6 8 10 12
AllUnited States
EgyptPoland
MaltaMalaysia
South AfricaFiji
NetherlandsIndia
GermanyHong Kong
GreecePhilippines
LebanonVietnam
ChinaFormer Yugo.
ItalyNew Zealand
United KingdomAustralia
Per cent
Page T 6
Illustration H
Illustration I
Illustration H: Lifetime risk for lung cancer to age 74 years
5
15
25
35
45
55
65
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
Males - Western Australia Females - Western AustraliaOne in
Illustration I: Prevalence of human immunodeficiency virus (HIV) and hepatitis C virus (HCV) infection by injecting history, clients of needle and syringe programs, NSW 1995 to 1998
HIV HCV
0
20
40
60
80
100
1995 1996 1997 1998
Year
Injecting <3 years
Injecting 3+ years
Per cent antibody positive
0
20
40
60
80
100
1995 1996 1997 1998
Year
Injecting <3 years
Injecting 3+ years
Per cent antibody positive
Page T 7
Illustration J
Illustration K
Illustration J: Principal causes of death, ACT, 1991-96
Source: Causes of death Australia 1991-96 . ABS Catalogue No. 3303.0
0
5
10
15
20
25
30
35
1991 1992 1993 1994 1995 1996
Pro
po
rtio
n o
f d
eath
s
Cancer
Heart disease
Cerebrovascular
Accidents, poisoningsand violence
Respiratory
HOSPITAL SEPARATIONS, Cause of Injury or Poisoning(a)---1998-99
0 5 10 15 20 25 30
Other
Intentional self harm
Unspecified accidental exposures
Complications of medical & surgical care
Transport accidents
Accidental falls
Exposure to inanimate mechanical forces (b)
Assault
Males identified as Indigenous
Females identified as Indigenous
(a) Data are from public and most private hospitals. Cause of injury is based on the first reported external cause where the principal diagnosis was 'injury, poisoning and certain other consequences ofexternal causes'.(b) Includes injuries due to accidental contact with machinery or other objects, accidental discharge from firearms, explosions, & exposure to noise.
Source: AIHW National Hospital Morbidity Database.
% of injury or poisoning separations
Page T 8
Illustration L
Illustration L: Childhood cancers (0 to 14 years)
Females
0 10 20 30 40 50
Leukaemia
Central nervous system
Neuroblastoma
Wilms' tumour
Lymphomas
Soft tissue sarcoma
Bone tumours
Melanoma
Retinoblastoma
Other
Proportion (%)
Males
0 10 20 30 40 50
Leukaemia
Central nervous system
Lymphomas
Neuroblastoma
Soft tissue sarcoma
Bone tumours
Retinoblastoma
Wilms' tumour
Melanoma
Other
Proportion (%)
NSW Health Better health graphs – Volume 1
Appendix 3. Questionnaire
27
Page Q 1
Questionnaire +
Introduction
The NSW Department of Health has asked the Hunter Valley Research Foundation (HVRF) to conduct a study to determine guidelines for designing informative and useful graphs. Graph design can determine whether the reader correctly interprets the information contained in the graph. This questionnaire and booklet of graphs is one component of this study. Please note that this questionnaire is not a test. Even very experienced people can have trouble understanding graphs that are not designed properly. Your answers will help us identify what aspects of graphs are hard to understand so that we can develop guidelines to improve published graphs. We appreciate you taking the time to answer these questions even though some might seem difficult.
Why your participation is essential
To make recommendations for the design of a good graph we need to know how people interpret different styles of graph. Even if you are not a frequent graph user, your input is valuable.
How to use this questionnaire
The questionnaire asks questions about the graphs in the bookletof graphs. Please write your answers in the questionnaire. The questionnaire should only take 20 minutes to complete.
Instructions
You may use any tool that you might use in real life to make interpretations. That is, any technique (ruler, pen etc.) that you already use when interpreting a graph in health publications or other media (e.g. newspapers).
Knowledge of the topic in each graph is not a requirement of the study. Answer the questions from the information in each graph, not your knowledge of the subject.
Returning the completed questionnaire
Please post your completed questionnaire in the self addressed, reply paid envelope to: The Researcher The HVRF: Graph Project PO Box 3023 Hamilton DC NSW 2303
Any questions? Should you have any questions regarding this research, feel free to contact Andrew Searles at the HVRF on (02) 4969 4566 (ext 525). Alternatively, call David Muscatello at NSW Health on (02) 9391 9408.
Bet
ter
heal
th g
raph
s pr
ojec
t
Page Q 2
Approximately, how long did you take to answer the questions about illustration A ? Minutes: _______ Seconds: _______
Please record your start time (below) and finish time (at the end of the questionnaire) as we would like to record how much time you required to complete the survey.
Start time: ______________________
Referring to Illustration A in your booklet of graphs
Qa1) Which statement best describes the incidence rate of female cancer in 1997?
Circle the number of your answer:
1 Out of every 100,000 females, there were 330 who were newly diagnosed with cancer
2 Out of every 100,000 females, there were 330 with cancer
3 For every 100,000 females, there were an additional 330 who had cancer
9 Don’t know
Qa2) As the graph uses “age-standardised” data, which of the following statements is the most correct?
Circle the number of your answer:
1 Age-standardisation means differences between the rates of cancer in males and females
could be due to differences in the pattern of ages in the male and female populations
2 Age-standardisation means differences between the rates of cancer in males and females
are not due to differences in the pattern of ages in the male and female populations
3 Age-standardisation means that a single figure represents the rate of new cases of cancer in males and females
9 Don’t know
Page Q 3
Approximately, how long did you take to answer the questions about illustration B ? Minutes: _______ Seconds: _______
Referring to Illustration B in your booklet of graphs
Qb1) What is the (approximate) difference in DALYs per 1,000 population between
female anxiety disorders at age 20 and at age 60?
Please write your answer here: _________________________ 9 Don’t know
Qb2) At age 30, which gender has the higher incident DALY rate for anxiety disorders, males or females?
Circle the number of your answer:
1 Males
2 Females
9 Don’t know
Qb3) Which statement best reflects the trend in male depression?
Circle the number of your answer:
1 Peaks at age 20, drops to age 30, remains stable to age 40, then declines
2 Rises to a peak at age 50, then declines
3 Fluctuates throughout the age groups
9 Don’t know
Qb4) Compared with females, males are less likely to develop mental disorders at 60 or more years of age? Circle the number of your answer:
1 True
2 False
9 Don’t know
Page Q 4
Approximately, how long did you take to answer the questions about illustration C ? Minutes: _______ Seconds: _______
Referring to Illustration C in your booklet of graphs
Qc1) For male COPD (chronic obstructive pulmonary disease), which is the larger
value, YLL or YLD?
Circle the number of your answer:
1 YLL
2 YLD
9 Don’t know
Qc2) For males, which respiratory disease caused the highest disability burden?
Circle the number of your answer:
1 Other respiratory
2 Asthma
3 COPD (chronic obstructive pulmonary disease)
4 Cannot be answered from the graph
9 Don’t know
Qc3) For asthma, do males or females have the greatest burden from deaths (YLL)?
Circle the number of your answer:
1 Males
2 Females
9 Don’t know
Qc4) Which disease has the greatest overall burden (DALYs) for females?
Circle the number of your answer:
1 COPD (chronic obstructive pulmonary disease)
2 Asthma
3 Other respiratory
9 Don’t know
Page Q 5
Approximately, how long did you take to answer the questions about illustration D ? Minutes: _______ Seconds: _______
Approximately, how long did you take to answer the questions about illustration E? Minutes: _______ Seconds: _______
Referring to Illustration D in your booklet of graphs
Qd1) What was the approximate total rate of YLLs for females in rural towns?
Circle the number of your answer:
1 60
2 65
3 80
9 Don’t know
Qd2) Correct or incorrect?
Circle the number of your answer:
Statement A: Rural areas had higher rates of YLLs compared with metropolitan areas?
1 Correct
2 Incorrect
9 Don’t know
Statement B: Males had lower rates of YLLs than females, regardless of where they lived
1 Correct
2 Incorrect
9 Don’t know
Referring to Illustration E in your booklet of graphs
Qe1) In Queensland, males aged 19 or less outnumber females aged 19 or less.
Circle the number of your answer:
1 True
2 False
9 Don’t know
Qe2) Which is the most accurate statement for this illustration?
Circle the number of your answer:
1 The two graphs show that Central Zone has more people than Queensland
2 The two graphs show that Central Zone has less people than Queensland
3 Cannot answer from the graph
9 Don’t know
Qe3) Which is the most accurate statement for this illustration?
Circle the number of your answer:
1 Both Central Zone and Queensland have more younger people (aged 19 or less) than older people (aged 60+)
2 Both Central Zone and Queensland have more older people (aged 60+) than younger people (aged 19 or less)
3 Cannot answer from the graph
9 Don’t know
Page Q 6
Approximately, how long did you take to answer the questions about illustration F ? Minutes: _______ Seconds: _______
Referring to Illustration F in your booklet of graphs
Qf1) Can we be certain that mothers born in Greece and those born in the Philippines
really differed from each other in their chance of having a premature birth?
Circle the number of your answer:
1 Yes
2 No
9 Don’t know
Qf2) Comparing mothers born in the Philippines with those born in Lebanon:
Circle the number of your answer:
1 Mothers born in the Philippines had a higher proportion of premature births
2 Mothers born in the Philippines had a lower proportion of premature births
9 Don’t know
Qf3) Mothers born in Lebanon had a lower proportion of premature births than mothers born
in Australia?
1 True
2 False
9 Don’t know
Page Q 7
Approximately, how long did you take to answer the questions about illustration G? Minutes: _______ Seconds: _______
Referring to Illustration G in your booklet of graphs
Qg1) This graphs shows that, compared with most Australians ….
Circle the number of your answer:
1 NT Aboriginal people have a higher risk of death
2 NT Aboriginal people have a similar risk of death
3 NT Aboriginal people have a lower risk of death
9 Don’t know
Qg2) For the age group 70-74 how many times greater is the risk of a NT Aboriginal women
dying compared with all Australian women in the same age group?
Circle the number of your answer:
1 4.0
2 4.5
3 5.0
9 Don’t know
Qg3) For the age group 45-49 the approximate value for females is 7. Which of the following
best describes the meaning of this result for people aged 45-49:
Circle the number of your answer:
1 Compared with Aboriginal males, Aboriginal females are seven times more likely to die
2 The risk of a Northern Territory Aboriginal female dying is seven times as high as an Australian female overall
3 Seven Northern Territory Aboriginal males die for every 1 Aboriginal female
9 Don’t know
Page Q 8
Approximately, how long did you take to answer the questions about illustration H? Minutes: _______ Seconds: _______
Approximately, how long did you take to answer the questions about illustration I? Minutes: _______ Seconds: _______
Referring to Illustration H in your booklet of graphs
Qh1) In 1996, would a male or a female have been more likely to develop lung cancer in
Western Australia?
Circle the number of your answer:
1 A female
2 A male
9 Don’t know
Qh2) What is the direction of male lifetime risk for lung cancer?
Circle the number of your answer:
1 Slightly increasing risk
2 Slightly decreasing risk
3 Steady (no trend)
9 Don’t know
Qh3) In 1993, what was the lifetime risk for females?
Please write your answer here: One in _________________________ 9 Don’t know
Referring to Illustration I in your booklet of graphs
Qi1) In 1997, approximately what proportion of clients who had been injecting for
less than 3 years had HCV infection?
Please write your answer here: _________________________ 9 Don’t know
Qi2) Which group had the lower prevalence of HCV infection between 1995 and 1998?
Circle the number of your answer:
1 Those injecting for 3 or more years
2 Those injecting less than 3 years
3 Both have the same prevalence
9 Don’t know
Qi3) Which infection was more prevalent among injecting drug users in 1996?
Circle the number of your answer:
1 HIV
2 HCV
3 HIV and HCV were about the same
9 Don’t know
Qi4) In 1997, the gap in prevalence between short and long term injectors was approximately the
same for HIV and HCV?
Circle the number of your answer:
1 True
2 False
9 Don’t know
Page Q 9
Approximately, how long did you take to answer the questions about illustration J? Minutes: _______ Seconds: _______
Approximately, how long did you take to answer the questions about illustration K? Minutes: _______ Seconds: _______
Referring to Illustration J in your booklet of graphs
Qj1) Approximately what proportion of deaths were due to cancer in 1996?
Please write your answer here: _________________________ 9 Don’t know
Qj2) In 1995, the lowest proportion of deaths was associated with ….
Circle the number of your answer:
1 Cancer
2 Heart disease
3 Accidents, poisonings and violence
4 Respiratory
9 Don’t know
Qj3) Which cause of death shows the most increasing trend between 1991 and 1996?
Circle the number of your answer:
1 Cancer
2 Heart disease
3 Cerebrovascular
4 Respiratory
9 Don’t know
Referring to Illustration K in your booklet of graphs
Qk1) Does intentional self harm account for a greater proportion of hospital
separations for indigenous males or indigenous females?
Circle the number of your answer:
1 Males
2 Females
3 Both are the same
9 Don’t know
Qk2) What is the most common cause of hospital separations for injuries in indigenous males?
Circle the number of your answer:
1 Transport accidents
2 Complications of medical and surgical care
3 Assault
9 Don’t know
Page Q 10
Approximately, how long did you take to answer the questions about illustration L? Minutes: _______ Seconds: _______
Referring to Illustration L in your booklet of graphs
QL1) What is the most common childhood cancer for males?
Please write your answer here: _________________________ 9 Don’t know
QL2) For females, are there more neuroblastomas or central nervous system cancers?
Circle the number of your answer:
1 Central nervous system
2 Neuroblastomas
3 Both are the same
9 Don’t know
QL3) Do males or females have a greater proportion of central nervous system cancers?
Circle the number of your answer:
1 Males
2 Females
3 Both are the same
9 Don’t know
QL4) What is the least common cause of cancer in females?
Circle the number of your answer:
1 Melanoma
2 Retinoblastoma
3 Bone tumours
9 Don’t know
QL5) Approximately what proportion of childhood cancers for girls does melanoma account for?
Please write your answer here: _________________________ 9 Don’t know
Page Q 11
These questions will help ensure our sample included a range of people
DEM1) In what language would you have felt most comfortable completing this questionnaire? Circle the number of your answer:
1 English 2 Other Please identify your preferred language: _________________________
DEM2) What is the highest level of education you have completed? Circle the number of your answer:
1 Never attended school 2 Primary school only 3 Secondary school (Up to year 12 / 6th form / HSC / Leaving Certificate) 4 TAFE or equivalent technical qualification 5 University or CAE 6 Postgraduate studies 8 Other Please identify: _________________________
DEM3) How frequently do you use graphs in your daily activities? (This includes graphs that you might interpret or create yourself. They might be for work or non-work activities such as reading a newspaper or for your studies). Circle the number of your answer:
1 Never 2 Rarely (i.e. less than a few times a year)
3 Occasionally (i.e. a few times a year to less than once a month) 4 Often (i.e. at least once a month)
DEM4) How would you rate your visual ability to see the detail in the graphs in this study? (This refers to your ability to see the labels and diagrammatic detail either unaided, or if you have corrected vision, with eye glasses, contact lenses or other aides). Circle the number of your answer:
1 Good (could read all labels and notes on the sample graphs – even when the font size was small) 2 Average (could read labels and notes on the sample graphs – with slight difficulty) 3 Poor (had difficulty reading labels and notes on the sample graphs)
DEM5) What is your age category?
1 Under 24 2 24 to 34 3 35 to 44 4 45 to 54 5 55 to 64 6 65 and over
Continued over the page
Page Q 12
These questions will help ensure our sample included a range of people
DEM6) And your gender? Circle the number of your answer:
1 Male 2 Female
DEM7) How would you describe your current work position? Please write your occupation in the space below
_____________________________________________________________________________
DEM8) If you have completed the questionnaire in one sitting (and provided a start time on page 2), please answer 6a. If the questionnaire was completed over multiple sittings please answer 6b.
6a) Finish time: ______________________
6b) Approximately how much time in total did you need to complete this questionnaire? Please write the length of time in minutes here: _________________________
Thank you for your help! Please return your completed questionnaire in the reply paid envelope.
SHPN (HSP) 060048