Displaying data
and interpreting results
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Overview
 Summarizing data
 Displaying results
 Interpretation of results
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing data
 Summary statistics that are used for summarizing STEPS
data:
– Prevalence
– Mean
– Median
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing data - prevalence
 Prevalence = proportion of a population
– e.g., % of people who smoke daily
 We use WHO-recognized cut-points for continuous variables.
– e.g. % of people who are overweight (have a BMI of 25 or higher)
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing data – mean and median
 Mean and median are measures of central tendency.
 Mean provides the average of a set of values, and is used for
normally distributed variables (such as BMI).
 Median provides the mid-point of a set of values, and is used
for non-normally distributed variables (such as time spent in
physical activity).
– The median enables you to say: 50% of the respondents had a
response more or less than [median value].
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing data – estimates and true values
 All sample-based surveys lack some amount of precision due to non-sampling
and sampling error.
 Prevalence, mean or median from a sample are estimated values.
 Standard error (SE) is usually calculated to show the amount of uncertainty, or
error, in an estimated value. It takes into consideration the sample size &
distribution (standard deviation) of your sample.
 Confidence intervals (CI) are calculated from the SE and are a more userfriendly way to show to show the amount of precision in your results.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing Data - estimates and true values, cont.
 Epi Info and most other stats programs automatically calculate the
Standard Error and the Confidence Intervals for you.
 Keep in mind …
– The more precise your results, the more confidence you can have in them!
– Thus, the shorter the interval the better!
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Confidence Intervals (CI)
 Typically calculated by multiplying the standard error (SE) by 1.96:
CI = SE * 1.96
 Expressed as a range around the estimate:
42% (40% - 44%) or 42% (+2%)
 The interval shows the range of the estimates that would be obtained were all
possible samples used.
 A 95% CI suggests that if 100 samples were drawn, the estimate obtained from
each (a mean or prevalence value) would fall within that interval 95 of 100 times.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summarizing data – median and IQR
 When a sample median is reported, an inter-quartile range (IQR)
[25%; 75%] should be reported too.
 Example:
median time (in mins) spent in physical activity per week = 60, IQR
[20; 180]
→ Half of the sample engages in 60 minutes of PA per week or
less, ¼ of it in 20 minutes or less, ¾ of it in 180 minutes or less
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Displaying data
 Visual methods can make the point much stronger than simply
describing the data.
 Appropriate use of tables and graphs can enhance the
message you are delivering.
 But they have potential to confuse or convey wrong messages.
 Always be sure to explain your table or graph in the text of
your report.
– For every table or graph, there should be a reference to it in the text.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Tables
 Good for when exact numbers need to be presented.
 Because your STEPS report will be the reference for results
from your STEPS survey, it will have a lot of tables!
 Best Practice:
– Clear table title and column / row headings
– Minimal use of grid lines
– Leave enough space so columns/rows are easy to read
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Tables, example
 Example of a typical "prevalence table" from the STEPS Data Book:
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Graphs
 For STEPS data, bar graphs and pie charts are most often
used.
 Bar graphs show quantities represented by horizontal or
vertical bars and are useful for displaying:
– Several categories of results at once (e.g. males vs. females)
 Pie charts show proportions in relation to a whole, with each
wedge representing a percentage of the total and are useful
for displaying:
– Parts of a whole in percentages
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Graphs: Best Practice
 Emphasize one idea at a time in a figure.
 Pay careful attention to the scaling of the graph.
 Provide a title, units and labels: the graph or table should be
self-explanatory!
 If possible, mention the total sample size of the data set for
which the graph or chart is made.
 Be sparing and consistent with use of colour, fonts and
"enhancements".
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Watch the scale!
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Title and Label!
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Title and Label!
Smoking status
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Show the sample size
Include
error bars
where
possible.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Keep it simple
"Exploding" pie charts and other 3D graphics are not
necessary. They distract from your data and may even
make your graph harder to read.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Pie Charts
 Pie charts should be used sparingly and are generally more useful
when there are marked differences between the segments:
Value labels make them easier to
read.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Bar Chart example
Bar Charts can
also be used to
show parts of a
whole.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Bar Chart example
Remember that
in addition to
males vs.
females, other
sub-group
comparisons,
such as age
group, may be
interesting to
graph.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Interpretation of results
 How to test for subgroup differences?
 For which population are results representative?
 What could have influenced results?
 How to interpret results in a context?
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Interpretation of results, cont.
 Look for overall pattern, then deviation from the pattern.
 Look for extreme values (outliers) and gaps.
 Locate center and spread of distribution.
 Compare graphs with same scale - look for max, min.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Confidence Intervals Method for Testing for Subgroup Differences
 Possible subgroups
– Males and Females
– Age groups (25 – 34, 35 – 44, etc)
– Countries, Regions, Cities, etc.
 Question – Are they different?
 Answer – Look at the 95% confidence intervals (CI’s)
– If the CI’s overlap – they are not statistically different
– If the CI’s do not overlap – they are statistically different
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Are they different?
 Participants who are overweight
Males - 16.4% (14.5-18.2)
Females – 12.6% (11.4-13.7)
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Are they different?
 Participants who are overweight
Males - 16.4% (14.5-18.2)
Females – 12.6% (11.4-13.7)
11
12
13
14
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
15
16
17
18
Are they different?
 Participants who have raised blood pressure
Males - 8.4 (5.4-11.4)
Females – 11.2 (9.2-13.2)
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Are they different?
 Participants who have raised blood pressure
Males - 8.4 (5.4-11.4)
Females – 11.2 (9.2-13.2)
5
7
9
11
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
13
15
Representativeness of results
 Questions that should be discussed:
– What population are these results representative for (ages, regional
coverage)?
– How was the response-rate?
– Can any information be obtained about persons who did not respond?
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Influence on results
 What could have influenced the results?
 Is it possible that results over- / underestimate true values?
(every study has limitations)
–
–
–
–
Season
Social desirability
Missing values
Methods used (e. g., for sampling, data analysis)
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Results in a context
 If possible, results should be put in a context for interpretation:
– Comparison with previous surveys in the same population
– Comparison with surveys using the same methodology in other
populations
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases
Summary
 Prevalence, mean and median are used as summary statistics for
reporting STEPS data.
 CI should be provided with prevalence and mean values, IQR with
median values.
 The most effective ways of presenting data are tables and graphs.
 Results should be interpreted with regards to accuracy,
representativeness and methods used.
 If possible, results should be compared to other surveys.
Surveillance and Population-based Prevention
Department for Prevention of Noncommunicable Diseases