1. Effective graphs and maps
The visual display of quantitative and geographical information
EPIET/EUPHEM introductory course
Chris Williams, Adapted from slides Y. Hutin

2. Communicating quantitative information
Analogue communication (graphs, maps)
Graphic overview
“hand waving”
Less precise
Digital communication (e.g. Tables)
Precise
Numeric (text also?)
Provides detailed and exact description
7:00 am
Digital and analogical information

3. Table, graph or map? (or none of the above)
Table –you really need to know exact time. Map/picture shows border of Haiti and Dominican Republic- deforestation in former. Graph conveys change over time.

4. Vaccine coverage in Yamoussoukro district, Ivory Coast, 1995
The exact % does not add anything and may be distracting
Digital and analogical information
EPIGEPS course of field epidemiology

5. Graphs: Key areas What is the message?
Note the singular
Choose the right graph for the right data, that demonstrates the message clearly
Style and etiquette
Message:ink ratio
Message- think of linked text in paper. What graph best shows this –imagine seen as flashcard/subliminal.

6. Decide on a message Don’t use a graph if there is nothing to say
Frame the idea to communicate
Identify the graph that matches this idea
Eliminate unnecessary information
If there are two ideas, use two slides
Message is the only thing you would say about the graph if you had to describe it- or better still, what you wanted to show before you made the graph
Framing the message

7. No clear message Symptom surveillance by region
Is there a message? How should this have been presented? Perhaps OK for a regular surveillance report

8. Two messages- too many? Weekly norovirus tests by result
Shows number of tests and % positive. Also juxtaposed bars

9. One message- varying proportion positive
Add graph on tests per week

10. Choosing the right graph

11. Frequency distribution
Histogram
Graphic representation of the frequency distribution of a continuous variable
Rectangles drawn in such a way that their bases lie on a linear scale representing different intervals
Areas are proportional to the frequencies of the values within each of the intervals
Epidemic curve is an example of histogram
Analog information

12. Histogram
Urinary iodine excretion status, 24 N Parganas, West Bengal, India, 2004 80 60 40
Percentage 20
Ringed category is twice as wide as 50-99, in turn twice
0-19.9
> 300
Urinary Iodine Excretion levels (µg/L)

13. Epicurve: gastroenteritis at a music festival
One case one square

14. COLD/FLU CALLS: Daily ‘cold/flu’ calls as a proportion of total calls (7-day moving average) by region( ).
Presentation of time series data

15. Proportions of a total presenting selected characteristics
Breakdown of a total in proportions:
Pie chart
Breakdown of more than one total into proportion:
Stacked bar charts adding to 100%
Analog information

16. Respiratory virus isolates by type: Pie chart
Whole is all positive samples

17. Alignment allows comparing proportions across groups
Cumulated bar chart for the breakdown of many totals in proportions
Estimated and projected proportion of deaths due to non-communicable diseases, India,
100%
90%
80%
Injuries
70%
60%
Communicable
Proportion (%)
50%
diseases
40%
Non communicable
30%
diseases
20%
10%
Alignment allows comparing proportions across groups 0%
1990
2000
2010
Year

18. Comparing proportions across groups
No logical order: Horizontal bar chart
Sort according to decreasing proportions
Logical order: Vertical bar chart
Not a continuous variable : Do not display axis
Continuous variable: Display axis
This is where proportions don’t have to add to 100% across groups, or where there are too many groups for a pie or stacked bar to be clear
Analog information

19. Causes of non vaccination as reported by the mothers, Bubaneshwar, Orissa, India, 2003
Lack of awareness
Child sick
Irregularity by health staff
Lack of motivation
Lack of time
No logical order (categorical?) – no axis as would imply category between irregularity and lack of motivation
Lack of facility
Lack of money 0%
20%
40%
60%
80%
100%
Analog information
India FETP

20. Vertical bar chart: Gradient, but not quantified (No x axis)
Maternal education of mothers, vaccine coverage survey, Yamoussoukro, Ivory Cost, 1995 70 60 50 40
Frequency (%) 30 20
Logical order (ordinal) 10
None
Prim
Sec
Sup
Level of education
Analog information
EPIGEPS course of field epidemiology

21. Can use x axis as age is continuous
Prevalence of hypertension by age and sex, Aizawl, Mizoram, India, 2003 6 5 4 % 3 2
Can use x axis as age is continuous 1 3 - 3 9 4 - 4 9 5 - 5 9 6 - 6 9 7 + A g e g r o u p ( y e a r s )
Vertical bar chart: Quantified gradient (x Axis)

22. Deciding the type of graph that is needed (for reference)

23. Putting it on the map: using maps in epidemiology
Lecture notes

24. Learning objectives Understand the principles of mapping
Understand maps of counts and rates
Understand how maps can generate hypotheses
How to create maps

25. Two basic tips Maps in background Politics
Include if location is relevant to your key message
Politics
Some maps/borders/ comparisons are sensitive
Avian influenza in Norfolk

26. Using maps in field epidemiology
Principals of case mapping – spot/count and population-adjusted maps
Using maps to generate hypotheses
Practical and analytic aspects of maps

27. First, catch your hare Where are the cases?
What do we mean by “where”
Once we know where they are, we can:
Go back to them for case management/investigation
Create a map to generate hypotheses or convey a message

28. Geolocation Can use GPS or address- shows postcode location Case Onset
Other
Geolocation
Case 1
24-May
....
Case 2
25-May
Case 3
26-May
Case 4
27-May
Case 5
28-May
Case 6
29-May
Case 7
30-May
Case 8
31-May
Case 9
01-Jun
Case 10
02-Jun
Case 11
03-Jun
Case 12
04-Jun
Case 13
05-Jun
Case 14
06-Jun
Case 15
07-Jun
Can use GPS or address- shows postcode location

29. Choosing a geo-locator
Cases are people (or animals!)
People move around
Options:
Place of residence (temporary residence)
Place of work or study
Healthcare provider location
Bed/room?
Merging into exposure locations e.g. Pools, water sources
Note time also – location when?

30. General points on mapping
Two types of maps:
Spot maps
Incidence maps
Use a key
Add a title with time, place and person information
Now on to maps.
Do not use tables to show geographical information: You would loose a lot of information!
Field epi map

31. Techniques for mapping
Pen and paper
Drawing packages (including Excel)
Transparency method
Simple mapping- EpiMap, HPZone
GIS packages- ArcView, MapInfo
Can generate derived geographical variables

32. Geographic information system (GIS)
A database linked to geographical information
Location obtained:
Directly via GPS devices or similar
Indirectly from information on address/GP/lab
Can serve as case database
Also can generate maps and test hypotheses
Essentially a database linked to geolocation data
GPS

33. Drawing a spot map during an outbreak investigation
Rough sketch of the setting of an outbreak
One dot = One case
Other locations of potential importance are also recorded
Does not adjust for population density (OK in small places)
This is the basic spot map drawn in the field to spot cases.
Field epi map

34. Using simple packages for maps: Excel
Map of cases of respiratory disease in a community hospital – any thoughts on mode of spread? In reality staff involved (not mapped!)

35. Spot map: ornithosis cases by place of residence, East of England 2008, n=3
First 3 cases were all hospitalised near there place of residence

36. Spot map: ornithosis cases by place of residence, East of England 2008 (n=10)
We asked cases where they worked, and they answered “poultry factory” – they did not know the address

37. Spot map: ornithosis cases by place of work, East of England 2008 (n=10)
Cases were transported to work via minibus, so they did not need to know the address

38. An incidence map adjusts for population density
List the cases
Regroup cases by location for which population denominator is available
Look up census data
Divide the number of cases by the population denominator
Choose gradients of colours to represent increasing incidence
The incidence density map adjusts for population size. –where might this be helpful?
It uses the count, divide and compare approach.
Field epi map

39. Incidence map Invasive meningococal disease, East of England
This is 5 years’ data. You can spot the cities . Census data provided population

40. Using colours in maps The cold / warm scale
Represents violent contrasts
Increasing density in one colour
Represents increasing levels of magnitude
Complementary colours
Use equivalent intensity
Represents unrelated notions
Tips

41. Generating hypotheses
Brainstorm likely geographical links
Map cases plus other relevant features
Water sources
Roads
Cooling towers
Can also map epidemiological indicators by area
Type of ground
Deprivation
Prevalence of another infection/disease

42. Mapping an area exposure/determinant
Also use surveillance data, eg. Prevalence. Test has to be on table data – meningitis risk higher in more deprived wards (local area)

43. Q fever: incidence and spot map
To generate a hypothesis about spread – dots show goat farms, shaded areas are reported Q fever incident. Link to case study

44. Take home messages
Know how to obtain geolocator information and link to cases
Use spot maps and incidence maps to generate hypotheses
Adjust for population size with rates when needed
Communicate efficiently the spatial distribution of health information
Practice map preparation to produce them rapidly when needed in practice

45. Graph exercise: reported reasons for not swimming at Lazareto (% responses)

46. Questions
Message
Graph type
Style

47. First attempt: Select, Insert graph
What’s wrong? Lots, but start with message. Should we put secular trend with comparison of proportions?

48. Message: comparing proportions of responses
But best to compare proportions using horizontal bar – and to sort by descending frequency

49. Type: Bar chart showing proportion by category
Categories are not ordered, so order by proportion

50. Style: reduce unnecessary ink

51. Don’t forget the jellyfish
There were two messages- don’t forget

52. QUESTIONS?

53. A quick and not-so-dirty electronic map in three steps
A hardcopy of your map
A transparency
Cello tape
Permanent markers
Computer
Drawing software
Tips

54. Step 1: Place transparency on the hardcopy of map to draw map with permanent marker
Tips

55. Step 2: Stick the transparency on the screen with cello tape and follow the guide to draw map with the mouse in a drawing software
Tips

56. Step 3: Remove the transparency and edit the map in the drawing software
Tips

57. Your map is ready for use
Your map is ready for use! Edit lines and fill, add title, legends and footnotes
Tips

58. Fraction of HBV infections attributable to sharps injuries among health care workers, by region, 2000
Eastern Europe
Western Pacific
W Europe
Former USSR
North America
S. E. Asia
M. East
South Asia
Gulf
Australia Japan
Africa E
Africa D
Latin America
< 1%
1-9%
10-19%
20-29%
30-49%
> 50%
World Health Report, 2002
Tips

59. Further functions of GIS
Point in polygon
Distance to vector (e.g. river, road)
Raster – defining exposure areas
Satellite images
”Hotspot” maps
Interactive maps :