1.11

# What is a good map?

Note: The animation above was created by Prof. Menno-Jan Kraak (University of Twente) and shows the diffusion of Black Death over Europe during the 14th century. Please note that this animation does not contain any audio.

Maps provide the opportunity to visually analyze the distribution of disease over space, e.g. the location of health facilities in relation to the distribution of the patients. We often refer to this as “visual analytics”. We can learn a lot from maps, as our human brain is very well equipped to detect spatial patterns and relate spatial elements.

What are we mapping within the GeoHealth domain? We can divide the map types into the following categories:

• Locations of patients or health facilities
• Absolute number of disease cases.
• Area of disease (where does the disease occur)
• Prevalence rates or ratios (how many disease cases per standardized unit)
• Catchment areas of medical facilities
• Variation in space and time (diffusion of disease)

## Point Maps

Visualizing point data is probably the easiest type of visualization. This applies to mapping patient locations as well as absolute number of disease cases. We can use the size of the point to indicate how many disease cases it represents. But representing absolute numbers is dangerous as some areas may have many inhabitants and others are almost uninhabited. The interpretation of this type of information is not always easy.

Point map, each point represents a health center, the size of the dot indicates the number of Measles cases

Choropleth map, each area represents a district. Colour shows a classified number of disease cases per 1000 inhabitants.

## Polygon (Area) Maps

Polygon Maps often visualize disease incidence aggregated to administrative units, by using choropleth maps. These maps show the outlines of the administrative units (districts) and use color or shading to represent disease incidence. In such a case we cannot use the absolute number of disease cases but need to use relative information (e.g. the number of disease cases per 1000 inhabitants). They can only be used when your data shows rates or ratios and the phenomenon you try to map is continuous (can be measured anywhere in space). For choropleth maps, the number of classes you use and the classification method (how you divide your data into classes) will determine the pattern that you will see.

Sometimes we are not visualizing disease prevalence but an area in which a certain disease occurs. This can be done by selecting administrative units (countries) and giving them a colour indicating that the disease occurs in these countries. You have to realize though that the complete country will be coloured and in case the disease only occurs in a small part of this country this is a big over-estimation of the risk. The smaller the administrative units you use for such a visualization, the better you will represent the actual area at risk.

Light blue area shows the area in which a disease occurs (for example, the area of tick bites causing Lyme disease). When country boundaries are used, this can be misleading indicating a complete country as being at risk.
Map showing the allocation of points (households, to a medical facility (blue squares) by drawing lines between the facility and the household.

When we want to visualize the service area of a medical facility we can generate a polygon (areas) around this facility indicating a certain distance (either over the road or as the bird flies) or travel time. We can also visualize the catchment by drawing lines (we call these desire lines) between a village, community or house location (communities or houses) and the hospital they go to.

## Animations

A map normally only captures one moment in time. When we constantly collect data via a surveillance system, we do not only want to know where the disease occurs or which people visit a certain hospital but also want to visualize changes that occur over time. As the amount of spatial-temporal data is growing fast, we will need more spatial-temporal visualizations like animations in future. Play the enclosed animation of the spread of Black Death over Europe on top of this page and you will know what we mean.

Maps can be misleading in many different ways. We already introduced some examples of this in the text above like the fact that you need to use relative data when you are using choropleth maps, and that your visualization will change when you change the number of classes or the classification method. Color can also be very misleading. Our brain associates certain colors with danger (e.g. the color red), but associates other colors with safe (green).

## Get a taste of this course

Find out what this course is like by previewing some of the course steps before you join: