Visualising Disease Distribution Across the US

Visual Design Pie Charts
#1

Hi Data Visualizers,

I’ve attempted to show the distribution of common diseases through the continental US. The goal is to identify trends in the data and deduce a causation from the correlation. I’ve selected measles, TB, meningitis, variola virus and streptococcus cases between 1900 - 2014. The data is pulled from (https://www.tycho.pitt.edu).

PNG
PNG1435×711 240 KB

Visual Design Type: Choropleth Map, Piechart

Name of Tool: Tableau

Country: United States

Disease: Measles, TB, Meningitis, Variola virus, Streptococcus

Year: 1900-2014

DOI: 10.25337/T7/ptycho.v2.0/US.14189004

Visual Mappings:

Colour:
State colour intensity calculated as percentage case numbers of total.
Distinct disease cases shown as different colours on pie chart.

Position: Cases are bound to their geographical positions

Unique Observation:

We can see the majority of cases occur in the north east with Texas, California, Washington and Colorado also having a high number of cases. A correlation could be drawn in those states having large cities as compared to other mid-western states.

Distribution is fairly consistent with measles being the most diagnosed ailment.

Louisiana is an outlier with proportionately greater number of TB cases than other states.

Streptococcus seems to be more prominent in the mid-west.

Data Preparation:

Case numbers calculated as percentage of cases per state over all reported cases.

Mapping state/country to altitude/longitude.

So here’s some questions, any feedback welcomed.
Is my palette choice optimal to distinguish between diseases?
Have I selected too large a time-frame to extract any useful information?

Thanks for your time.

Chloropleth Map: https://datavizcatalogue.com/method/choropleth.html

#2

In terms of color choices to distinguish different diseases you could check if your choices align with a colorbrewer 2.0s suggested palettes.

If you’re worried that your period is too long you could implement the same visualisation in a language like D3 allowing for interaction such as a time scale to filter data.

#3

Is there merit in separating the two encodings? At the moment the map has an overplotting issue and the information about smaller sectors of the pie is lost.

As an alternative, you can have the map and a side panel with the pie charts. Since you’re using Tableau, you can allow user to draw initial inference from the choropleth and the follow up with a detailed view that includes one or more pie charts (depending on how many states the user selects on the map).

A more general point on choropleth maps - the size of the area has an perceptual influence on your users: larger states command more visual attention than smaller states. However, this attention doesn’t necessarily align with your communication aims and is separate from the underlying disease data (which is population-based, not geographic area-based).

#4

Hi @Corrul,

Regarding your questions, I don’t necessarily think there is an issue with colours you have chosen, but the pie charts are so small (and sometimes overlapping) making it difficult to distinguish the smaller segments.

Perhaps you could plot only the most prevalent disease per state to get a clearer idea of distribution? It would likely lead you to make the same unique observations you have already made.

All the best! :slight_smile:

Ben