Climate Change and Pharmaceutical Use
Another visualisation project at the European Centre for Environment and Human Health that I've been meaning to write up for a while now. This one was inspired again by Dr Clare Redshaw (who was also a driving force behind the Pharma Town graphic). We actually began talking about this project before that one, but it ended up being pretty extensive, and was only published earlier this year. Now it's in print, I am at liberty to blog about the design process for the paper figures.
The first thing that Dr Redshaw presented me with was a series of three tables. The first of which looked like this:
A lot of people are very apologetic about presenting me with Excel tables. Which is odd, because I love them. It gives immediate access to the data you'll be working with. Way better for a designer like me than some data locked up in a horrible bar chart or twelve, anyway.
This data is all relational rather than numeric, but it's quite complex in its own way. The left column shows diseases, the second is a series of climatic conditions that will lead to increasing prevalence of these diseases. The third column is the medications used to treat these illnesses, and the fourth gives some examples of the drug names these are marketed under. The challenge was to try and give an overview of which pharmaceutical products would be used more often because of climate change. Many are used to treat multiple conditions, and climatic changes often lead to multiple changes in disease prevalence.
To begin with, I started thinking about the three categories of drug, disease and climatic factor (by drug meaning the general medication rather than the specific brand). Our first thoughts were to focus on the person, linking diseases to body parts, and then medications to diseases, in concentric rings, joined with lines:
However, we quickly realised that this wasn't the overview we were looking for. We went for another level of abstraction, and just thought about how the three sets of information were related:
The idea on the left is actually the fundamental structure of the display we ended up with. However, put so simply I felt it was a little too abstract. I then thought about giving each disease a box, and then symbols within that would show the drugs used to treat it:
Then I realised that this wouldn't show how often a particular drug was used, so I decided to link these to other symbols outside the boxes:
Then I used the Gabriel Rogers Principle (so named for a researcher I worked with during my PhD, who once cynically remarked that if an information graphic didn't look good enough, most designers just rearranged it into a circle and hey presto - cool graphic). In this instance, I think the use of the circle is justified though, as it means that all the diseases can be linked to the central area, preventing very long lines from one end right across to the other (at least, that's what I like to tell myself...)
At this point, I also added bars around the outside to show which climatic conditions were related to which disease. A slightly more detailed sketch, and I was good to go:
I used the real data for the final sketch, including all the names of climatic conditions and diseases. To check that the medications fitted in the centre, I counted how many times each appeared, and filled out the correct number of links. Although I didn't actually connect them to the correct diseases at this stage, it gave me an idea of how much space I would need. I also sketched out the other two graphics in the same way, to check that the volume of data would be similar enough for the same treatment. You might notice there is a gap at the top. This is because the different graphics needed different numbers of segments, the maximum being 16. As it's far easier to divide a circle into 16 than 15 by eye, I just left a gap for this first figure that only needed 15 disease wedges. The serious calculations would come later, when I fired up Illustrator:
The first version wasn't pretty, but there were a lot of carefully-calculated angles behind that! The coloured lines are just there as paths to put text on. The pink is for the titles, and the blue for explanatory text. I chose some (fairly boring) colours to represent the five different climatic conditions we were presenting.
Then I added a scale and the drugs into the centre. Using the same drug symbol as past graphics, I hoped to continue the visual language I was establishing in this area.
After adding in all the drugs, linking them all to the diseases and finding a unique colour for each one, I had one of those moments that every designer dreads. I realised that the graphic had become quite cramped in the centre - and that the second graphic had even more drugs to squeeze in. I probably had to take a few deep breaths before I selected the drugs and links layers and hit the delete button...
I agreed with Dr Redshaw at this point that the disease descriptions were not adding much to the diagram, and could be removed to create more space for the drugs, so I cut a larger circle in the centre:
I also felt that using entire wedge sides as arrows to show where diseases were linked to each other wasn't really clear enough, so I added in some actual arrows. I also flipped over the text for the bottom half of the diagram to make it legible. I also wasn't happy with the pills in the centre. It wasn't easy to see how many conditions were linked to each one, so I tried a couple of different representations:
At this point I actually began working on the other diagrams. For reasons that will become clear, I needed to add in some extra parts at the edge to make the graphics a coherent set:
At this stage, Dr Redshaw came up with a brilliant idea, which I could now include. I had earlier considered using the same colours for the central pharmaceuticals as the climatic conditions around the edge, but at this stage the circles were still solid, and I couldn't work out how to colour circles that were linked to diseases caused by different climatic conditions. Now that I was using the "wedges" representation for the drugs, it was easy to colour each segment according to the link line from the diseases:
There was really no need to colour each drug, and probably more confusing than anything else. This also allowed me to use a brighter colour palette:
The key needed a bit of work to associate its title better with the content. Also I had to restructure it a bit as I needed to add in a 6th item (we combined two kinds of antibiotics together to make a larger group):
Lastly, this gave me space to explain what was represented by the different parts of the diagram with another part of the key:
The second graphic was rather more challenging, as it had a fourth layer of information to include (hence the outer ring of gradients in the earlier graphic) This second graphic dealt with "vector-borne" diseases (carried by insects known as "disease vectors" - such as mosquitos and ticks). These vector organisms generally live on "host" organisms, such as rats, birds and even humans. I represented these on an outer circle of gradients, with the vectors themselves shown using the lines that had been used in the previous graphic to show climatic conditions:
This time it was more challenging to fit in the pharmaceuticals in the centre:
However, I was limited by the size of the print copy of the journal, so couldn't enlarge the size any more. I decided that, while the extra drugs gave it a cramped feel, this did at least show that there were more different treatments for the vector-borne diseases (and in the end, it provided a good comparison with the third diagram, which was very sparse).
This version was challenging to colour, as there were more segments. In the end, I decided to limit the number of colours used, as there was often an obvious link between the vectors and a particular group of hosts:
For the last graphic, the pattern was quite well-established. However, I found the different sections more challenging to label (appropriate diagrams for fecal-contaminated food anyone? or anal/oral sex?) Also, three of the health conditions were primarily vaccinated against, rather than treated with pharmaceutical products as such. I represented these with outlines rather than solid blocks of colour (as with the single surgical intervention).
We realised that these extra kinds of intervention were more of a distraction from the central message of the paper, so we decided to remove them (along with rehydration salts, which aren't really a pharmaceutical product as such). This also allowed us to remove hepatitis C from the diagram, and thus the anal/oral sex (phew!). We also realised that some of the products used to treat shigellosis and salmonella were actually linked to secondary arthritis (which is good, because if we didn't have that to fill the gap I'd have been redrawing the whole diagram). I also redrew the food/water illustrations to be more like the rest, which were all vector-traced from photographs. I wasn't at all happy with the illustrations though (particularly the thing that's supposed to be a prawn in the bottom right. Some things just don't outline well...)
Somewhere around here we realised that three of the conditions could also be picked up from swimming with an open wound. Not conditions that were next to each other, of course... I also realised that our labelling of the categories wasn't consistent. The raw and uncooked meat diseases were grouped together, but were separated for seafood. I brought these categories together, and rearranged the diagram substantially:
A quick rotation was needed to give an obvious horizontal axis along which I could flip the text from one orientation to the other:
I then redrew some of the illustrations around the outside - the clam shell being much more recognisable than the prawn attempt I think. The final version of this one:
As you might have guessed from the way that the nicer key magically appears at the end of each graphic, this isn't the actual sequence that we went through when designing the diagrams. All three were produced concurrently to an extent. I've produced a short time-lapse video to show the order in which the different versions were done:
The full paper is published in the Journal of Toxicology and Environmental Health, Part B.
Comments