ONE MILLION DEATHS, TWO PAGES
Background: Public-health experts need death stats to monitor disease and assess interventions, but quality mortality data are scarce in most developing countries.
75 percent of the 60 million deaths around the globe are in low- and middle-income countries such as India, where cause of death is often misclassified or unreported.
One research group is determined to get a clearer picture how people die in India, and they have recently published their Million Death Study (MDS), a massive effort that involves biannual in-person surveys of more than 1 million households across India.
In a recent issue we explained the first results of the study graphically on a two-page spread (as seen above, or find the pdf here).
Design challenge, Part 1: Information design
The most obvious challenge was deciding what to include from the huge survey. Writer Erica Westly and editor Brendan Maher worked with the Nature art team to narrow down the most interesting findings, such as a surprising number of snakebite deaths.
Designer Jasiek Krzysztofiak from the art team used Erica and Brendan’s first draft sketch (second image) to create a two-page spread that 1) told the story of how the data was collected; 2) graphically displayed the findings; and 3) put the data in context.
This from Jasiek:
“We decided it was important to explain the research process at the beginning, before plowing into the data, to give readers an immediate impression of the unique ‘door-to-door’ nature of the project. We used big numbers to emphasize scale and small cartoon illustrations to convey a sense of humanity.
We felt the central visual should be a map of India presenting the geographic distribution of the key findings. (See more about the map from Chris, below.) The map serves to orient the reader at a glance, and shows the key concept of population density (rural deaths were a focus of the study, as they are not always properly documented).
The 6 causes of death flagged on the map are then explained in more detailed on the second page. I created icons for the 6 causes to help with wayfinding, and then set to work trying to visualise all the data in a clear, legible way.
The main challenge here was to compare MDS and WHO estimates, show all the data for male vs female, home vs hospital and urban vs rural together, in one comprehensive graphic. Some of our initial concepts turned out to be too creative and difficult to understand, such as plotting causes on a matrix with urban/rural versus number of deaths, or representing numbers with human icons.
Eventually we settled for simple, easy to compare solutions, using bar and pie charts to present information. “
Design challenge, Part 2: The map
As noted above, the main visual element on the spread is a map of India that shows current population density. As the concept of ‘urban vs rural’ is central to the rationale of the MDS, we thought a density map with current numbers would be indispensable.
Easy, right? Wrong! This from Chris Ryan, who developed the map:
"Create a map of map of India with each of the districts coloured according to their population density". Simple, huh? Well sort of…
The population density data from the Indian census of 2011 is freely available on the Indian government’s website and they also publish maps with all of the districts labelled. Great!
The drag is that there are soooo many districts. Around 640 depending on who you talk to. Whilst I could sit down with Adobe Illustrator and colour each district ‘by hand’ the potential to introduce errors is pretty high.
To get around this - and save myself from a nasty case of RSI - I devised a ‘pure data visualisation strategy’. That is, to write a script to tell my computer to reach out to the internet to grab the data and then present this data as a map that can be extracted and used for the print layout.
That was the plan anyway. Unfortunately finding two datasets that agreed on the names and locations of all of India’s districts proved to be quite a challenge. For a start, India either has around 598 or 641 districts depending on where you look. That jump happened between the 2001 census and the most recent in 2011.
Also there are many inconsistencies in how the districts are named that caused my system to break down. For instance a human can quite easily ascertain that the labels ‘Dharbanga’ and ‘Darbhanga’ refer to the same place but it’s hard to program a computer to come to the same conclusion.
In the end I had to manually edit around 140 entries in the population density spreadsheet so that they married up with the district names in the map. In the process risking the introduction of errors and a nasty case of RSI.
The next problem was devising a colour scheme that accounts for the wide disparity in population density within India. Heavily populated areas can have upwards of 37,000 people living pre square kilometre, whereas remote rural regions will only have around 30.
Using a colour scale that ranges from 0 people per Km2 to the maximum produces a map that, whilst technically correct, seems to suggest that most of India is empty.
To get around this we clamped the scale at 16,000 - meaning any district with a population density above 16,000 people per Km2 will receive the same colour. This compromise allowed us to indicate the densely populated areas without loosing the granularity for the less inhabited rural areas.”
If you’d like to take a look at the raw SVG map it can be seen here:
And all the code can be accessed on github:
-Kelly Krause, Chris Ryan, and Jasiek Krzysztofiak