How To Show How We Die

A recent Guardian data blog article unveiled an infographic showing causes of death for 2011 in the UK. These are the official causes recorded on the coroner’s report, so “old age” isn’t generally considered detailed enough.

Mortality rates big graphic

There’s a problem with this way of depicting categories and subcategories of data. Quite apart from the problems we as readers have in estimating the areas of circles and comparing one to another, the hub-and-spoke design used here repeats some of the data points several times.

Mortality rates big graphicFirst, the central circle (the total) is fighting against the main story the article is trying to tell, that more people now die of cancer than circulatory diseases. It actually takes a moment’s thought to realise the enormous circle in the centre is made up of all the circles connected to it; that it’s just the dataset itself.

This problem becomes increasingly severe as you work your way into the data. Look at the 2,164 hanging/strangulation deaths—they’re actually a subset, not a sister category of the 3644 intentional self-harm deaths (which in turn are included in the 17,590 category total). None of this is obvious at first glance, and a journalist in a hurry could quite easily write down both numbers as if they were independent. The hanging/strangulation data is actually represented on the chart four times (these three numbers, plus the main hub), whereas some categories like skin diseases only appear twice.

One way around this problem is to represent each data point or group of data once and once only, so that the size of categories accurately reflects their importance. I created an infographic to summarise the equivalent data for New Zealand, with the help of Dr Siouxsie Wiles (Auckland University) and Dr Paul Gardner (University of Canterbury). I decided to omit the detailed numbers and percentage increase/decrease the Guardian used, instead showing the general proportions of each cause of death. Siouxsie ran an online survey (results summarised here) to test whether the public’s perceptions of how they might die matched reality.