Knowing that this theatre has 1,000 seats, we get an idea of how common the given adverse effect is. The visualisation is particularly effective when two or more graphs are placed next to each other, one showing baseline risk, the other the effect of treatment, or comparing different risk scenarios.
This visualisation makes it possible to show relatively rare risks that are often difficult to represent (anything smaller than a couple percentage points would look negligible, e.g., in a pie chart, but there’s a big difference between a 2 per cent risk and a 0.02 per cent risk), allows to associate a risk with a quantity and context that a person can instinctively relate to, and by not giving prominence to exact figures, it implicitly deals with the big issue of uncertainty in risk.
Overall, I feel there are clear merits for this type of visualisations, and surely enough merit to deserve a modern graphical implementation of this technique enabling its wider adoption, be it among medical professionals or data journalists.
Before moving on to more examples and introducing our own implementation of this concept, I will state the obvious: this visualisation is effective in some cases, but not so much in others; sometimes, we do have a very high degree of certainty in a given data point; getting familiar with research data may be useful, but context remains fundamental and consultation with professionals may be necessary for a better interpretation of risk statististics (indeed, the book is at least partly addressed to doctors seeking to present risk to their own patients during such consultations).
A modern visual implementation of the risk characterisation theatre
Solutions that share some of the features of the risk characterisation theatre designed by Rifkin and Bouwer are not unseen in data visualisation. For example, when showing the impact of “long Covid” according to a recent study, the data team at The Economist showed the frequency of different symptoms by changing the shade of the relevant share of dots out of a bunch of 100 dots: this shows relative and absolute numbers, but somehow lacks the natural association with a real world scenario that defines the risk characterisation theatre. Instead of showing a bunch of dots, for example, they could have shown them as places in a bus: if you imagine a typical bus with 50 places, then a figure such as “22 per cent of patients discharged from hospital after covid-19 reported hair loss” could be depicted as 11 individuals on a bus full of people who have left hospital after receiving care for covid-19: a situation that can easily be imagined by anyone familiar with a bus. The data become immediately less abstract.
To facilitate producing this kind of graph, I have created a package for the R programming language,
riskviewer. It currently offers two basic scenarios:
- one based on a typical configuration of many airplanes common in European routes such as the Airbus A320 or the Boeing 737, with about 30 rows, 6 seats per row, and close to two hundreds seats in total.
- one inspired by the Verona Arena, a Roman amphitheatre still used for performances and concerts (here is the official seating chart); in my implementation (not an accurate replica), the Arena has approximately 10,000 seating places;
Both settings should be relatively familiar to many in Europe, or at least it should be easy for them to visually imagine the size of the crowd involved. I see the advantage of choosing seating charts of smaller size (such as buses), but it is not uncommon to be in the position to represent relatively rare risks, whereby larger seat charts can be more useful.
Let’s start from one of the examples Rifin and Bouwer outline in their book: the British Doctors Study, a 50 years long study that ran from 1951 to 2001 and determined the increased risk of lung cancer associated with smoking tobacco.
There are many ways to present these data, and many data points that outline, for example, the relative benefits for those who quit smoking earlier. But let’s just look at one of the most blunt data points: the number of men who lived from age 35 through age 80. Among smokers, 74 per cent died before reaching the age of 80, compared to 41 per cent of the non-smokers (figures relate to people born in the first decade of the last century).