Using Evidence, Visualization and Statistical Thinking to Improve Healthcare--Howard Wainer

Although not necessarily a book for the expert in the field of data visualization, Wainer refines the finer points of displaying medical evidence accurately and in a compelling informative manner. The use of examples from the literature citing such pioneers as Edward Tufte and even reaching back to origins of evidenced-based medicine by Pierre Charles Alexandre Louis's 1835 published account discrediting the efficacy of "bloodletting".

One of the most interesting stories presented in the Medical Illuminations text centers around the introduction of Will Burtin as a pioneer of scientific visualization beginning in Nazi Germany. Burtin was a designer coming to notoriety at the brink of a world war and developing a lengthy association with UpJohn Pharmaceuticals--including a collaboration to construct a large model of a red blood cell at the San Francisco 1958 American Medical Association meeting.

Visualizing data displaying the effectiveness of penicillin, streptomycin and neomycin on 16 different bacteria provided an informative graphic to help guide clinicians and researchers both in the era of the new "wonder drug" of medicine.

---

His display is a direct lineal descendent of Florence Nightingale's famous Rose, in which the radii of the segments convey the amount of data, rather than a traditional pie chart in which the angle of each segment is the carrier of the information. Burtin does several clever things in this display, two of them, of special interest are described here.

First, he saw the huge range of values that the data took and realized that some sort of re-expression was necessary. He chose a logarithmic transformation. Such a re-expression is obvious to someone with statistical training, but is reassuring that a designer should come to the same conclusion...

His resulting display is compact, accurate and informative. But with the wisdom borne of a half century of work on statistical display and exploratory data analysis, can we improve matters?

Burtin's diagram compares impacts of Penicillin, Streptomycin and Neomycin on a range of bacteria (Scope, Fall 1951)

From here, you are artfully led toward potential improvements to an already spectacular graphic helping you to refine your eye and the use of graphical tools.

You will need to refer to the book directly for details that suggest avenues for additional improvement but here is a handy tool below.

A few suggestions to keep in mind...

---

1. To improve visibility, light colors should be avoided--don't assume users have access to color printing so include distinctive markings that will be retained in black & white printing
   
2. Label both ends of a graph if lines are crossing and there may be some confusion
   
3. Select appropriate spacing intervals for axes
   
4. This is a pet peeve of mine--labels MUST be easily read and positioned clearly
   
5. When including a "Total" or an "All" label it should be clearly distinct from other labels
   
6. Use the x-axis label to clearly describe the display (it should be explicit)
7. Remove extra gridlines where warranted and possible
   
8. Points should not sit on the axis--add appropriate spacing
   
9. Remove the extraneous plot points once the graphic is rendered
   
10. Font selection should be deliberate and effective--try something new to provide visual interest
    

---