Only Charlotte Needs a Web

Recently my colleague and HealthDataViz Senior Consultant Janet Steeger sent me the graph below and an associated article, with the subject line “Blog Posting?”.

(Click to expand)

Incredulousness was quickly followed by inspiration. How, we wondered, could super-smart people use such dumb graphs to display their data? In response, Janet came up with a great re- design (see below).

In a nutshell, the article is about the Value-Based Health Care Delivery (VBHCD) initiative at Harvard Business School, which engages with leading healthcare provider organizations to measure and manage patient-level costs over complete cycles of care for a variety of medical conditions.

The Radar Chart displayed above (sometimes referred to as a Spider Chart, Web Chart, Polar Chart, or as Star Plots) is supposed to help us understand and compare three different provider results on cost, complications, and patient outcomes that are part of a study conducted by VBHCD.

Why this chart doesn’t work

A radar chart allows the viewer to compare multiple quantitative variables; some would also argue that it is most useful for viewing data outliers and displaying performance. Here, however, are a few of the reasons why this chart simply doesn’t work:

The values being displayed are different, and have therefore been rendered on the same scale so that they can be displayed together on this chart. The notes at the bottom of the chart indicate that a score of 100 equals the lowest cost, which to my perception is completely counter-intuitive, annoying and confusing.

Not only is it hard to visually compare the lengths of the different spokes shown; it is also hard to hold the scale in our memories and accurately judge the radial distances.

The lines connecting the data may occlude (obstruct) one another as the values grow closer, or if they are identical. It’s inevitable that when multiple series are plotted, some values will eventually be on top of and therefore blocking each other.

The area of the shapes presented increases as a square of the values rather than linearly. This may cause us to misinterpret the data displayed, because a small difference in the values results in a significant change in the area, so the difference is visually exaggerated.

Now we get to the part about the inspiration that followed our astonishment over anyone’s using such a misleading display format. Let’s agree to leave these radar charts that look very much like Charlotte’s web to our friends the spiders, and consider a better way to display this data (with thanks again to Janet for her re- design).

(Click to expand)

By using bar charts arranged one right below the other, and displaying the actual data (versus the converted form used to display the data in the radar chart), we can easily see and understand that:

Surgeon B has the highest cost for Total Hospital Stay and Post Acute Care, the highest Postoperative Occurrence Rate and Readmission Rate, and the lowest Patient Improvement Scores across the board. (Which brings to mind the iconic Ricky Ricardo’s line “Lucy, you got some ‘splainin’ to do.” But I digress.)

We can also quickly and easily see that although Surgeon C’s Total Hospital Stay Cost is the second highest at $14,000 (it is still $4,000 less then Surgeon B’s), his/her Postoperative Occurrence and Readmissions rates are by far the lowest at 4.0% and 2.0% respectively, and Patient Improvement Scores are the highest.

At first glance it would appear that the extra $4,000 for Total Hospital Care connected with Surgeon C was money well spent. We’d want to dig into the data a bit more to see if we couldn’t learn a thing or two to share with the other surgeons to help them achieve the same or similar results.

Here again, dear Subscriber, we are faced in the radar chart with a display that may look cool, but that simply doesn’t work. I think of what a wise little spider once told her dear friend Wilbur:

“Trust me, Wilbur. People are very gullible. They’ll believe anything they see in print.”

E. B. White, Charlotte’s Web

Just because some very smart people decided to use a seemingly cool chart, we need not be so gullible as to believe it is the best way to display our data.

Share
This entry was posted in Communicating Data to the Public, Data Visualization, Newsletters. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *


− four = 2