Best practices for healthcare data visualization
As more health systems work to harness their data for quality improvements and efficiency, many are realizing that it's about a lot more than just numbers. Intuitive and communicative charts and dashboards are key to ensuring the right story gets told.
At the HIMSS Big Data and Healthcare Analytics Forum in San Francisco on Tuesday, Teresa Larsen, director of the Foundation for Scientific Literacy, offered tips for spotting suboptimal data visualizations and recommendations to improve them.
Key to her message was that, just as with spoken and written words – visual images have their own rules of vocabulary and grammar that should be followed for effective communication.
Whether for frontline clinicians or bean-counting C-suite members, analytics reports that tell the the story clearly, convincingly and truthfully are the ones most likely to deliver results. Toward that end, Larsen offered some best practices for creating charts or graphs.
First, the number one principle of good visualization is to "stay true to the data," said Larsen.
To illustrate this, she showed a 2014 Reuters chart (sadly apropos given the tragic recent events in Orlando) purporting to show that Florida gun deaths had markedly decreased since the adoption of the "stand your ground" law.
That's how it might appear on first glance, at least. But a closer look shows that the number '0' is on top of the graph, rather than on the bottom (where ingrained assumptions based on years of looking at similar charts would put it). In fact, the number of gun deaths had risen precipitously since the law was enacted.
"Don't lie," said Larsen. "If the data is telling you something, you don't want to massage it to say something else."
[Special Report: How real-time big data analytics strategies are improving care quality and efficiency]
Updating Mark Twain's famous adage, she cited the dangers of "lies, damn lies and computer graphics."
Second, Larsen said it's key to communicate by design. Image is not unlike language. Like sentences, a design can be parsed.
Larsen called herself a "card carrying member of the grammar police" – and she flashed a badge in her wallet to prove it. (Motto: "to correct and serve").
When it comes to designs and charts, sound grammar is key to clear communication, she said. As an example of what not to do, she pointed to pie charts, which have inexplicably become go-to standard for displaying numerical breakdowns.
"Someone decided pie charts are intuitive," said Larsen. In fact, they're not only not intuitive, they're often not even accurate. (How many charts have you seen where the numbers don't directly correspond to the size of the wedges?)
The best pie chart she's ever seen, she said, looked like the one illustrating this story.
Rather than a pie chart, why not try a stacked histogram? It's much easier to understand and appreciate with a quick glance – what she calls the "blink test."
Third, Larsen said the importance of color choice should not be underestimated.
Don't stick with default choices, she said, which are often garish and oversaturated reds, blues and yellows.
"There is a visual vocabulary," she said. "Colors matter."
To drive the point, she showed soothing photos of Yosemite and the Seychelles – very different landscapes that nonetheless shared similar palates of blues, greens and grays.
Then she showed a photo of a yellow jacket, and one of a snake. "Yellow and black is nature's synonym for danger," said Larsen. Often mankind's too: consider traffic hazard signs and crime scene tape.
A chart with blaring colors will be difficult to parse. "Would you write a sentence with every letter of the alphabet, in all caps?" she said.
Using a more natural palate appeals to people's intuitive understandings. "Colors have meaning – take advantage of what your audience already knows, and they're going to get it."
Finally, Larsen offered a fourth bit of best practice: Ensure that plot dimensions match data dimensions. To illustrate this, she pointed to what she calls a faux pas of visual grammar: a 3D histogram that should actually be the the simpler two-dimensional form.
The reason why probably won't surprise: "It makes it harder to read," she said. "Simplicity is king."
This article is part of our reporting on the Big Data and Healthcare Analytics Forum, taking place this week in San Francisco. Other stories include: Atul Butte says 'precision medicine makes doctors nervous,' and Wowrack partners with Security First to create cloud security services.
Twitter: @MikeMiliardHITN
Email the writer: mike.miliard@himssmedia.com
Like Healthcare IT News on Facebook and LinkedIn