I want to send a very big and public thank you to the Health Informatics and Financial Reporting leadership at Johns Hopkins Children’s Hospital in St. Petersburg, Florida, for instructing their teams to put an away message on e-mail, shut off their phones, and spend two days immersed in my data visualization training. Man, do they get it!! People need uninterrupted time to learn new skills in an engaged and focused manner and — wait for it — to THINK.
But here’s the deal: it’s not just training that requires that time. Finding solutions to problems, whether simple or gnarly, also requires time to THINK. (Contrary to popular belief, those problems can’t be solved in the blink of an eye by FM — Freaking Magic.)
Let’s look at a visualization I’ve been re-designing that illustrates this idea. I was able to say right away, “This thing is hard to understand” — but that didn’t mean I could come up with a solution equally quickly. That part took me some quiet, uninterrupted time to sketch and consider possible solutions.
This visualization displays the intersection of two indicators from a public health survey. The first indicator, a composite of respondents who answered “yes” to six questions about mood, anxiety, and depression in the 30 days preceding the survey, is labeled “Serious psychological distress.” It is crossed with the indicator “Health insurance coverage” (indicated by answers to the question, “Do you have a health insurance policy?”).
The challenge of this visualization is that the viewer has to try and match the “yes” responses displayed in the bars of one indicator with those in the other, then try to decide whether respondents who experienced serious psychological distress had health insurance.
If you look closely, you will notice that the labels are hard to decipher: the two bars on the left are whether or not respondents answered yes or no to “Yes,” Psychological distress; the two on the right are whether they responded yes or no to “Yes,” they had health insurance (I think). Then you have to match the blue to the blue bar and the red to the red one. I admit I’m still wildly confused.
As a first step, I arranged the responses to the two indicators in a simple matrix, then just studied them for a bit.
The simple act of creating this matrix helped me realize that the denominators for the Second Indicator (Health insurance) are the YES and NO responses to the First Indicator (Serious psychological distress). This may sound like a minor insight, but it was a key first step to figuring out how to better display the information.
Next, I started to sketch possible ways to display the results using a simple Excel sheet to create the following visual without any competing distractions. I allowed myself some time just to think about it.
In this new visual, I’ve directly displayed the results of stratifying Indicator 1 results by Indicator 2 results. Now we can see that:
- Of those respondents who answered YES, they had experienced serious psychological distress, only 5.7% had a health insurance policy compared to 89.1% who had no serious distress and did have health insurance.
- It may be worthwhile to analyze in greater detail those who suffered serious distress and do not have health insurance. What are their demographics (age, sex, education, income level)? Where do they live? Such analysis can help us focus efforts to create or deliver (for example) new resources and support services.
- There is power in taking time to create detailed, plain-language labels to eliminate the mystery of what is being displayed.
I love meeting, training, and collaborating with all our clients; but it is especially gratifying when those same clients give the projects we’re working on their full attention — and allow all of us time to think. That’s really the only magic required for creating great work.