How to Tell Stories with Data in an Unfolding Global Pandemic

Balancing accuracy, clarity, and responsibility

Four months ago, many of us never would have expected our lives to become dominated by charts and graphs. And yet, as the coronavirus began spreading around the world, it became more important than ever to find ways to visually depict the magnitude of the crisis.

As a health- and development-focused organization, Surgo Foundation analyzes the systemic and psychological drivers of health behaviors. When COVID took hold in the U.S., we created the COVID-19 Community Vulnerability Index (CCVI) to understand how vulnerable each community was to the effects of the pandemic, and the ways in which people’s behaviors were changing in response.

The index provides a trove of data about a community’s ability to fend off COVID-19 and the factors impacting vulnerability (from the relative strength of the health systems in an area, to access to transportation to medical appointments, to the proportion of elderly people living in a community).

But as with any massive data set, an index is only as good as its ability to be understood. And understanding can only happen with effective data visualization.

The challenge of visualizing COVID-19 data

Visually communicating the data had to be as central to our approach as our analysis.

This rich but complex data needed to be understandable by a wide audience, including non-experts, with top-line findings presented in a way that would instantly be grasped and formats that would enable users to explore data in-depth at national, state, and hyper-local levels.

Today, almost four months into our COVID-19 data visualization journey, here’s what we’ve learned so far:

Know when to bring in experts: A data scientist is not an expert in communications design — and vice versa.

Recognizing this, we enlisted the help of designer Federica Fragapane and front-end developer Paolo Corti. Federica’s work presents multi-faceted data sets in visually compelling ways, without losing sight of the human stories behind the statistics (as you can hear in her fascinating interview in the Design Emergency webcast series). We had worked with both of them before, so jumping into this journey with Frederica and Paolo was easy. They are great teachers and listeners!

Through our collaboration and learning process, we were also able to discern the instances in which we could rely on our team’s own internal data visualization skills.

Don’t forget about user experience: Effective data visualization is not just about the “look” of a tool, but also about how it “feels” to the user. Elements like menus, buttons, and mouse-overs must be responsive for an optimal user experience.

Adding to the challenge, our COVID-19 data is multi-dimensional — varying by region, state and county, reflecting differing timelines for policy announcements and behavior change, and with case counts that changed by the day.

We had to be nimble but persistent in shaping the design elements until they told the story successfully and in user-friendly ways — leading to the creation of our precisionforcovid.org website.

Iterate, but don’t delay: The best design results from an iterative process in which refinements are made based on internal feedback from the whole team, as well as from users once the tools or designs are published. COVID-19 challenged us to get good design done quickly, because our findings were time-sensitive and needed to be “out the door” fast in order to be useful.

We created some prototype visualizations using off-the-shelf packages like Tableau to get data out rapidly to our audience. At the same time, we created custom designs that told the story more clearly and with a better user experience than a pre-packaged design could deliver. Those visualizations then replaced the off-the-shelf ones, and we improved them incrementally over time — and will continue to do so, based on user feedback.

Use data both to explore and to explain: The maps and other visual tools on precisionforcovid.org allow users to explore data in depth, switching between states, counties, and dates to map geographic and temporal trends. But we were also able to present versions of these to tell stories in our other posts on Medium, interweaving them with narrative text that explains what we’ve learned, and the implications for policymakers and the public in the US and around the world.

Design elements in action

Since the advent of the pandemic, we have tried to tell the COVID-19 story visually in many ways. Our analysis often tries to answer a question.

Here are two examples of how we tackled the visualization challenge:

1. Which are the most vulnerable U.S. communities, and why?

The CCVI assigns each county in the U.S. a composite score based on its overall vulnerability to the effects of COVID-19. This score is made up of six themes — socioeconomic status, access to housing and transportation, household composition and disability, minority status and language, epidemiological factors, and healthcare-system factors — and each theme has its own score.

We wanted to represent the six theme scores visually, so that every county within every state could be compared for each theme. Federica’s visually-arresting solution was the “star graph,” with a line representing each county radiating from a central hub:

In the star graph, each theme is represented by a colored dot, whose position on the line indicates its score — the further out from the hub, the higher the score.

For example, anyone looking at the star graph for California (see above, left-hand side) can instantly see that epidemiological factors (yellow) are not a significant vulnerability for most counties, but that healthcare-system factors (red) are a greater concern, and minority status and language (purple) a considerable vulnerability.

To make the graph as versatile as possible, the user can sort counties by alphabetical order or by CCVI ranking, or apply a filter to display individual themes, allowing comparisons among counties to be made easily. Hovering over a line isolates an individual county and calls up a box with the county’s name, overall CCVI score, and population (above, right-hand side).

Based on feedback from users, we added a search bar so that counties could quickly be found within the graph (since some states have 50 counties or more), and a feature to allow people to compare counties across any two states.

2. How are Americans social distancing?

We also created a dynamic, daily changing illustration of the reported COVID-19 infection rate for each county, superimposed on a map of the U.S. The next challenge was to combine the CCVI with a representation of social distancing.

We had analyzed mobile-phone mobility data from Unacast that indicates the extent to which people in each county were moving around compared with a baseline of before the epidemic. But we wondered: how could we tell a nuanced story that reflected the vulnerability of each county, the number of cases it was experiencing, and its degree of social distancing — and show trends in how all this was changing over time?

We came up with a bubble graph that’s animated to show the change in data day by day.

In the bubble graph, each U.S. county is represented by a bubble, with the more vulnerable counties appearing towards the top of the graph. Bubbles increase in size as reported COVID-19 cases grow. As social distancing increases, a county moves towards the right of the graph — or back to the left as it reduces or abandons social distancing.

The animation easily conveys how cases have grown, and how social distancing tended to decrease from late April onwards. Different colors communicate other significant dimensions of the story: the differences among regions, such as the greater number of cases in some counties of the Northeast, and the higher proportion of vulnerable counties — and tendency for less social distancing — in the South.

Moving beyond today’s crisis

The COVID-19 pandemic illustrates the importance of communicating complex information in visual formats. Data visualization experts, journalists and others have written insightfully about this challenge. Done well, data visualization can be a tool to improve policy, empower individuals, and positively affect health outcomes. We’re still learning as we go — and the learning won’t stop when the crisis finally comes to an end.

Our combined experience as researchers, analysts, designers, and communicators has ramifications for the stories we can tell about other health and social issues, such as chronic disease, climate change, poverty, racial discrimination, and policing. The urgency of these stories, and their intersections, has become all too apparent in recent years — and even more so in recent months. With our growing knowledge comes the responsibility to visualize a better future.

This work was made possible by everyone at the Surgo Foundation, including but not limited to (in alphabetical order): James Baer, Bethany Hardy, Vincent Huang, Rahul Joseph, Hannah Kemp, Sema Sgaier, Peter Smittenaar, Nick Stewart, and Staci Sutermaster.