How to make a covidiot-proof COVID-19 dashboard
While we can’t all be Johns Hopkins University, there are a few guidelines that any data analytics team can follow
By now everyone has seen the gold standard JHU coronavirus tracker either directly online, or indirectly by news reports of the latest number of global positive cases and deaths.
In many ways it is the perfect dashboard — the data visualizations don’t get any fancier than your standard map, line chart, and badge (i.e. a box with just the variable name and a single number in huge font). It isn’t trying to predict anything — just the facts, ma’am. There aren’t a whole lot of colors. It is updated multiple times per day and renders fairly quickly given the massive amount of data that goes into it. Really, it doesn’t get much better than this, folks. No wonder it is the most famous, most used dashboard ever in the history of dashboards.
However, that doesn’t mean there’s no room for other teams of data analysts to build a COVID-19 dashboard for their own organizations. Healthcare systems and other research institutions around the world are also in need of a simple-to-digest way to monitor various metrics used by COVID-19 command centers to manage this crisis. We here at Pie and Donut Analytics have helped build and have seen our fair share of COVID-19 dashboards in the US; thus we can elaborate on the kinds of things stakeholders want to see and how they are being displayed.
Other useful things to know besides data on coronavirus testing
Of course the #1 thing any COVID-19 dashboard worth its salt must include is the answer to the questions (1) How many SARS CoV-2 PCR tests did we do? (2) How many of those are positive? Something like the below badges and bar graphs are usually front and center of a typical COVID-19 command center dashboard:
But there are a plethora of other measures that hospitals need to get a handle on, and it’s not just clinical related things like ventilator usage and ICU bed capacity. For example, you can see below that one healthcare system is interested in “% of calls abandoned and average call speed for calls into COVID info line.”
We would venture to say that the two other high profile aspects to the pandemic besides testing are PPE and pharmacological treatments. N95 and hydroxycholoroquine are now part of everyone’s vocabulary. Below are two components of a healthcare system’s COVID-19 dashboard that monitor hospital supplies and drug utilization. Note that these are simply displayed as rows and columns, no elaborate data viz to be found. Sometimes all people really want to see are just straight numbers, not pretty bubble charts.
While deceptively plain-looking, the below dashboard component of a university hospital system is a clinic (pardon the pun) in dashboard best practices. It packs a punch in such a small amount of real estate — showing the number of PUI (persons under investigation) and breaks that down into patients in either the ICU, emergency department, or “floor” (i.e. not in the ICU or ED, here called “GC” presumably General Care). It also shows how many patients have been discharged from the hospital. The component reports these numbers for each of the hospital’s facilities, and then for each facility you can see the breakdown by hospital unit. For example, Facility #1 has 5 PUI of which 3 are in the ED, 1 is in ICU unit A, and 1 is in ICU unit B.
Achtung! Hospital Marketing and HR Departments
Where once healthcare was traditionally thought of as a recession-proof industry, this pandemic has spared no one, not even people who work in hospitals. Many healthcare workers are being laid off or furloughed due to the cancellation of elective procedures and surgeries to free up beds in anticipation of a surge in COVID-19 cases (which in many cases never materialized). Surgeries are huge revenue generators for hospitals; thus major hospital systems serving patients from multiple states usually have a marketing department that does outreach to physician groups in the surrounding region. A preeminent research hospital in another state may be some patients’ only chance to get an organ transplant or experimental treatment, for example.
With such an infrastructure already in place well before the pandemic’s arrival, one healthcare system’s marketing department had the ability to rapidly pivot and switch their focus to providing education on the topic of COVID-19 to providers in the four-state area they serve. Below are a stacked bar trend graph and a map of states in the Mountain time zone. It is very clear to see that this healthcare system’s COVID-19 town halls have been getting a fair amount (although understandably declining as the pandemic wears on) of registrants, with a decent percentage (~10%) from outside of the most populous state Colorado.
Some Good News
This last COVID-19 dashboard component pulls the testing and the HR aspects of the pandemic together. There are several noteworthy things to point out on this one. First, it is very rare for a hospital to be so forthcoming as to report on furloughs and positive cases among employees on a dashboard; this topic is usually very closely guarded. Also, it is one thing to report on testing for a hospital’s patients; quite another for a hospital’s employees since they wouldn’t have necessarily gotten tested at the same hospital where they work. So, it is not as straightforward to obtain this necessary data. Lastly, you can infer from the downward trend in furloughed employees and the upward trend in employees who returned to work that this hospital system is committed to rehiring workers who had been displaced. Kudos all around for this hospital and their great dashboard!
Credit where credit is due
Finally, let’s come back to the granddaddy of them all — the JHU coronavirus tracker. Recently there was an interview with Lauren Gardner, the head of the team who built it. It is very interesting and accessible even if you are not at all data savvy, but we think it should be required reading for leadership at companies with data analytics departments. If anyone thinks that a great dashboard which people actually use is something that just one person can put together from start to finish after a crash course in Tableau, they would be sorely mistaken.
Not all healthcare workers wear scrubs — there are a lot of dedicated folks employed in the industry whose work directly supports those on the front lines keeping us safe from this devastating pandemic. Healthcare data analysts especially those who are involved in creating these much needed dashboards for COVID-19 command centers included!
