I get annoyed with the way COVID-19 data is reported. There is a lot of emphasis put on the number today and on the aggregate (total cases, total deaths, etc.) These make good headline numbers but don’t do much to help people understand the current trend – are we getting better or worse.

So I built a report using the COVID-19 Tracking Project at The Atlantic’s data feed and Power BI for visualization. You can access the dashboard at bit.ly/NNCOVID19DB. The data is automatically updated several times daily.

There are definitely prettier dashboards out there, including on the COVID-19 Tracking Project website, but none that I see show two moving averages to get rate of change visualizations. That’s a problem.

In my view what really matters is whether we are seeing upward or downward trends. Thus I wanted to see – and suggest more people look at – moving average (MA) comparisons. You want to see the shorter moving average trending below the longer moving average to indicate a downward trend (which in the case of this dashboard is the good direction.) And, vice versa, when the shorter is above the longer it indicates an uptrend (i.e., hospitalizations, cases, or deaths are increasing.)

Additionally, I’m a big fan of looking at the hospitalization trend to see if the problem is getting better or worse. It eliminates lots of side-discussion about whether or not COVID-19 infections are among the young and/or how we are protecting the elderly and vulnerable populations. Ultimately, if you end up hospitalized for a respiratory illness that’s bad, so looking at COVID-19 hospitalization provides the secondary data point (after cases) to see how we are doing.

Back to trends: they really matter a lot more than top line numbers. The top line numbers shift and a single moving average is either too responsive to change or too slow to change, making it hard to visualize if a measure is accelerating or decelerating.

Lets look at an example to understand the value and importance of this data.

Comparing COVID-19 Moving Average Rate of Change in Slope for Trend Analysis

Above (#1) we see an increasing (bad) trend. Throughout the entire period shown the MAs are going up. But, if you were looking at just one daily MA (DMA) trend line you would miss the important change from 1 to 2. In 1 the slope of the yellow line (7 DMA) is increasing faster than the blue line (21 DMA) indicating not just a growing problem but a problem that is escalating as it grows faster each day. Hence, we are failing to manage the growth of the problem, let alone slowing the problem.

Then in #2 the trend is continuing upwards (bad) but there is a change in the 7 DMA slope relative to the 21 DMA slope, as they begin to run parallel. This indicates that we are seeing progress towards a solution. Generally, in absence of other information, this would indicate that the actions being taken are working and should be continued and (possibly) accelerated.

In this case, ultimately, we saw the 7 DMA crest and eventually turn down crossing below the 21 DMA, indicating the problem was getting better.

The short term 7 DMA cresting and crossing below the long term 21 DMA.

Lets look at another example of using DMA slopes to see the value of this analysis.

In # 1 we see a steadily improving situation, both lines sloping down in parallel. This indicates the problem is improving but nothing we are doing is improving the problem faster. For example, if we implemented certain closures to get a problem under control and then increased the closures (hoping to speed up improvement) then # 1 would indicate the additional closures were not delivering benefits.

Then in # 2 we see the shorter term (7 DMA) line bottom out and turn slightly upwards. Ultimately the two trends merge together (# 3) indicating we are now in stasis, able to keep the problem under control but not reduce it further.

Using Two DMA lines to Visualize Remediation Impact

My goal was to encourage people to look beyond simple data points and take the time to understand the trends. Numbers on their own can be misleading. Even graphs of a single data trend can lead to erroneous conclusions. Comparative rates of change are ultimately the best way to visualize and understand a problem.

Feedback is welcome in the comments.

Again, the dashboard Link: bit.ly/NNCOVID19DB