​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Using Shewhart Charts to Understand Variation in COVID-19 Reported Deaths​

​IHI and our colleagues have developed a COVID-19 Data Tracker application showing the reported deaths from COVID-19 globally using Shewhart charts.

Because the Shewhart charts, a cornerstone of improvement methods, are designed to distinguish between common cause (random) and special cause (non-random) variation, they are ideally suited to help decision makers understand if the reported number of daily deaths is predictable within limits or unpredictable (i.e., special cause variation).​

The IHI COVID-19 Data Tracker has Shewhart charts for​: 1) individual countries and 2) US states and territories. 

Shewhart charts are updated daily for each country and each US state and territory. The model tracking the trajectory of COVID-19 reported deaths detects when the number of deaths by geographic location is entering ​rapid exponential growth (Epoch 2) and when exponential growth has ended. We believe that the Shewhart chart approach will prove to be useful to subject matter experts, decision makers, care providers, and the public as they address COVID-19 challenges.

Instructions​

• View the detailed instructions for the COVID-19 D​ata Tracker to learn how the Data Tracker is organized, key terms used to display the data, and gu​idance on how to navigate the various screens.
• It is important to understand the key terms used to create and interpret the Shewhart charts​:

   

• Epochs: Pre-Exponential Growth (Epoch 1); Exponenti​al Growth (Epoch 2); Post-Exponential Growth with flat trajectory or exponential decline (Epoch 3); Stability after descent (Epoch 4)
• Extensions: When a phase is “active” the upper and lower control limits and center line are extended as reference markers for 10 days
• Adjusted versus Unadjusted Data: The raw data (unadjusted) reflect the number of deaths submitted by numerous sources tracking COVID-19 deaths, which include “spikes” or extreme data points that do not typify the general historical patterns in the reported data; the Adjusted data attempt to create a more stable stream of data from the various data sources
• Ghosting unusually large data values: Each daily record is examined to assess if the record is unusually large compared to days preceding and following the record; on the Shewhart charts these unusually large data points are referred to as being “ghosted”
• Use the Full Screen mode (click arrow icon, bottom right) to see the charts in a full browser window.
  

Charts for Individual Countries​: COVID-19 Reported Deaths by Country

​ Select Country: On the far left, click ​the circle to the left of the country name. (Use Full Sc​reen mode to view larger.)

• Countries are listed in alphabetical order on the left side. Note that the screen opens with the United States ​as a default.
• The data shown on the chart are Adjusted Data. To view the chart with Unadjusted Data, click on the box in the lower left corner labeled “Click for Unadjusted Daily Deaths.”
• You can also view the Unadjusted number of reported deaths by clicking on the “2” in the label at the bottom center of the chart that looks like this: <1 of 2>​
• To view a specific date range on a chart, use the “s​lider” (located along the bottom of the chart) to select the date range you wish to review.​
• Download detailed instructions