There have been many reporting formats for interpreting data during the Pandemic.  But how many methods have helped healthcare leaders and managers predict with a high degree of belief how COVID-19 would be affecting their operations in the next 15, 30 or 60 days?  A great deal of effort has been done to collect data.  But the analysis of the data has been lacking from an operational perspective.

You may have noticed that the data is being reported in data tables or maps generally with cumulative numbers.  Cumulative data “smooths” the variation, but only tells us a picture in time where we have been, much like driving a car looking at the rearview mirror, a recipe for disaster.  Histograms and talks of “flattening the curve” cause people to make predictions with a wide range of assumptions, little theory, and lots of opinions.  When the numbers go up and down, people tend to over-react or under-react. Shewhart called these Mistake 1 & Mistake 2.  Over-reacting during a pandemic can naturally cause hysteria, or fear of making a critical mistake.  

Walter Shewhart in 1924, devised a method to use data, over time (and not cumulative data) to see the variation and not assume any statistical model.  Shewhart’s charts allow the data to create its own path, defined by upper and lower control limits.  The Shewhart chart (sometimes called a control chart) was born, with the first statistical signal (when the data jumped out of the path).

 Over the years, his theory has been updated and expanded upon for understanding different types of data, rare events, trends and now the exponential model, which fits the COVID-19 Pandemic behaviors.  Statistical signals have been complicated until API (a group of statisticians who worked with Dr. W. Edwards Deming, authors of The Improvement Guide and the Model for Improvement), standardized five basic patterns in probability (see Table 2/3 THOR).


Recently, in a ground-breaking article published in US World Report, our API partner, Lloyd Provost et al took COVID-19 death statistics and created a new Exponential Shewhart chart1 .  This chart used statistical theory based on Log 10 data transformation, to model the behavior of a contagious disease such as COVID-19.  Using the death statistics, we could now predict the trajectory of the disease without intervention and impact of intervention.

Using this same theory-based model, we have expanded the usefulness of this tool and answered questions such as:

· What percent of the tests are positive?

· Can we predict the trajectory of positive tests?

· Can the model help predict hospitalizations and bed occupancy needs and their trajectory?

· Can we predict the impact of hospitalizations and our capabilities in the ICU?

· Can we see the trajectory of ventilator use and determine how many we might need with confidence?

When we are not able to take care of our more serious patients, can it help us predict the number of patients we will need to transfer so that we can plan and partner with others for their care?

How have we done this?  By applying the algorithm to daily positive tests, daily bed occupancy, daily ICU bed occupancy, daily ventilator usage and finally transfers of care.  Early, we started with a Shewhart chart.  



Once we had either a run of eight points above the mean, 2 out of 3 consecutive points near the upper control limit, or points outside the upper control limit, we transferred the same data into the exponential control chart to quantify the rate of increase or trajectory.


By adjusting the days, generally to the next 14 days, we used the trajectory of the chart to predict needs, assess capability and, understand when the special causes signal temporary or sustained improvement.


Once a sustained special cause signal is identified, the exponential chart signals the trajectory has changed.  Aimi allows us to use the phasing capability tools to understand decreasing rates by calculating the new trajectory.