Wise Decision Making in a Pandemic Requires Data Maturity

Joy Eakins and Chris Wyant Cornerstone Data share their data analytics with a client.

Monday’s alarmist tweet by Major Brandon Whipple of Wichita, KS stated that positive rates in Sedgwick County had a 3% bump – he implied it was due to Labor Day and admonished citizens to be careful.

COVID testing chart. Snapshot of Mayor Brandon Whipple's tweet in COVID 2020.

But the context behind the number is important – and one that Mayor Whipple, City Council, County Commissioners, and School Board members should understand before tweeting, retweeting, or making decisions that affect residents, businesses, and school children.
Each week, Sedgwick County updates their data with the number of positive and negative tests per day. The percent positive rate is calculated and then an average of the last 14 days is taken to come up with this figure. For the 14 days in question in his tweet, the data shown was:


If you average the percent positives for these 14 days, you will get the 7.8% shown in Whipple’s tweet


Notice the percent positive on that last day is 44.4% – but only 9 tests are reported on September 20. That small outlier day with only 9 tests causes the bump of 3 percent.

Sedgwick County is using standard definitions for rolling averages to calculate this percent positive. The problem with using rolling average in this dashboard is that each day in the 14-day period “weighs” the same – even when all tests results are not back. An alternative method to describe percent positives would be to take the sum of all positives over 14 days and divide by all the tests taken in that same 14 days. This approach would weigh every test equally instead of weighing each day equally. Using this method, the result would have been 4.9% positive.



Another issue that all our local leaders need to understand is something we have noted since the beginning of the pandemic – the report of negative test results lags positive test results. Sedgwick County Health Department officials have addressed this in multiple Commission meetings. With limited resources, their goal is to notify positive citizens first to limit spread and perform contract tracing. But this means that it takes longer to post the negative test results for the public.


Here is an example we noted to county commissioners in July.


We pulled the positive and negative data on July 20th and July 28th for test dates from June 25 to July 7.


You can see for yourself that the negative test results were entered much later and at higher rates than the positive test results.


In this example, for the period up to 33 days prior, 206 positive results were added and 1,596 negative results were added.

The County has done a good job communicating this information to Commissioners over the last 2 months.  But perhaps they can find a way to incorporate this context into the dashboard to educate citizens and other leaders in the community.  Along with descriptions of the context, the County might consider showing all but the last 2 or 3 days of data until a certain minimum threshold of results is received.  


Regardless of whether the County changes their dashboard, elected leaders should show restraint.  When wise leaders see a major change in a trendline, they stop to ask questions about what caused it before taking to Twitter.  We ask our government officials to be wise.



UPDATE: As of September 22, 2020, the test results on the Sedgwick County dashboard show that on 9/20 there were 16 positives and 237 negatives, making a 14-day rolling average of 5.16%. This is a decrease of 2.6% in the percent positive calculation overnight.