Thread by @BrettKelman

Tennessee’s coronavirus stats went wonky today because both the Nashville and state governments separately changed how they present data. I know many of you attempt to follow this closely, so I’m going to try to explain it all. This is going to get weedy.

.@TNDeptofHealth made two major changes. First, they revised how they define “recovered” coronavirus patients, resulting in a big reduction in the count of active infections in every county. Second, they corrected about 1,700 cases that were listed in the wrong county. Whoopsie.

There is no one standard for what it means to be “recovered” from coronavirus. Previously, Tennessee classified you as recovered in one of two ways:

1. Infected people should have daily convos with contact tracers, who deem them recovered when symptoms are safely over.

2. But, if you can’t or won’t talk to contact tracers, the state automatically marked you as recovered after 21 days (unless you died.) This threshold was needed so people who don't answer their phone aren't classified as “active infections” indefinitely. It's a failsafe.

Today, the state reduced that failsafe threshold from 21 days to 14 days. Officials said they did this because growing knowledge of the virus shows infections don’t last as long as we once thought.

The impact of this change is lots of people who were listed as active infections on Wednesday are no longer considered active. As a result, statewide active infections dropped from 35,359 to 16,163.

This is NOT because the outbreak shrunk.

We are measuring on a new scale.

The state is also dropping the label of “recovered” people in favor of “recovered/inactive.” Officials say this new label is more representative of people whose infections have run their course but still could suffer complications. Seems fair.

Tennessee’s coronavirus stats also shifted because the state corrected 1,700ish infections attributed to incorrect counties. Generally, these cases came from zip codes that straddle two counties and were placed in the wrong one.

If your county lost cases today, this is why.

Considering all of that, here is the latest: Tennessee gained 1,715 coronavirus infections today on 26,633 tests; daily positivity rate of 7.6%. Average hospitalizations per day steady at 64. Eighteen deaths lift the statewide death toll to 1,815. Active infections now at 16,163.

Chapter 2: Nashville also changed how it reports coronavirus data.

The city is now using a revised, more transparent method to calculate positivity rate, a key virus metric. The new method is the same one used by the state government, the White House and myself.

The method divides the daily number of positive tests (not cases, tests!) by the daily number of total tests. This accounts for people who are tested multiple times in the numerator and denominator, providing a true accounting of test positivity, not people positivity.

This method is widely used, but not without flaws.

For example, two people get tested. Person A is negative. Person B is positive. B later gets tested again, hoping to be cleared for work. He is negative. Half of the people had the virus, but only one third of the tests did.

This is an exaggerated example that would be largely smoothed out over thousands and thousands of tests. But, as I understand it, this method may slightly underrepresent the positivity of people while accurately representing the positivity of tests.

Prior to now, Nashville’s positivity rate was confusing because it was built from stats that weren’t public. Instead of dividing by daily cases or tests, the city relied on a daily tally of new “investigations,” or entries, in the National Electronic Disease Surveillance System.

This database is used by health officials, but I can't see it. I’m told by officials the tally in this system is the equivalent of deduplicated tests and therefore basically the same as dividing daily positive cases (cases, not tests!) by all daily tests. I can’t verify this.

Fun fact: People call this database “NBS.” That stands for “NEDSS Base system," which in turn stands for National Electronic Disease Surveillance System Base System. "

It’s an acronym inside of an acronym that includes the word "system" twice.

I hate it.

Because Nashville switched this methodology, the city also disclosed lots of duplicate tests that weren’t previously in their public data. (People tested for a 2nd or 3rd time.) So, it may appear Nashville gained 79,000 tests today. It did not. Most of those are old.

For all you armchair epidemiologists out there, the good news is that Nashville officials and state officials now calculate positivity rate in the same way. The bad news is that their numbers still won’t line up. Why?

First, the state releases a positivity rate for each individual day but the city just publishes a weekly average rate. I hope to start calculating a daily rate for the city soon, but I’m not fully confident I can do it quite yet. (I’m figuring a great deal of this out as I go.)

Second – and this is a big one – the city and state organize data on different schedules. The state counts test results based on the day they are received. The city backdates those results to the day of the test. I don’t know which method is better, but they are clearly different

In closing, there were lots of changes in the virus data today. Hopefully, some of it will make more sense now. If it still doesn't make sense, maybe this thread will help explain why. Also ...

I should have done my math homework in high school.