Today was a big day for a number of reasons.  First, the daily reported new lab confirmed infections were 611 – beating the previous day’s record by 136 – a huge increase.  Seems like terrible news at first glance, yet at the same time those numbers were being released Governor DeWine announced that our state will begin to reopen on May 1st.  This is huge!  This represents a massive shift in his rhetoric from the day before.  While I still believe we can open safely sooner than May 1st, I want to celebrate this enormous shift in our direction.  The support that this effort and few others across the state has given to allies in the legislature is having a truly significant impact.  Dr. Bulas and I, along with the Ohioans that are out of work, missing their loved ones and missing their favorite activities thank you.

Now let’s get back to the data.  When yesterdays’ numbers set a surprising record, I said don’t panic.  First, with any data set, you shouldn’t get overly worked up over any single point.  The real world is messy and data describing the real world is noisy.  This is why I keep focusing on the trend in one of our data sets – the trend in new infections based on date of onset (the data previously referred to this as date of first infection).  When someone gets sick with COVID-19, we know based on their age that there is a certain probability that they will need hospital care or potentially die.  The trends in hospitalizations and deaths will follow from the trend in infections.  While we ultimately care more about the latter data points, the trend in cases better tells us the direction we are heading on any particular day.

To understand why I’m not panicking over the recent uptick, you have to understand how the data is reported.  The 611 cases reported on April 16 were not all tested or infected on April 16.  Various hospitals and medical providers are running tests with different lab providers that take varying times to return a result.  These providers in turn now have to figure out date of onset and then report this data back to the state authorities who then have to import the data into their reporting tool.  Some are very fast at getting this done, others aren’t.  Practically, this means that those 611 cases are actually spread out over several previous days.  In fact, the earliest case in that total actually had an onset date of March 1st.  While it isn’t unusual to see a few anomalies like that, I’ve found that each daily release follows a fairly predictable pattern.  Let’s break it down:

Date of Onset Cases Reported 4/16 Cases Reported 4/15
4/16 12 n/a
4/15 102 13
4/14 273 87
4/13 50 150
4/12 29 95
4/11 10 26
4/10 25 23
4/9 19 10
4/8 4 2

 

The chart shows the last two days of reported cases.  Notice how in each case, the vast majority of the cases reported came over a 3-day period with the being being 2 days before.  While we have other cases going back before 4/8 that ultimately add up to the total, the consistency of the pattern of having most cases falling into a predictable 3-day period is important.  For the 16 days that I’ve been analyzing the data, I’ve calculated the trend line based on two weeks of cases prior to that 3-day period.  That’s led to a remarkably stable analysis with strong predictive power.

Today’s graph is starting to flatten out based on the recent upwards trend and tomorrow’s version may even go slightly positive.  That’s what I would expect based on the last two days and I’m not happy about it.

The important questions then becomes why is this occurring and should we expect it to continue?  How does this affect our ability to Restart Ohio safely and quickly?

I dug even deeper and found a few things that explain the uptick in new cases that imply that we (probably) haven’t started a trend for the worse.

  1. Testing availability has increased.  We’ve had the most daily tests the last two days and I’ve heard anecdotally that we may be starting to report on some of the serological antibody testing.  I’m working to get confirmation if that’s true, but changes to testing availability and methodology means we’ll definitely have to look at things differently going forward.  I’ll continue to monitor this going forward.
  2. Several counties had large totals that were very abnormal based on their history.  This could be related to new testing, it could also be representative of a localized outbreak or a combination of both.  If you have any info on these, please reach out to me on Facebook.
    1. Marion County – 201 cases over a 4-day stretch, 99 of which were in yesterday’s total while their previous high was 9.  This is definitely related to an outbreak at the Marion Correctional Institution.  This is an awful situation, but it should end up being a blip in the overall trend.
    2. Lucas County (Toledo) – 76 cases over a 2-day stretch with a previous high of 30 and out of line with their recent downward trend.  Could be a lot of things.
    3. Pickaway County (Circleville) – 103 cases over a 4-day stretch with a previous high of 13.  Likely another prison outbreak.
    4. Franklin County (Columbus) – 67 cases on April 14 with a previous high of 44 and out of line with their current trend

All in all, a significant portion of the recent bump are two prison outbreaks that should run their course and shouldn’t affect the long term trend.   I’m looking forward to seeing what tomorrow’s numbers bring.