The story of the last few days has been a sudden rise in new cases reported in Ohio.  Yesterday’s (4/18) total of 1081 lab confirmed cases blew away the previous record.  Even as tracking COVID-19 by onset date spreads out the new cases to some extent, we now have a very sharp positive trend showing up in the raw data.

In fact, this is the strongest positive trend we’ve ever seen including back in mid-March before the shutdown orders went into effect.  I’ve heard a lot of speculation that this is a way to get the cases to actually match the models.  DeWine and Acton burned a ton of credibility earlier in the month continuing to assert that a surge in new cases was coming that never materialized, so I understand the mistrust.  Even so, I don’t think this is a way to drive the numbers up for a few reasons.  First, I sat on a conference call Friday morning with the OSU professors who created the model and had a chance to ask questions.  I found them to fully understand the data and that their conclusions were very much in alignment with mine.  They also shared a graph of their updated model that showed a peak of about 420 new cases/day on 4/18.  I think that number was quite realistic.

  • The original models were always based on onset date of the illness (not reported date) and assumed a consistent testing protocol (we would test only the sickest)
  • Their explanation for overestimating cases per day was that they underestimated the impact of individuals social distancing on their own

On 4/11, the Ohio Department of Corrections began an effort to test all inmates in most of our institutions based on evidence of outbreaks beginning in a few institutions.  This means that for this population, we changed protocols and we can’t compare the previously reported numbers to these.  Unlike our efforts in the rest of the state where we had relatively high death and hospitalization rates due to limitations on only testing the sick, we began testing the 49,000 or so inmates where only a handful actually had symptoms.  As we add all these confirmed cases without symptoms, look what it does to our overall hospitalization and death rates:

Date Hospitalization Rate Death Rate
April 11 30.3% 5.4%
April 12 29.8% 5.3%
April 13 29.1% 5.2%
April 14 28.0% 5.0%
April 15 27.2% 4.9%
April 16 24.9% 4.5%
April 17 24.6% 4.4%

Did the deadliness of the disease actually decrease by 18.5% (1/5.4) in a week?  Of course not, we just started testing people that were less sick than before.  Once this effort finishes into early next week, we should go back to the usual protocol of only testing the sickest and our numbers will be comparable again.

So how big of an impact is this?  The answer is a lot.  According to the Department of Correction report, 1659 inmates and staff are confirmed positive.  Some of them were positive prior to April 11, but given that Marion and Pickaway counties, home to the largest outbreaks have had 1099 positive cases since April 11 and would have been expected to have 50 or so cases based on their trends otherwise, we can assume that the vast majority of these are the inmates.  I took the time to adjust the data to pull out these positive tests that come from the changed methodology.  Let’s look at our graph on onset date again:

The yellow line is our new series.  Notice how our trendline is flat over two weeks versus the strong positive trend in the raw numbers.  This is why we have to dig into the data and understand it in order to drive our decision making.  DeWine and Acton, as long as they listen to their advisors, will understand that the prison outbreak/testing should not affect our ability to restart the state.